Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

On the influence of effective stress and micro-structure on suffusion and suffosion Slangen, Paul Herman Henricus 2015

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
24-ubc_2015_september_slangen_paul.pdf [ 69.97MB ]
Metadata
JSON: 24-1.0166340.json
JSON-LD: 24-1.0166340-ld.json
RDF/XML (Pretty): 24-1.0166340-rdf.xml
RDF/JSON: 24-1.0166340-rdf.json
Turtle: 24-1.0166340-turtle.txt
N-Triples: 24-1.0166340-rdf-ntriples.txt
Original Record: 24-1.0166340-source.json
Full Text
24-1.0166340-fulltext.txt
Citation
24-1.0166340.ris

Full Text

On the influence of effective stress and micro-structure on suffusionand suffosionbyPaul Herman Henricus SlangenB.Sc., Delft University of Technology, 2006M.Sc., Delft University of Technology, 2008A THESIS SUBMITTED IN PARTIAL FULFILLMENTOF THE REQUIREMENTS FOR THE DEGREE OFDoctor of PhilosophyinTHE FACULTY OF GRADUATE AND POSTDOCTORAL STUDIES(Civil Engineering)The University Of British Columbia(Vancouver)June 2015c© Paul Herman Henricus Slangen, 2015AbstractThe research presented within this thesis covers the development of a flexible wall permeame-ter, and a parametric laboratory investigation of the factors influencing seepage-induced internalinstability in gap-graded granular materials. A flexible wall permeameter comprising a double-walled triaxial cell, a seepage control system, and instrumentation, with a novel measurementof volume change, has been designed and built. The apparatus was successfully commissionedand the test procedure demonstrated to yield repeatable results. Two commissioning tests, 23tests on eight glass beads gradations and 16 tests on ten soil gradations were conducted. Allgap-graded gradations were reconstituted using the modified slurry deposition method, isotrop-ically consolidated to a cell pressure between 50 and 150 kPa, and subsequently subject toupward multi-stage seepage flow.Analysis of the test results identifies two distinct seepage-induced internal instability phe-nomena. First, migration of fine particles from a soil, termed suffusion, is characterised by amass loss without change in volume, or with a small non-progressive change in volume, ac-companied by an increase of hydraulic conductivity. Second, local or overall collapse of thesoil structure, termed suffosion, is characterised by a seepage-induced mass loss, accompa-nied by a reduction in volume and a change in hydraulic conductivity. It is demonstrated thatmeasurement of total volume change is necessary to avoid any mis-interpretation of the phe-nomenological response to seepage flow.It was found that the differential pore water pressure at the onset of suffosion increaseswith increasing mean effective stress. The micro-structure of the specimen was found to influ-ence the susceptibility to seepage-induced internal instability: the portion of non-load bearingfine particles appears a useful parameter to quantify the potential for suffusion, whereas theproposed state parameters are predictors of the relative susceptibility to suffosion. Althoughparticle shape does not affect the suffusive response in a transitional clast-supported micro-structure, sub-angular particles are found to yield a transitional micro-structure that is moreresistant to suffosion than a similar micro-structure of spherical particles. A unified approachis presented to characterise suffosion.iiPrefaceChapter 2: Versions of Sections 2.1 and 2.4.3 have been published as follows:• Fannin, R.J., and Slangen, P. 2014. On the distinct phenomena of suffusion and suffosion.Geotechnique Letters 4(4): 289 - 294.I was the lead author and lead investigator of the literature review for the above manuscript.R.J. Fannin was responsible for the conceptual formation, most notably of Fig. 2.5.The remaining parts of this dissertation are original, unpublished, independent work of theauthor, P. Slangen.iiiTable of ContentsAbstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iiPreface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iiiTable of Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ivList of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ixList of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiList of Acronyms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xviiiGlossary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xixAcknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xxiv1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.1 Research hypotheses . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31.2 Scope and objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41.3 Outline of the thesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42 Literature review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 62.1 Seepage-induced internal instability phenomena . . . . . . . . . . . . . . . . . 62.1.1 Experimental evidence of internal instability phenomena . . . . . . . . 72.1.2 Conceptual framework . . . . . . . . . . . . . . . . . . . . . . . . . . 82.1.3 Concluding remarks . . . . . . . . . . . . . . . . . . . . . . . . . . . 92.2 Micro-structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92.2.1 Micro-structures of gap-graded mixtures . . . . . . . . . . . . . . . . . 92.2.2 Particle shape . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132.3 Seepage flow in granular materials . . . . . . . . . . . . . . . . . . . . . . . . 142.4 Experimental investigations of seepage-induced internal instability . . . . . . . 172.4.1 Rigid-wall permeameters . . . . . . . . . . . . . . . . . . . . . . . . . 17iv2.4.2 Flexible wall permeameters . . . . . . . . . . . . . . . . . . . . . . . 172.4.3 Variables to quantify seepage-induced internal instability . . . . . . . . 182.5 Factors governing seepage-induced internal instability . . . . . . . . . . . . . . 192.6 Reflection on research hypotheses . . . . . . . . . . . . . . . . . . . . . . . . 212.6.1 Hypothesis No. 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212.6.2 Hypothesis No. 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 212.6.3 Hypothesis No. 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222.6.4 Hypothesis No. 4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222.7 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 223 Apparatus, materials and testing program . . . . . . . . . . . . . . . . . . . . . . 383.1 Flexible wall permeameter . . . . . . . . . . . . . . . . . . . . . . . . . . . . 383.1.1 Double-walled triaxial cell . . . . . . . . . . . . . . . . . . . . . . . . 383.1.2 Membrane, base pedestal and top cap . . . . . . . . . . . . . . . . . . 393.1.3 Seepage control system . . . . . . . . . . . . . . . . . . . . . . . . . . 403.1.4 Instrumentation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 413.2 Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 433.2.1 Glass beads . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 443.2.2 Soils . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 453.3 Test procedure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 453.3.1 Specimen preparation . . . . . . . . . . . . . . . . . . . . . . . . . . . 453.3.2 Stage 1: Consolidation . . . . . . . . . . . . . . . . . . . . . . . . . . 463.3.3 Stage 2: Multi-stage seepage flow . . . . . . . . . . . . . . . . . . . . 463.4 Test program . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 473.5 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 484 Results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 674.1 Commissioning tests . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 674.1.1 Test GB-F-100 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 674.1.2 Test 6.5GB25-100 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 684.1.3 Findings of the commissioning tests . . . . . . . . . . . . . . . . . . . 684.2 Tests on glass beads . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 694.2.1 Gradation 3.3GB20 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 694.2.2 Gradation 4.8GB20 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 704.2.3 Gradation 4.8GB35 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 714.2.4 Gradation 6.0GB20 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 734.2.5 Gradation 6.0GB25 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 754.2.6 Gradation 6.0GB30 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 764.2.7 Gradation 6.0GB35 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 78v4.2.8 Gradation 6.5GB35 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 804.3 Tests on soils . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 824.3.1 Gradation 5.1BT20 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 824.3.2 Gradation 5.7BT20 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 834.3.3 Gradation 5.7BT35 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 844.3.4 Gradation 7.0BT20 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 844.3.5 Gradation 7.0BT35 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 854.3.6 Gradation 8.6BT20 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 864.3.7 Gradation 8.6BT35 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 864.3.8 Gradation 10.4BT25 . . . . . . . . . . . . . . . . . . . . . . . . . . . 874.3.9 Gradation 10.4BT30 . . . . . . . . . . . . . . . . . . . . . . . . . . . 884.3.10 Gradation 10.4BT35 . . . . . . . . . . . . . . . . . . . . . . . . . . . 884.4 Repeatability of the test results . . . . . . . . . . . . . . . . . . . . . . . . . . 904.5 Synthesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 905 Analysis of the test results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1185.1 Analysis of the seepage regime . . . . . . . . . . . . . . . . . . . . . . . . . . 1195.2 Analysis of micro-structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1205.2.1 Micro-structures of glass beads gradations . . . . . . . . . . . . . . . . 1205.2.2 Micro-structures of soil gradations . . . . . . . . . . . . . . . . . . . . 1255.3 Analysis of tests on glass beads . . . . . . . . . . . . . . . . . . . . . . . . . . 1255.3.1 Gradation GB-F . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1255.3.2 Gradation 3.3GB20 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1265.3.3 Gradation 4.8GB20 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1265.3.4 Gradation 4.8GB35 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1285.3.5 Gradation 6.0GB20 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1295.3.6 Gradation 6.0GB25 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1305.3.7 Gradation 6.0GB30 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1315.3.8 Gradation 6.0GB35 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1315.3.9 Gradation 6.5GB25 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1335.3.10 Gradation 6.5GB35 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1335.4 Analysis of tests on soils . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1345.4.1 Gradation 5.1BT20 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1345.4.2 Gradation 5.7BT20 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1355.4.3 Gradation 5.7BT35 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1355.4.4 Gradation 7.0BT20 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1365.4.5 Gradation 7.0BT35 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1365.4.6 Gradation 8.6BT20 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137vi5.4.7 Gradation 8.6BT35 . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1375.4.8 Gradation 10.4BT25 . . . . . . . . . . . . . . . . . . . . . . . . . . . 1375.4.9 Gradation 10.4BT30 . . . . . . . . . . . . . . . . . . . . . . . . . . . 1385.4.10 Gradation 10.4BT35 . . . . . . . . . . . . . . . . . . . . . . . . . . . 1385.5 Synthesis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1396 Factors governing suffusion and suffosion . . . . . . . . . . . . . . . . . . . . . . 1526.1 Compilation of experimental database on suffusion and suffosion . . . . . . . . 1536.2 Test of hypothesis No. 1 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1546.2.1 Glass beads tests of this study . . . . . . . . . . . . . . . . . . . . . . 1546.2.2 Glass beads tests of other studies . . . . . . . . . . . . . . . . . . . . . 1566.3 Test of hypothesis No. 2 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1576.3.1 Glass beads tests of this study . . . . . . . . . . . . . . . . . . . . . . 1576.3.2 Glass beads tests of other studies . . . . . . . . . . . . . . . . . . . . . 1586.4 Test of hypothesis No. 3 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1596.4.1 Glass beads tests of this study . . . . . . . . . . . . . . . . . . . . . . 1596.4.2 Glass beads tests of other studies . . . . . . . . . . . . . . . . . . . . . 1636.5 Test of hypothesis No. 4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1646.5.1 Glass beads and soil tests . . . . . . . . . . . . . . . . . . . . . . . . . 1646.6 Factors governing suffusion and suffosion . . . . . . . . . . . . . . . . . . . . 1676.6.1 Volume change . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1676.6.2 Effective stress . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1686.6.3 Micro-structure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1696.7 A unified approach for suffosion . . . . . . . . . . . . . . . . . . . . . . . . . 1716.8 Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1727 Conclusions, recommendations and implications for practice . . . . . . . . . . . 1927.1 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1927.1.1 Conclusions derived from the literature review . . . . . . . . . . . . . 1927.1.2 Conclusions concerning the test method . . . . . . . . . . . . . . . . . 1937.1.3 Conclusions derived from the analysis of test results . . . . . . . . . . 1947.1.4 Factors governing suffusion and suffosion . . . . . . . . . . . . . . . . 1957.2 Novel contributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1977.3 Recommendations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1987.4 Implications for practice . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 199Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 200viiA Measurement uncertainty . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 208A.1 Uncertainty in measured quantities . . . . . . . . . . . . . . . . . . . . . . . . 208A.2 Propagation of uncertainty . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209B Membrane compliance . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 213B.1 Measurement of membrane compliance . . . . . . . . . . . . . . . . . . . . . 213B.2 Effect of membrane compliance in this study . . . . . . . . . . . . . . . . . . 214C Variation of effective stress during multi-stage seepage flow . . . . . . . . . . . . 216C.1 On the scale effect of hydraulic gradient . . . . . . . . . . . . . . . . . . . . . 216D Forensic observations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219D.1 Visual observations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 219D.2 Post-test particle size analyses . . . . . . . . . . . . . . . . . . . . . . . . . . 237viiiList of TablesTable 2.1 Descriptions of seepage-induced internal instability phenomena (source: Fan-nin and Slangen (2014), with permission from ICE Publishing). . . . . . . . 24Table 2.2 Overview of selected studies on seepage-induced internal instability using aflexible wall permeameter. . . . . . . . . . . . . . . . . . . . . . . . . . . . 31Table 3.1 Characteristics of test materials. . . . . . . . . . . . . . . . . . . . . . . . . 50Table 3.2 Gradations of the test specimens. . . . . . . . . . . . . . . . . . . . . . . . 50Table 4.1 Initial test conditions: glass beads gradations. . . . . . . . . . . . . . . . . 93Table 4.2 End-of-test conditions: glass beads gradations. . . . . . . . . . . . . . . . . 94Table 4.3 Initial test conditions: soil gradations. . . . . . . . . . . . . . . . . . . . . . 95Table 4.4 End-of-test conditions: soil gradations. . . . . . . . . . . . . . . . . . . . . 96Table 5.1 Reference points in the micro-structure identification diagram for glass beadsand soil gradations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142Table 5.2 Micro-structure: glass beads test specimens. . . . . . . . . . . . . . . . . . 143Table 5.3 Seepage regime: tests on glass beads gradations. . . . . . . . . . . . . . . . 144Table 5.4 Responses: tests on glass beads gradations. . . . . . . . . . . . . . . . . . . 145Table 5.5 Micro-structure: soil test specimens. . . . . . . . . . . . . . . . . . . . . . 146Table 5.6 Seepage regime: tests on soil gradations. . . . . . . . . . . . . . . . . . . . 147Table 5.7 Responses: tests on soil gradations. . . . . . . . . . . . . . . . . . . . . . . 148Table 6.1 Experimental database on suffusion and suffosion compiled from the literature.174Table 6.2 Seepage-induced internal instability in tests on glass beads. . . . . . . . . . 175Table 6.3 Evaluation of portion of non-load bearing particles in gap-graded gradations. 176Table 6.4 Comparison of select tests on glass beads and soils. . . . . . . . . . . . . . 177Table A.1 Uncertainty in measured quantities (Part 1 of 2). . . . . . . . . . . . . . . . 210Table A.2 Uncertainty in measured quantities (Part 2 of 2). . . . . . . . . . . . . . . . 211Table A.3 Uncertainty in derived quantities. . . . . . . . . . . . . . . . . . . . . . . . 212ixTable B.1 Measurement of membrane compliance. . . . . . . . . . . . . . . . . . . . 215xList of FiguresFigure 2.1 Non-destructive seepage-induced internal instability in a test conducted bySkempton and Brogan (1994) (source: Fannin and Slangen (2014), withpermission from ICE Publishing). . . . . . . . . . . . . . . . . . . . . . . 32Figure 2.2 Destructive seepage-induced internal instability in a test conducted by Li(2008) (source: Fannin and Slangen (2014), with permission from ICEPublishing). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32Figure 2.3 Seepage-induced instability in a test conducted by Li (2008) (source: Fan-nin and Slangen (2014), with permission from ICE Publishing). . . . . . . 33Figure 2.4 Schematic illustration of seepage-induced instability phenomena: (a) suffu-sion; (b) suffosion; and (c) fluidisation (source: Fannin and Slangen (2014),with permission from ICE Publishing). . . . . . . . . . . . . . . . . . . . 34Figure 2.5 Conceptual framework for seepage-induced instability phenomena (source:Fannin and Slangen (2014), with permission from ICE Publishing). . . . . 35Figure 2.6 Inter-coarse and inter-fine void ratios (adapted from Thevanayagam, 1998,with permission from ASCE). . . . . . . . . . . . . . . . . . . . . . . . . 35Figure 2.7 Micro-structure cases (source: Thevanayagam et al., 2002, with permissionfrom ASCE). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 36Figure 2.8 Definitions of particle size and shape (source: Altufi et al., 2013, with per-mission from ASCE). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37Figure 2.9 Plot of sphericity versus convexity for determination of particle shape (source:Altufi et al., 2013, with permission from ASCE). . . . . . . . . . . . . . . 37Figure 3.1 General configuration of flexible wall permeameter arrangement. . . . . . . 51Figure 3.2 Double-walled triaxial cell. . . . . . . . . . . . . . . . . . . . . . . . . . . 51Figure 3.3 Schematic plan view of double-walled triaxial cell, including volume changemeasurement systems. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52Figure 3.4 Pressures in the double-walled triaxial cell. . . . . . . . . . . . . . . . . . 52Figure 3.5 Dummy specimen mounted between base pedestal and top cap in double-walled triaxial cell (without acrylic tubes). . . . . . . . . . . . . . . . . . . 53xiFigure 3.6 Exploded view of connections between base frame, base pedestal, speci-men, and top cap. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54Figure 3.7 Relation between rate of volume change of inner chamber and cell pressurefrom calibration testing (t from 16 to 22 h after application of cell pressure). 55Figure 3.8 Particle size distribution of the glass bead components using QicPic dFmin . . 55Figure 3.10 Microscope image of component GB-C4 of glass beads. . . . . . . . . . . 57Figure 3.11 Microscope image of component GB-F of glass beads. . . . . . . . . . . . 57Figure 3.12 Particle size distribution of the 3.3GB gradation. . . . . . . . . . . . . . . 58Figure 3.13 Particle size distribution of the 4.8GB gradations. . . . . . . . . . . . . . . 58Figure 3.14 Particle size distribution of the 6.0GB gradations. . . . . . . . . . . . . . . 58Figure 3.15 Particle size distribution of the 6.5GB gradations. . . . . . . . . . . . . . . 59Figure 3.16 Particle size distribution of the components of soil. . . . . . . . . . . . . . 59Figure 3.18 Microscope image of the component BT-F of soil. . . . . . . . . . . . . . . 61Figure 3.19 Microscope image of the component BT-C3 of soil. . . . . . . . . . . . . . 61Figure 3.20 Particle size distribution of the 5.1BT gradation. . . . . . . . . . . . . . . 62Figure 3.21 Particle size distribution of the 5.7BT gradations. . . . . . . . . . . . . . . 62Figure 3.22 Particle size distribution of the 7.0BT gradations. . . . . . . . . . . . . . . 62Figure 3.23 Particle size distribution of the 8.6BT gradations. . . . . . . . . . . . . . . 63Figure 3.24 Particle size distribution of the 10.4BT gradations. . . . . . . . . . . . . . 63Figure 3.25 Variation of percentage finer fraction content per layer of a trial specimenof gradation 4.8GB20. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63Figure 3.26 Variation of percentage finer fraction content per layer of a trial specimenof gradation 6.5GB25. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64Figure 3.27 Variation of percentage finer fraction content per layer of a trial specimenof gradation 6.5GB35. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64Figure 3.28 Variation of percentage finer fraction content per layer of a trial specimenof gradation 6.0GB20. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64Figure 3.29 Variation of percentage finer fraction content per layer of a second trialspecimen of gradation 6.0GB20. . . . . . . . . . . . . . . . . . . . . . . . 65Figure 3.30 Test program. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66Figure 4.1 Response variables. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97Figure 4.2 GB-F-100 test results. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98Figure 4.3 6.5GB25-100 test results. . . . . . . . . . . . . . . . . . . . . . . . . . . . 99Figure 4.4 3.3GB20 test results. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100Figure 4.5 4.8GB20 test results. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101Figure 4.6 4.8GB35 test results. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102Figure 4.7 6.0GB20 test results. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 103Figure 4.8 6.0GB25 test results. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104xiiFigure 4.9 6.0GB30 test results. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105Figure 4.10 6.0GB35 test results. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106Figure 4.11 6.5GB35 test results. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107Figure 4.12 5.1BT20 test results. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 108Figure 4.13 5.7BT20 test results. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109Figure 4.14 5.7BT35 test results. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110Figure 4.15 7.0BT20 test results. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 111Figure 4.16 7.0BT35 test results. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112Figure 4.17 8.6BT20 test results. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 113Figure 4.18 8.6BT35 test results. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114Figure 4.19 10.4BT25 test results. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115Figure 4.20 10.4BT30 test results. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 116Figure 4.21 10.4BT35 test results. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117Figure 5.1 Micro-structure identification diagram for glass beads. . . . . . . . . . . . 149Figure 5.2 Micro-structure types. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 149Figure 5.3 Identification of initial micro-structure of glass beads test specimens: a)overview; b) detail. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 150Figure 5.4 Micro-structure identification diagram for soils. . . . . . . . . . . . . . . . 150Figure 5.5 Identification of initial micro-structure of soil test specimens: a) overview;b) detail. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 151Figure 6.1 Reader guide for discussion on factors controlling suffusion and suffosion. 178Figure 6.2 Experimental database compiled from literature. . . . . . . . . . . . . . . 179Figure 6.3 Suffosion: definition of average unit rate of deformation Ev. . . . . . . . . 179Figure 6.4 Suffosion: variation of Ev and D′15/d′85 in suffosive responses of glass beads. 179Figure 6.5 Suffosion: variation of volumetric and axial strain in tests on gradation4.8GB35. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180Figure 6.6 Suffosion: variation of volumetric and axial strain in tests on gradation6.0GB25. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180Figure 6.7 Suffosion: variation of volumetric and axial strain in tests on gradation6.0GB30. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 180Figure 6.8 Suffosion: variation of volumetric and axial strain in tests on gradation6.0GB35. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181Figure 6.9 Suffosion: variation of volumetric and axial strain in test on gradation6.5GB25. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181Figure 6.10 Suffosion: variation of volumetric and axial strain in tests on gradation6.5GB35. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181Figure 6.11 Onset of suffosion in tests on gradation 4.8GB35. . . . . . . . . . . . . . . 182xiiiFigure 6.12 Onset of suffosion in tests on gradation 6.0GB25. . . . . . . . . . . . . . . 182Figure 6.13 Onset of suffosion in tests on gradation 6.0GB30. . . . . . . . . . . . . . . 182Figure 6.14 Onset of suffosion in tests on gradation 6.0GB35. . . . . . . . . . . . . . . 183Figure 6.15 Onset of suffosion in test on gradation 6.5GB25. . . . . . . . . . . . . . . 183Figure 6.16 Onset of suffosion and condition at failure in tests on gradation 6.5GB35. . 183Figure 6.17 Onset of suffosion: variation between mean effective stress and differentialpore water pressure across the specimen in tests on glass beads gradations. 184Figure 6.18 Suffosion: upper limit for onset of suffosion and failure in tests on glassbeads gradations FR7 and FR8 (of Li, 2008). . . . . . . . . . . . . . . . . 184Figure 6.19 Seepage-induced internal instability phenomena in tests on glass beads gra-dations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184Figure 6.20 Variation of micro-structure of glass beads gradations with seepage-inducedinternal instability phenomena. . . . . . . . . . . . . . . . . . . . . . . . . 185Figure 6.21 Suffusion: variation of relative increase in hydraulic conductivity and por-tion of non-load bearing fine particles in C-T micro-structures of glass beads.185Figure 6.22 Variation of inter-coarse and inter-fine state parameters in glass beads testspecimens. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 185Figure 6.23 Onset of suffosion: variation of differential pore water pressure with char-acteristic state parameter in tests on glass beads gradations. . . . . . . . . . 186Figure 6.24 Suffosion: variation of volumetric and axial strains in tests on gradation10.4BT35. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186Figure 6.25 Onset of suffusion in tests on glass beads and soils. . . . . . . . . . . . . . 186Figure 6.26 Onset of suffosion in tests on gradation 10.4BT35. . . . . . . . . . . . . . 187Figure 6.27 Upper limit for onset of suffosion and failure in tests on soil gradationT-0 (adapted from Moffat et al., 2011, reproduced with permission fromCanadian Science Publishing). . . . . . . . . . . . . . . . . . . . . . . . . 187Figure 6.28 Upper limit for onset of suffosion and failure in tests on soil gradationT-5 (adapted from Moffat et al., 2011, reproduced with permission fromCanadian Science Publishing). . . . . . . . . . . . . . . . . . . . . . . . . 188Figure 6.29 Variation between p′so and ∆uso in soil gradation GS (data extracted fromChang and Zhang, 2013). . . . . . . . . . . . . . . . . . . . . . . . . . . . 188Figure 6.30 Seepage-induced internal instability phenomena in soils. . . . . . . . . . . 189Figure 6.31 Suffusion: variation of relative increase in hydraulic conductivity and por-tion of non-load bearing fine particles in C-T micro-structures of glassbeads and soils. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 189Figure 6.32 Variation of inter-coarse and inter-fine state parameters in glass beads andsoil test specimens. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190xivFigure 6.33 Onset of suffosion: variation of differential pore water pressure across thespecimen with characteristic state parameter in tests on glass beads and soilgradations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 190Figure 6.34 Unified approach for characterisation of suffosion. . . . . . . . . . . . . . 191Figure B.1 Membrane compliance: variation between particle size and constant of pro-portionality of membrane penetration per unit are (source: Frydman et al.,1973. Reprinted, with permission, from the Journal Testing and Evaluation1(1) (1973), copyright ASTM International, 100 Barr Harbor Drive, WestConshohocken, PA 19428). . . . . . . . . . . . . . . . . . . . . . . . . . . 215Figure C.1 Scale effect of hydraulic gradient. . . . . . . . . . . . . . . . . . . . . . . 218Figure D.1 4.8GB35-50: end-of-test side view shows small sign of local distress, nearthe top cap. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 220Figure D.2 6.0GB25-50: end-of-test front view shows signs of local distress at the tophalf of the specimen. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221Figure D.3 6.0GB30-50: end-of-test front view shows clear signs of local distress onthe right side. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 222Figure D.4 6.0GB30-150: end-of-test side view shows clear signs of local distress. . . 223Figure D.5 6.0GB30-150: top view, after dis-assembly of the device. . . . . . . . . . . 223Figure D.6 6.0GB35-50: end-of-test side view shows clear signs of distress. . . . . . . 224Figure D.7 6.0GB35-50: top view, after dis-assembly of the device, shows an abun-dance of fine particles. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 224Figure D.8 6.0GB35-100: end-of-test side view shows clear signs of local distress. . . 225Figure D.9 6.0GB35-100: top view, after dis-assembly of the device, shows an abun-dance of fine particles. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 225Figure D.10 6.5GB35-100: end-of-test side view. . . . . . . . . . . . . . . . . . . . . . 226Figure D.11 6.5GB35-100: top view, after dis-assembly of the device, shows an abun-dance of fine particles. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 226Figure D.12 10.4BT25-50: end-of-test side view shows coarse particles at the top halfof the specimen protruding under the membrane. . . . . . . . . . . . . . . 227Figure D.13 10.4BT25-50: top view, after dis-assembly of the device, shows an abun-dance of fine particles. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 227Figure D.14 10.4BT30-50: end-of-test front view shows coarse particles at the top halfof the specimen protruding under the membrane. . . . . . . . . . . . . . . 228Figure D.15 10.4BT30-50: top view, after dis-assembly of the device, shows an abun-dance of fine particles. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 229xvFigure D.16 10.4BT35-50: end-of-test side view, shows signs of local distress near thetop cap. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 230Figure D.17 10.4BT35-50: top view, after dis-assembly of the device, shows an abun-dance of fine particles. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 231Figure D.18 10.4BT35-50(R): end-of-test side view, shows signs of local distress at thetop half of the specimen. . . . . . . . . . . . . . . . . . . . . . . . . . . . 232Figure D.19 10.4BT35-50(R): end-of-test front view shows signs of local distress at theright side of the specimen. . . . . . . . . . . . . . . . . . . . . . . . . . . 233Figure D.20 10.4BT35-50(R): top view, after dis-assembly of the device, shows an abun-dance of fine particles. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 234Figure D.21 10.4BT35-100: end-of-test side view, shows signs of local distress on theleft side of the specimen. . . . . . . . . . . . . . . . . . . . . . . . . . . . 235Figure D.22 10.4BT35-100: end-of-test front view, shows signs of local distress at theright side of the specimen. . . . . . . . . . . . . . . . . . . . . . . . . . . 236Figure D.23 3.3GB20-50: post-test particle size analyses. . . . . . . . . . . . . . . . . 238Figure D.24 4.8GB20-150: post-test particle size analyses. . . . . . . . . . . . . . . . . 238Figure D.25 4.8GB35-100: post-test particle size analyses. . . . . . . . . . . . . . . . . 238Figure D.26 4.8GB35-150: post-test particle size analyses. . . . . . . . . . . . . . . . . 239Figure D.27 6.0GB20-50: post-test particle size analyses. . . . . . . . . . . . . . . . . 239Figure D.28 6.0GB20-100: post-test particle size analyses. . . . . . . . . . . . . . . . . 239Figure D.29 6.0GB20-150: post-test particle size analyses. . . . . . . . . . . . . . . . . 240Figure D.30 6.0GB25-150: post-test particle size analyses. . . . . . . . . . . . . . . . . 240Figure D.31 6.0GB30-150: post-test particle size analyses. . . . . . . . . . . . . . . . . 240Figure D.32 6.0GB35-100(R): post-test particle size analyses. . . . . . . . . . . . . . . 241Figure D.33 6.0GB35-150: post-test particle size analyses. . . . . . . . . . . . . . . . . 241Figure D.34 6.5GB35-50: post-test particle size analyses. . . . . . . . . . . . . . . . . 241Figure D.35 6.5GB35-100: post-test particle size analyses. . . . . . . . . . . . . . . . . 242Figure D.36 5.1BT20-50: post-test particle size analyses. . . . . . . . . . . . . . . . . . 242Figure D.37 5.1BT20-150: post-test particle size analyses. . . . . . . . . . . . . . . . . 242Figure D.38 5.7BT20-50: post-test particle size analyses. . . . . . . . . . . . . . . . . . 243Figure D.39 5.7BT20-100: post-test particle size analyses. . . . . . . . . . . . . . . . . 243Figure D.40 5.7BT20-150: post-test particle size analyses. . . . . . . . . . . . . . . . . 243Figure D.41 5.7BT35-100: post-test particle size analyses. . . . . . . . . . . . . . . . . 244Figure D.42 7.0BT20-50: post-test particle size analyses. . . . . . . . . . . . . . . . . . 244Figure D.43 7.0BT20-150: post-test particle size analyses. . . . . . . . . . . . . . . . . 244Figure D.44 8.6BT20-50: post-test particle size analyses. . . . . . . . . . . . . . . . . . 245Figure D.45 10.4BT25-50: post-test particle size analyses. . . . . . . . . . . . . . . . . 245Figure D.46 10.4BT30-50: post-test particle size analyses. . . . . . . . . . . . . . . . . 245xviFigure D.47 10.4BT35-50(R): post-test particle size analyses. . . . . . . . . . . . . . . 246Figure D.48 10.4BT35-100: post-test particle size analyses. . . . . . . . . . . . . . . . 246xviiList of AcronymsBT sub-angular soil particlesDAQ data acquisition systemDEM discrete element modellingDPT differential pressure transducerGB glass beadsI-CHD inflow constant-head deviceICOLD International Committee on Large DamsLVDT linear variable differential transformerMP measurement portO-CHD outflow constant-head deviceTDH total dynamic headTPT total pressure transducerUBC the University of British ColumbiaxviiiGlossaryα stress reduction factor∆min f mass loss inferred from analysis of seepage regime and deformations∆mobs mass loss established through forensic observations∆t duration of a stage of seepage flow∆u differential pore water pressure∆u f differential pore water pressure at failure∆ul differential pore water pressure at the last stage of seepage flow∆uso differential pore water pressure at the onset of suffosion∆usu differential pore water pressure at the onset of suffusion∆V volume change∆Vm volume change resulting from membrane compliance∆vm membrane penetration per unit area∆Vt total measured volume change, including membrane complianceεa axial strainεa,u axial strain during isotropic unloadingεv volumetric strainεv,u volumetric strain during isotropic unloadingγs unit weight of solidsγw unit weight of waterµ coefficient of static frictionxixµw dynamic viscosity of waterφ piezometric headΨ f inter-fine state parameterΨs modified inter-coarse state parameterρ densityρs density of solidsρw density of waterσc cell pressureσ ′v f vertical effective stress on the fine particle fractionσ ′v vertical effective stressA projected particle areaa threshold grain size ratioAm soil surface covered by the membraneAR aspect ratioAs specific surface of solidsB area of convex hullb portion of load-bearing fine particlesCc coefficient of curvatureCp pore shape factorCt tortuosity factorCu coefficient of uniformityCx convexityD specimen diameterd particle sizeD′ particle size of the coarse component of a gap-graded gradationxxd′ particle size of fine component of a gap-graded gradationD′15/d′85 ratio of the particle size of the coarse fraction corresponding to the 15th percent masspassing, and the particle size of the fine components corresponding to the 85th percentmass passingDc,D95 constriction size corresponding to the 95th percentile of the theoretical densest constric-tion curve for the coarse fractiondFmax maximum Feret diameterdFmean mean Feret diameterdFmin minimum Feret diametere void ratioec void ratio at the end of consolidationecub void ratio corresponding to simple cubic fabrice f inter-fine void ratioe f ,max maximum index void ratio of the fine fractione f ,min minimum index void ratio of the fine fractionemax maximum index void ratioemin minimum index void ratioes inter-coarse void ratioese equivalent inter-coarse void ratioes,max maximum index void ratio of the coarse fractiones,min minimum index void ratio of the coarse fractionEv average unit rate of volumetric deformationf functionFb buoyant force acting on the solidsFd drag acting on the solidsFg weight of the solidsxxiFr resultant force on the solidsg gravitational accelerationGs specific gravity of solidsi hydraulic gradientk hydraulic conductivityK0 lateral earth pressure coefficient with zero lateral strainki initial hydraulic conductivitykl hydraulic conductivity at the last stage of seepage flowl specimen lengthLs length scalen porosityn f inter-fine porosityP projected perimeterp′ mean effective stressp′c mean effective stress at the end of consolidationp′f mean effective stress at failurep′so mean effective stress at the onset of suffosionp′su mean effective stress at the onset of suffusionq deviatoric stressR roundnessrc sample correlation coefficientRd size disparity ratioRe Reynolds numbers standard deviationseq equivalent coarse grain spacing constantxxiiS f finer fraction contentS∗f critical finer fraction contentSf,L limit fine fraction contentsi standard deviation of base variable zisx standard deviation of derived variable zxSm constant for proportionality of membrane penetration per unit areaSQP sphericityT temperaturet timeu pore water pressureV specimen volumev specific dischargevs seepage velocityVp,c volume of particles constituting the coarse fractionVp, f volume of particles constituting the fine fractionvRe=1 upper limit of specific discharge in Darcian flow regimeVv total volume of voidsx¯ arithmetic mean of xz z coordinatez¯i predicted value of variable zizi base variable zixxiiiAcknowledgmentsI’d like to thank my research supervisor, Dr. R. J. Fannin and my committee members Dr. Z.Shawwash, Dr. M. Taiebat, and Dr. D. Wijewickreme for their support. The assistance ofDr. Y. Vaid during the development of the apparatus is much appreciated, as is the technicalsupport of S. Jackson, H. Shrempp and B. Leung. I’d also like to thank Dr. C. O’Sullivanfor use of the QicPic apparatus during my visit to Imperial College London. Funding for theresearch was provided by the Natural Science and Engineering Research Council of Canada, BCHydro, the UBC Faculty of Graduate and Postdoctoral Studies, and Vice President, Researchand International Office.xxivChapter 1IntroductionInternal erosion is broadly defined as the migration of soil particles by seepage flow withinearth structures, such as an embankment dam or its foundation. Statistical analysis establishedthat internal erosion is the governing failure mode for approximately half of the failures ofembankment dams (Foster et al., 2000). The contemporary design philosophy involves theconstruction of zoned embankment dams whereby filters and drains provide means of control-ling the seepage flow through the embankment and preventing migration of particles betweenvarious zones of the embankment and its foundation (USBR, 2011). However, many existingembankment dams have not been provided with adequate filters and drains (ICOLD, 2013) andthe safety assessment of these dams requires a fundamental understanding of the factors gov-erning internal erosion. The European Working Group of International Committee on LargeDams (ICOLD) recently summarised the state-of-art of internal erosion in embankment dams(ICOLD, 2013) and proposed a framework for the safety assessment of embankment dams withrespect to internal erosion, which was adopted from Fell and Fry (2007). For the assessment ofthe likelihood, and effect, of internal erosion, four phases are distinguished: (i) the initiation ofinternal erosion; (ii) the continuation of erosion; (iii) the progression of erosion to a larger zone;and (iv) the initiation of a breach in the structure. It is recommended to consider the followingmechanisms through which internal erosion can initiate: erosion in concentrated leaks, such ascracks; backward, or retrogressive erosion of particles that are not retained by a downstreamfilter; contact erosion at the interface between two materials; and “suffusion” or “internal insta-bility”, which is described as the transport of small particles by seepage flow through the poresof the coarser particles. A combination of empirical and physics-based tools are presented todetermine the conditions at the onset, and subsequent progression, of the first three internalerosion mechanisms. In contrast, the description of the factors governing internal instability isminimal and is mainly of a qualitative nature. This investigation aims to address this knowledgegap and is an attempt to advance the fundamental understanding of seepage-induced internalinstability.1Notably, Richards and Reddy (2007) and Moffat et al. (2011) recommend to distinguishbetween the seepage-induced internal instability phenomena of suffusion, “where the finerfraction of an internally unstable soil moves within the coarser fraction without any loss ofmatrix integrity or change in total volume” (Moffat et al., 2011), and suffosion, “where parti-cle migration yields a reduction in total volume and a consequent potential for collapse of thesoil matrix” (Moffat et al., 2011). Three factors controlling suffusion and suffosion are gen-erally distinguished: material susceptibility, effective stress and hydraulic conditions (Garnerand Fannin, 2010). The material susceptibility has historically been investigated by conduct-ing laboratory tests. These investigations have yielded several useful empirical tools, whichare exclusively based on the particle size distribution, to assess the susceptibility of materialsto seepage-induced internal instability, without distinguishing between suffusion and suffosion(e.g. Burenkova, 1993; Kenney and Lau, 1985, 1986; Kezdi, 1979; Li, 2008; Sherard, 1979;Wan and Fell, 2008). Recently, Crawford-Flett (2014) confirmed a previous postulate that soilswith a stable, coarse-particle dominated skeleton, exhibit a potential for suffusion (e.g. Kenneyand Lau, 1985; Kezdi, 1979; Skempton and Brogan, 1994; Wittmann, 1978). Skempton andBrogan (1994) were one of the first researchers to specifically address the hydraulic conditionsat the onset of seepage-induced internal instability, by subjecting gap-graded materials to up-ward seepage flow in a rigid wall permeameter. Research conducted at the University of BritishColumbia (UBC), using a rigid wall permeameter, indicated that the onset of seepage-inducedinternal instability was also controlled by the effective stress, in addition to the hydraulic con-ditions (Crawford-Flett, 2014; Li, 2008; Moffat, 2005). Chang and Zhang (2013), based ontests on one granular soil in a flexible wall permeameter, suggested that the onset and progres-sion of suffosion is dependent on effective stress. Driven by the postulates regarding the roleof the effective stress on the onset of seepage-induced internal instability, recent studies (Shireand O‘Sullivan, 2013; Shire et al., 2014), using discrete element modelling (DEM), establishedthat in certain gap-graded materials, the fine particles exhibit a smaller effective stress than thecoarse particles .Seepage-induced internal instability can thus be considered an emerging field of research,where the principle focus of investigation has shifted from the susceptibility of materials tothe conditions at which seepage-induced internal instability initiates. The understanding of thefactors governing seepage-induced internal-instability has advanced to a largely qualitative ap-preciation of the role of the hydraulic conditions and the effective stress. A further advance to aquantitative approach is believed to be contingent on the generation and analysis of experimen-tal data on seepage-induced internal instability, using suitable laboratory testing equipment,with control of the hydraulic conditions and effective stress.21.1 Research hypothesesThe aim of this study is to establish causative relations between the factors governing seepage-induced internal instability. Four research hypotheses are proposed to achieve this aim.Considering the visual observations of Moffat et al. (2011) that “significant voids in thespecimen” associated with suffosion, developed in internally unstable materials, volume changeis postulated to be a characteristic variable of seepage-induced internal instability. It is hypoth-esized that volume change can serve to distinguish between suffusion, which is associated withno volume change, and suffosion. The first hypothesis is proposed as follows:Hypothesis no. 1: Volume change is a characteristic variable of seepage-induced in-ternal instability and serves to distinguish between suffusion and suffosion.The influence of effective stress on the onset of seepage-induced internal instability hasmainly been established from tests in rigid wall permeameters (e.g. Li, 2008; Moffat, 2005),which exhibit shortcomings in the control of the effective stress. The recent advance towardsthe use of flexible wall permeameters (e.g. Bendahmane et al., 2008; Chang and Zhang, 2013;Ke and Takahashi, 2014b; Luo et al., 2012; Xiao and Shwiyhat, 2012), with better control of theeffective stress state, enables a more thorough examination of the influence of effective stresson seepage-induced internal instability. Notwithstanding the influence of other factors, the sec-ond hypothesis is defined to advance the understanding of the influence of effective stress onthe onset of suffosion:Hypothesis no. 2: The onset of suffosion is dependent on effective stress.Although suffusion has typically been attributed to the migration of fine particles from astable coarse-particle-dominated skeleton, the causality between the soil micro-structure andsuffosion has not been addressed. The third hypothesis has been defined to investigate the rela-tion between the soil micro-structure and suffusion and suffosion:Hypothesis no. 3: The micro-structure of internally unstable materials controls thephenomenological response to seepage flow.Finally, considering the DEM work on gradations of spherical particles, and the generallyrecognised influence of the particle shape on the shear strength of soils, it seems prudent toexamine the role of the particle shape on suffusion and suffosion. The fourth hypothesis isdefined with the purpose of establishing a causative relation between particle shape and thesusceptibility to seepage-induced internal instability:3Hypothesis no. 4: Gap-graded gradations of rounded particles are more susceptible toseepage-induced internal instability than identical gap-graded gradations of sub-angularparticles.1.2 Scope and objectivesThe scope of this work is the laboratory element testing of reconstituted gap-graded granularmaterials. This laboratory investigation comprises a parametric sensitivity study where thedependent variable is the response of a material subject to seepage flow and effective stress.The independent material variables are the finer fraction content, ratio of the particle size of thecoarse and fine fractions, and particle shape. The independent environmental variables are themean effective stress and differential pore water pressure across the specimen. Five researchobjectives have been defined for this study, with the intend of partitioning the investigation inmanageable projects. The research objectives of this study are:1. To summarise the state-of-art of seepage-induced internal instability and in particular:factors controlling suffusion and suffosion; flexible wall permeameters used for seepage-induced internal instability testing; and the micro-structure of gap-graded materials.2. To design a flexible wall permeameter, with control of effective stress and hydraulic load,and with measurement of volume change.3. To commission the flexible wall permeameter and to assess the repeatability of the testprocedure.4. To determine the phenomenological response to seepage flow of tests on different gap-graded materials, subject to varying stress conditions.5. To investigate the suitability of the volume change as a variable to quantify seepage-induced internal instability, and to determine the influence of effective stress, soil micro-structure, and particle shape on suffusion and suffosion.1.3 Outline of the thesisThe investigation is addressed in seven chapters:• Chapter One introduces the subject of seepage-induced internal instability, reports onrecent advances and identifies knowledge gaps. Four research hypotheses are presented.• Chapter Two summarises the state-of-art knowledge on seepage-induced internal insta-bility, with particular focus on suffusion and suffosion, experimental investigations, the4micro-structure of gap-graded materials and the influence of the particle shape (researchobjective No. 1).• Chapter Three describes the flexible wall permeameter, which has been developed forthis study (research objective No. 2), the selection of the materials, and the test program.• Chapter Four commences with the commissioning of the flexible wall permeameter. Theresults of the test program are subsequently presented, from which the repeatability ofthe test procedure is derived (research objective No. 3).• Chapter Five presents the analysis of the test results, which yields the phenomenologicalresponse to seepage flow in each test (research objective No. 4).• The research hypotheses are tested in Chapter 6 (research objective No. 5). Causativerelations between the dependent variables are sought, and form the basis of a broaderdiscussion on the factors governing suffusion and suffosion.• Chapter Seven summarises important findings of this investigation, accentuates novelcontributions, presents recommendations for future research and discusses implicationsfor practice.5Chapter 2Literature reviewThis Chapter summarises the state-of-art on seepage-induced internal instability (research ob-jective No.1). First, distinct seepage-induced internal instability phenomena are identified andtermed in Section 2.1, followed by a summary of the micro-structure of soils, including the roleof the particle shape, in Section 2.2. Several important aspects of the seepage flow of waterthrough a porous medium are summarised in Section 2.3. Advances in laboratory testing ofinternally unstable materials are summarised in Section 2.4, where a distinction is made be-tween the two most common types of devices: rigid wall permeameters (Section 2.4.1) andflexible wall permeameters (Section 2.4.2). The main findings of the experimental work aresummarised in Section 2.5. A reflection on the research hypotheses, in light of the literature re-view, is presented in Section 2.6. A summary of key aspects of the literature review is providedin Section 2.7.2.1 Seepage-induced internal instability phenomenaThere appears to be consensus in the literature that internal instability is a phenomenon wherebyfine particles are transported from a non-plastic soil by seepage flow (see Table 2.1). A sub-tle distinction has been proposed between the migration of particles within a soil, and out ofa soil (Kovacs, 1981). A more significant distinction has been made between a washed-outsoil structure that remains intact, and one in which some form of destruction, or collapse, ofthe structure accompanies the seepage-induced migration of fine particles (e.g. A˚ berg, 1993;Burenkova, 1993; Garner and Sobkowicz, 2002; Kezdi, 1979; Li, 2008; Moffat et al., 2011;Richards and Reddy, 2007; Wittmann, 1978). Herein the term structure is used to take intoaccount both the soil fabric and its stability, as suggested by Mitchell and Soga (2005); a moreelaborate review of soil structure is presented in Section 2.2.Various terms have been used to describe internal instability, with and without some formof collapse of the soil structure. More unfortunate is the use of the same term for each of these6distinct phenomena: suffosion has been used to describe both destructive (e.g. Burenkova,1993; Garner and Sobkowicz, 2002; Moffat et al., 2011; Richards and Reddy, 2007) and non-destructive phenomena (e.g. Burenkova, 1993; Molenkamp et al., 1979; Wittmann, 1978) ofseepage-induced internal instability (see Table 2.1). In contrast, suffusion has only been usedto describe the non-destructive phenomenon of seepage-induced internal instability.2.1.1 Experimental evidence of internal instability phenomenaExperimental evidence suggests that the three variables of (i) a measured value of mass loss,(ii) a measured value of volume change, and, (iii) a value of change in hydraulic conductivity,deduced from measurements of hydraulic gradient and flow rate, are sufficient to quantify, andhence distinguish between, seepage-induced instability phenomena:• Skempton and Brogan (1994) report seepage-induced internal instability in a specimenof gap-graded sandy gravel subject to upward flow in a rigid-wall permeameter. The topsurface of the specimen was unloaded. The linear relation between hydraulic gradient,i, and the specific discharge through the specimen to i ≤ 0.11 (see Fig. 2.1) indicatesa constant value of hydraulic conductivity. At i > 0.11, a disproportionate increase ofseepage velocity with hydraulic gradient indicates an increase in hydraulic conductiv-ity. At i = 0.2, there was “strong general piping of fines throughout”, but “[t]he gravelparticles remain undisturbed.” The experimental findings are: (i) mass loss; (ii) no col-lapse of the soil structure and hence no volume change; and, (iii) an increase in hydraulicconductivity.• Li (2008) reports seepage-induced internal instability in a specimen of gap-graded glassbeads, subject to downward flow in a rigid-wall permeameter. The top surface of thespecimen was loaded. The linear relation between hydraulic gradient and specific dis-charge to i ≤ 2.9 (see Fig. 2.2) indicates a constant value of hydraulic conductivity. At i> 2.9, a disproportionate increase of discharge velocity with hydraulic gradient indicatesan increase in hydraulic conductivity. “Negligible mass loss and axial displacement wereobserved during these flow stages [..] Upon imposing a small increase in hydraulic gra-dient to i = 3.2, a modest amount of finer particles (6.9 %) was lost from the specimen[..] A total downward axial displacement of 2.5 mm was measured, resulting in an ax-ial strain of 2.6 %.” Thus, in contrast to the findings of Skempton and Brogan (1994),and notwithstanding the different hydraulic conditions, the experimental findings of Li(2008) are: (i) mass loss; (ii) contractive volume change associated with collapse of thesoil structure; and, (iii) an increase in hydraulic conductivity.In the classic description of piping by heave of sand, Terzaghi and Peck (1948) observe:“This process greatly increases the permeability of the sand [..] and [t]he surface of the sand7then rises.” Accordingly, this seepage-induced instability phenomenon may be similarly quan-tified by the same three variables:• Li (2008) reports seepage-induced instability of a specimen of broadly graded sand andgravel, subject to upward seepage flow in a rigid-wall permeameter. The top surface ofthe specimen was loaded. The linear relation between hydraulic gradient and dischargevelocity to i ≤ 15 (see Fig. 2.3) indicates a constant value of hydraulic conductivity.“Upon imposing an increase in hydraulic gradient to i = 16.0 [..] a large upward dis-placement of 4.2 mm was measured, resulting in a strain of 1.3 %.” The experimentalfindings are: (i) no mass loss; (ii) expansive volume change; and (iii) an increase inhydraulic conductivity.2.1.2 Conceptual frameworkIn an internally unstable soil, the phenomenon whereby fine particles are transported by seep-age flow and the soil structure remains intact, may be quantified by a mass loss, no change involume and an increase in hydraulic conductivity (see Fig. 2.4a). The term suffusion has beencommonly used to describe this non-destructive response (see Table 2.1), and its continued useis therefore advocated. In contrast, the internal instability phenomenon whereby the transportof fine particles by seepage flow is accompanied by a collapse of the soil structure, may bequantified by a mass loss, a volumetric contraction and a change in hydraulic conductivity (seeFig. 2.4b). It is a different response, and one with potential for unacceptable deformation, forwhich the term suffosion is recommended.An internally stable soil may be either uniformly graded or broadly graded. In the pres-ence of upward seepage flow, the instability phenomenon may be quantified by a volumetricexpansion, accompanied by an increase in hydraulic conductivity (e.g. see Fig. 2.4c). The termfluidisation has been used to describe this response (Vardoulakis, 2004) and its continued useappears appropriate.The recommendation to distinguish between the three phenomena of suffusion, suffosionand fluidisation, based on mass loss, volume change and change in hydraulic conductivity, is il-lustrated in a revised version of the Venn-diagram (see Fig. 2.5), originally proposed by Garnerand Fannin (2010). The conceptual framework presented herein can thus successfully distin-guish between the seepage-induced internal instability phenomena of suffusion and suffosion,in both a qualitative and a quantitative manner.82.1.3 Concluding remarksA conceptual framework is developed such that a distinction can be reasonably made betweenphenomenological responses based on mass loss and volume change, which can be measureddirectly, and change in hydraulic conductivity, which can be deduced from measurement of hy-draulic gradient and flow rate. Recognising the important distinction between non-destructiveand destructive phenomena in engineering practice, it is recommended that:• suffusion be characterised as seepage-induced mass loss without change in volume, ac-companied by an increase of hydraulic conductivity; and,• suffosion be characterised as a seepage-induced mass loss accompanied by a reductionin volume and change in hydraulic conductivity; and,• fluidisation be characterised as a seepage-induced volumetric expansion, accompaniedby an increase in hydraulic conductivity, with no mass loss.2.2 Micro-structureMitchell and Soga (2005) made an important distinction between soil fabric and soil structure:“[t]he term fabric refers to the arrangement of particles, particles groups and pore spaces ina soil.” Structure is used to refer to “the combined effects of fabric, composition, and inter-particle forces”, i.e. soil structure refers to the soil fabric and its stability. The term micro-structure is used when considering aspects of the fabric and its stability on a particle level1. Inthis study, fabric and particle arrangement will be used interchangeably to improve readability.2.2.1 Micro-structures of gap-graded mixturesA suitable starting point for the discussion on fabric and structure is the theoretical packing ofequal-size spherical particles. The loosest theoretical fabric of equal-size spheres, for which theparticles are in contact with each other, is the simple-cubic arrangement with a theoretical voidratio e = 0.91, and the densest theoretical fabric of equal-size spheres is the tetrahedral packing,with a theoretical void ratio e = 0.35. Several researchers have experimentally investigated themaximum and minimum densities of spheres and found that a stable fabric of equal-size spherescan only exist in a narrow range of void ratios. McGeary (1961) investigated the maximum pos-sible density of equal-size spheres in a large container using mechanical vibration and obtainedvalues of maximum packing densities emin = 0.602. In their classic study, Scott and Kilgour1The definition of micro-structure, as a particle arrangement and its stability, in this study, is thus distinct fromthe adjective “microstructured” used by Leroueil and Hight (2003) to characterise the mechanical behaviour ofcertain soils.2It should be noted here that minimum and maximum void ratio values are variable and operator dependent;accordingly, Holtz et al. (2011) recommend considering emin and emax as minimum and maximum index void ratios,as at least theoretically, denser and looser states appear feasible. In this study, emin and emax are considered index9(1969) reported values for minimum and maximum densities of randomly packed steel ballsof nearly perfect uniformity of emax = 0.67 and emin = 0.57, when correcting for the effect ofthe container diameter, using the elegant procedure proposed by Scott (1960). McGeary (1961)also investigated the maximum possible density of binary mixtures of discrete fine and coarseparticle fractions and found that maximum densities could only be obtained by first placing andvibrating the coarse component to emin = 0.60, then placing the fine component, and vibratingagain. The findings of McGeary (1961) demonstrate that the maximum density of binary mix-tures of discrete fine and coarse particle fractions can only be attained if the ratio between theparticle size of the coarse fraction (D′) and the particle size of the fine fraction (d′), is D′/d′ >10.The concept of binary mixtures led Kenney and Lau (1985), and later Skempton and Bro-gan (1994), to the observation that seepage-induced internal instability is limited to a clast-supported micro-structure, which is characterised by a primary fabric of coarse particles. Clast-supported micro-structures were postulated to exhibit a potential for non-load bearing fine par-ticles. In different appearances, Skempton and Brogan (1994), Kezdi (1979), Kenney and Lau(1985) and Wittmann (1978) presented equivalent expressions to determine the theoretical crit-ical finer fraction content S∗f above which a clast-supported micro-structure cannot exist. Thedefinition of Kenney et al. (1985) is preferred for its simplicity:S∗f = 1−11+ es(1−n f ) (2.1)where es is inter-coarse void ratio (see Eq. 2.2), and n f is inter-fine porosity (see Eqs. 2.3and 2.4).The inter-coarse void ratio es is defined as the void ratio of the coarse grains if the fineparticles were considered voids (see Fig. 2.6):es =Vv +Vp, fVp,c=V −Vp,cVp,c=e+S f1−S f (2.2)where Vv is total volume of voids, Vp, f is volume of particles constituting the fine fraction,Vp,c is volume of particles constituting the coarse fraction, V is specimen volume, and S f isfiner fraction content.Similarly, in porous media with a high content of fine particles, the inter-fine void ratio, e f(see Fig. 2.6), can be calculated by ignoring the coarse particles:void ratios that refer to the practically, as opposed to theoretically, densest and loosest states, respectively, of agranular material.10e f =VvVp, f=eS f(2.3)The corresponding inter-fine porosity n f can then be calculated using Eq. 2.4:n f =e f1+ e f(2.4)In this thesis, the minimum and maximum index inter-coarse void ratios, es,min and es,max,are defined as the void ratios corresponding to the practically densest and loosest void ratios,respectively, of a packing of only coarse particles. Similarly, the minimum and maximum in-dex inter-fine void ratios, e f ,min and e f ,max, are defined as the void ratios corresponding to thepractically densest and loosest void ratios, respectively, of a packing of only fine particles.Thevanayagam et al. (2002) proposed a micro-structure identification diagram as a “guide-line to determine the anticipated behavior of gap-graded granular mixtures”. The following fivemicro-structure “cases” were identified, based on void ratio and finer fraction content (see Fig.2.7):3• Case i: “the fines are confined within the void spaces between the coarse grains withlittle contribution to supporting the coarse grain skeleton” (Thevanayagam et al., 2002).Accordingly, the coarse particle components has to yield a stable fabric with es < es,max,which was postulated to be possible only if D′/d′ > 6.5.• Case ii: “[the fines] are partially supporting the coarse grain skeleton” (Thevanayagamet al., 2002). The limits for this case are not clearly defined, but Thevanayagam (1998)appears to suggest that es ≈ es,max.• Case iii: “[the fines] partially separate the coarse grains” (Thevanayagam et al., 2002).The inter-coarse void ratio es > es,max indicates that the fine particles are partially sup-porting the coarse-particle dominated micro-structure.• Case iv-1: “the coarse grains are fully dispersed in the in the fine grain matrix [..] thebehavior of the soil mix is entirely governed by the fine grains” (Thevanayagam et al.,2002). A limiting finer fraction content Sf,L is proposed above which the coarse particlesare fully dispersed in the fine particle matrix (see Eq. 2.5).• Case iv-2: the coarse grains are partially dispersed and contribute as a reinforcing elementif S∗f < S f < Sf,L.3The specimen preparation technique influences the structure of the soil, related to the particle arrangement,composition and inter-particle forces, which governs the stress-strain response (e.g. Vaid and Sivathayalan, 2000).It should be noted that the void ratio and finer fraction content do not quantify these characteristics of structure11The limiting finer fraction content Sf,L is determined based on geometrical considerationsof a theoretical particle packing:Sf,L = 1− pi(1+ e)6s3eq=6s3eq−pi6s3eq +pie f(2.5)where seq is equivalent coarse grain spacing constant, which is defined as:seq = 1+ad′D′(2.6)with the threshold grain size ratio a = 10, based on experimental observations of Roscoe(1970, as cited in Thevanayagam and Mohan, 2000), that the zone of influence of shearing in auniform grained soil is about 10 times the grain size.Thevanayagam et al. (2002) also defined the portion of the fine grains that are load bearingb, with 0 ≤ b ≤ 1, to account for the secondary cushioning effects of the fines contributing tofabric stability in cases ii and iii: b = 0 indicates no contribution of the fines to the support of thecoarse-grain skeleton and b = 1 indicates that all fines are actively contributing to the supportof the coarse grain skeleton. The equivalent inter-coarse void ratio, ese, which accounts for thecontribution of the fine particles to the stability of the micro-structure, is introduced:ese =e+(1−b)S f1− (1−b)S f (2.7)Rahman et al. (2008) argued that b is a function of both S f and the ratio (D′/d′), based onthe findings of McGeary (1961) that the minimum void ratio of binary mixtures approaches atheoretical minimum void ratio only if (D′/d′) > 10; mixtures with small particle ratios yieldedsignificantly greater void ratios, which was attributed to the fine particles pushing the coarseparticles apart.Crawford-Flett (2014) constructed a micro-structure identification diagram, adapted fromThevanayagam et al. (2002), to investigate the potential for suffusion in gap-graded grada-tions of glass beads. The theoretical limits es,max = e f ,max = 0.95 and es,min = e f ,min = 0.35were assumed to develop the diagram, which appear to yield too wide a range of void ra-tios considering to the experimental work of McGeary (1961), and Scott and Kilgour (1969).Crawford-Flett (2014) defined only three micro-structure types, instead of the five cases definedby Thevanayagam et al. (2002):• A clast-supported coarse-particle dominated micro-structure with es < es,max, and e f >e f ,max. “[A] significant portion of the finer particles will not be fixed in place by inter-particle contacts” (Crawford-Flett, 2014).12• Matrix-supported fine-particle dominated micro-structure with e f < e f ,max, and es >es,max. “The coarser particles are therefore separated by the fixed matrix of finer par-ticles” (Crawford-Flett, 2014).• Transitional micro-structure with es < es,max, and e f < e f ,max. “Finer particles are in firmcontact with coarser grains of the soil skeleton; therefore, both finer and coarser particleswill contribute to stress transfer through inter[-]particle contacts” (Crawford-Flett, 2014).2.2.2 Particle shapeMeasures of particle shape are presented in Section 2.2.2.1. It is followed by a summary of theinfluence of the particle shape on the stability of particle packings in Section 2.2.2.2.2.2.2.1 Measures of particle shapeThe state-of-practice description of the particle shape follows the approach of ASTM D2488-09a (ASTM, 2009), based on the charts provided by Powers (1953) and Krumbein and Sloss(1963), where particles are characterised according to four classes of roundness: angular, sub-angular, sub-rounded, and rounded. Altuhafi et al. (2013) demonstrate that this approach issomewhat subjective, and propose an new method based on measures of sphericity and con-vexity. The QicPic apparatus (Sympatec, 2008), which is based on a digital imaging technique,was used to obtain measures of sphericity and convexity. The convexity Cx is defined as theratio between the projected particle area A and the sum of the projected particle area and thearea of the convex hull (A + B) (see Fig. 2.8):Cx =AA+B(2.8)The sphericity SQP is defined as the ratio of the perimeter of a circle whose area equals theprojected particle area, and the projected perimeter, P:SQP =2√piAP(2.9)The equivalent particle shape used by ASTM D2488-09a (ASTM, 2009), can then be de-termined from a plot of Cx versus SQP (see Fig. 2.9).Additional useful parameters are the Feret diameter and the aspect ratio. The Feret diam-eter is measured as the distance between two tangents on opposite sides of the particle (seeFig. 2.8). The maximum, minimum, and mean Feret diameters are given by dFmax , dFmin anddFmean , respectively. According to Altuhafi et al. (2013), dFmin is the measure that most closelycorresponds to the particle size determined using sieving. The aspect ratio AR is defined as theratio between the Feret diameters, see Eq. 2.10.13AR =dFmaxdFmin(2.10)Aspect ratio, convexity and sphericity are typically reported as median values for a cumu-lative distribution by volume as AR50, Cx50 and SQP50, respectively.2.2.2.2 Role of particle shapeAlthough the particle shape greatly affects the engineering response of granular soils (Holtzet al., 2011), it has received little or no attention in the investigation of seepage-induced inter-nal instability. The work of Casagrande on internal friction angles of granular soils (in Holtzet al., 2011) shows a substantial increase of internal friction angle for sub-angular particlescompared to the internal friction angle of sub-rounded particles, reconstituted to the same voidratio. Youd (1973) demonstrates that minimum and maximum index void ratios, and the differ-ence between minimum and maximum index void ratios, increase as the particles become moreangular. Mitchell and Soga (2005), referring to Hagerty et al. (1993), note that “[..] angularglass beads are more susceptible to breakage than round glass beads.” Cho et al. (2006) notethat angular particles yield a relatively looser packing, which exhibits a lower stiffness at smallstrains than a similar packing of rounded particles. Intermediate strain behaviour, associatedwith contact slippage or particle breakage, yielded higher compression and decompression in-dices of angular particles than rounded particles. At large strain behaviour, associated withparticle rotation and contact slippage, increasing angularity was postulated to increase the re-sistance against particle rotation, yielding a greater shear resistance.2.3 Seepage flow in granular materialsA review of seepage flow in granular media naturally starts with the pioneering work of Darcy(1856, as cited in Bear, 1972). From experiments, Darcy concluded that the specific discharge,v, is proportional to the difference in piezometric head, ∆φ , and inversely proportional to thespecimen length, l:v = k∆φl(2.11)where the coefficient of proportionality k is hydraulic conductivity.Mitchell and Soga (2005) list three hypotheses to account for a deviation from Darcy’slaw: 1) a non-Darcy flow regime; 2) particle migration; or 3) changes in specimen volume.The effect of non-Darcy flow is commonly, but erroneously, associated with the persistenceof turbulent flow. Laminar flow occurs when a fluid flows in parallel layers, whereas lateralmixing of fluid particles between layers occurs in turbulent flow. Dybbs and Edwards (1984),14in their excellent review of flow of liquids in porous structures, note the existence of four flowregimes in porous media, based on the Reynolds number Re (see Eq. 2.12):1. The Darcy flow regime occurs at Re < 1. The flow in this laminar, linear flow regime isdominated by the viscous forces.2. The inertial flow regime initiates at 1 < Re < 10 and persists to Re ≈ 150. The flow islaminar, but non-linear. The flow regime is sometimes referred to as the “Forchheimer”flow regime.3. An unsteady laminar flow regime at Reynolds numbers between approximately 150 and300. Unsteady flow occurs “in the form of laminar wake oscillations in the pores” (Dybbsand Edwards, 1984).4. A highly unsteady flow regime for Reynolds numbers greater than approximately 300,“qualitatively resembling turbulent flow” (Dybbs and Edwards, 1984).The Reynolds number Re is the ratio of the inertial forces and the viscous forces of a flow:Re =ρwv2sµwvs/Ls=vsLsρwµw(2.12)where ρw is the density of water, µw is the dynamic viscosity of water4, vs is the averageseepage velocity through the pores (see Eq. 2.13), n is porosity, and Ls is a length scale.The average seepage velocity through the pores vs is defined as follows:vs =vn(2.13)The Kozeny-Carman relation can be used to determine the saturated hydraulic conductivityof porous media for seepage regimes where Darcy’s law is valid. It is based on conceptualmodels of capillary tube flow for which the Navier-Stokes equation can be used. The relationis commonly held to be valid for granular materials, while its validity for fine grained soils,especially clays, is a subject of debate (e.g. Carrier, 2003; Chapuis and Aubertin, 2003, 2004;Hansen, 2004; Mitchell and Soga, 2005). In this study, the form presented by Mitchell andSoga (2005) is adopted:k =1CtCpgµwρwe3A2s G2s (1+ e)(2.14)4In this study, which primarily deals with laboratory investigations in a controlled environment, the water tem-perature T = 20 ◦C is constant, yielding ρw = 998.2 kgm−3, and µw = 1.002×10−3 Nsm−2.15where g is gravitational acceleration, Ct is tortuosity factor, Cp is pore shape factor, As isspecific surface of solids and Gs is specific gravity of solids.Assuming constant values for viscosity and density of water and solids, which is a reason-able assumption in most geotechnical engineering applications, Eq. 2.14 establishes that thehydraulic conductivity increases with increasing void ratio, decreasing specific surface of thesolids and decreasing tortuosity.The interaction between the porous medium and fluid flow can also be appreciated by con-sidering the forces acting on the solid particles of a porous medium, subject to seepage flow ina vertical direction, per unit volume of porous medium, with positive z in a upward direction(after Bear, 1972):• Weight of the solids Fg acting in a downward direction is a function of the unit weight ofthe solids γs:Fg =−γs(1−n) (2.15)• The buoyancy force Fb is equal to the resultant of the liquid pressure u acting on the solidparticles:Fb =−(1−n)∂u/∂ z (2.16)Note that for the hydrostatic condition ∂u/∂ z = γw, the buoyancy force is equal theweight of the displaced fluid: Fb =−γw(1−n).• The drag force is derived from the available energy per unit fluid, φ , which drives thefluid through the porous medium. This energy is dissipated as viscous friction at thesolid-fluid interface which generates a drag on the solid matrix in the direction of fluidflow. The drag force Fd of the fluid, at the solid-fluid interface, per unit volume of porousmedium is:Fd =−nγw∂φ/∂ z. (2.17)The resultant force Fr per unit volume of porous medium acting on the solid matrix is then:Fr = Fg +Fb +Fd =−(γs− γw)(1−n)− γw∂φ/∂ z (2.18)The first term on the right-hand side of Eq. 2.18 is the specific submerged unit weight of thesolid matrix, i.e. the weight of the solid particles and the buoyancy force on the solid particlesin the hydrostatic condition. The second term in the right-hand side of Eq. 2.18 will be referredto as the seepage force per unit volume: it is the sum of the change of buoyancy, associated withthe change of the pore water pressure distribution, and the drag force at the solid-fluid interface.The seepage force per unit volume is proportional to the hydraulic gradient i = ∂φ/∂ z across16the specimen and, in case of Darcy flow, proportional to the specific discharge v. However, themeasure of seepage force per unit volume does not correctly account for the seepage-inducedchanges of effective stress. This macro-scale effect can only be appreciated by integrating theseepage force per unit volume over the relevant volume of the soil. The pore water pressure isan appropriate measure for the seepage-induced changes of effective stress.2.4 Experimental investigations of seepage-induced internalinstability2.4.1 Rigid-wall permeametersHistorically, the susceptibility of a material to seepage-induced internal instability was tested insimple rigid wall permeameters, with constant head control. The susceptibility was determinedbased on mass loss and forensic analysis of the particle size distribution at the end of the test(e.g. A˚ berg, 1993; Kenney and Lau, 1985; USACE, 1953). The permeameter of Honjo et al.(1996) in addition permitted measurement of the axial deformation. The second generation ofrigid wall permeameters included improved seepage control systems and monitoring of the porewater pressure distribution, using standpipes and measurement of the flow rate, to allow for thedetermination of hydraulic conductivity (e.g. Ke and Takahashi, 2012; Skempton and Brogan,1994), often combined with measurement of mass loss and forensic analysis of the particle sizedistribution at the end of the test (e.g. Cividini et al., 2009; Lafleur et al., 1989; Sterpi, 2003;Wan and Fell, 2004; Wittmann, 1977). Most of these studies did not apply any vertical stress,although some applied a nominal top load (A˚ berg, 1993; Honjo et al., 1996; Kenney and Lau,1985). The third and most recent generation of rigid wall permeameters incorporated an axialloading system with measurement of the axial load at the top (Chapuis et al., 1996; and thesmall rigid wall permeameter of Li, 2008), or with measurement of the axial load at the top andbottom of the specimen (Moffat and Fannin, 2006; Sail et al., 2011).2.4.2 Flexible wall permeametersSeveral flexible wall permeameters have been used to investigate seepage-induced internal in-stability (see Table 2.2). Molenkamp et al. (1979), Sanchez et al. (1983) and Sun (1989) used aflexible wall permeameter in an attempt to eliminate any preferential seepage flow paths alongthe soil-wall interface of a rigid-wall permeameter, in order to more confidently establish thesusceptibility of materials. The ability to control the stress on the test specimen appealed tomore recent investigators (e.g. Bendahmane et al., 2008; Chang and Zhang, 2011; Ke andTakahashi, 2014b; Luo et al., 2012; Xiao and Shwiyhat, 2012), all of whom sought to investi-gate the conditions at the onset of instability. For this purpose, the hydraulic gradient, flow rateand eroded mass were monitored, typically in combination with deformation of the test spec-17imen, either by means of monitoring axial or volumetric deformations. Luo et al. (2012), andKe and Takahashi (2014b), deduced volumetric deformation from local measurements usingtwo and three pairs of strain gauges, respectively, whereas Chang and Zhang (2011) deducedradial deformation from planar deformations observed on digital photographs. Accordingly,the techniques yield an indirect determination of volume change of the specimen, through localmeasurements of deformation, which leads to substantial uncertainties in the measured values.In contrast, it is common practice in triaxial shear testing of unsaturated soils to measure to-tal volumetric deformation of the specimen by monitoring the cell fluid (e.g. Ng et al., 2002;Sharma, 1998; Wheeler, 1986; Yin, 2003).2.4.3 Variables to quantify seepage-induced internal instabilityEarly investigations relied solely on change in the grain size distribution as an indicator ofinternal instability (e.g. Kenney and Lau, 1985; Molenkamp et al., 1979; USACE, 1953). Massloss is commonly used to quantify internal instability: the loss may be characterised directly bycollection of soil eroded from the specimen (e.g. A˚ berg, 1993; Adel et al., 1988; Bendahmaneet al., 2008), else indirectly from change in gradation of the test specimen (e.g. Chapuis et al.,1996; Moffat et al., 2011; Wan and Fell, 2004). On occasion both methods have been used(e.g. Honjo et al., 1996; Lafleur et al., 1989; Li, 2008). Kovacs (1981) first explicitly postulatedthat a change in local permeability accompanies the migration of fine particles out of a clast-supported micro-structure, which USACE (1953) had earlier alluded to. Lafleur et al. (1989)first used temporal and spatial changes in hydraulic conductivity, deduced from measurementsof hydraulic gradient and flow rate, as indicators of internal instability. Although measurementsof hydraulic gradient and flow rate are commonly recorded (e.g. Chapuis et al., 1996; Garnerand Sobkowicz, 2002; Li, 2008; Moffat et al., 2011; Skempton and Brogan, 1994; Wan and Fell,2004), it appears that changes in hydraulic conductivity are seldom reported in a systematicmanner. Notably, Skempton and Brogan (1994), and later Li (2008), used measurements ofhydraulic gradient and specific discharge to quantify the onset of instability. Despite an earlyappreciation for the possible rearrangement of soil structure in response to the onset of internalinstability (e.g. Kezdi, 1979; Kovacs, 1981), measurement of deformation has not receivedthe same widespread recognition that is given to measurement of mass loss, hydraulic gradientand flow rate. Only Honjo et al. (1996); Li (2008) and Moffat et al. (2011), in studies using arigid-wall permeameter, report systematic measurements of axial deformation associated withseepage-induced internal instability. The recent development of a triaxial permeameter forinvestigation of internal instability has enabled the measurement of volumetric deformation(e.g. Chang and Zhang, 2011; Ke and Takahashi, 2014b; Luo et al., 2012; Xiao and Shwiyhat,2012).182.5 Factors governing seepage-induced internal instabilityEarly experimental investigations focused primarily on the susceptibility to seepage-inducedinternal instability of soils (e.g. USACE, 1953). Subsequent investigations established therole of the particle size distribution on the material susceptibility of granular materials, whichhas yielded several useful empirical tools, exclusively based on the particle size distribution,to determine the susceptibility to seepage-induced internal instability (e.g. Burenkova, 1993;Kenney and Lau, 1985, 1986; Kezdi, 1979; Sherard, 1979). Noteworthy is that these methodsdo not distinguish between the susceptibility to suffusion or suffosion. Chapuis (1992) demon-strated that the methods of Sherard (1979), Kezdi (1979) and Kenney and Lau (1985, 1986)essentially compare the secant slope of the particle size distribution with some limiting value.Li and Fannin (2008) proposed to combine the methods of Kenney and Lau (1985, 1986) andKezdi (1979) to eliminate some of the conservatism in both methods. Wan and Fell (2008),following the work of Sun (1989), noted that the methods of Sherard (1979), Kenney and Lau(1985, 1986) and Burenkova (1993) were conservative for sand and gravel soils with silty andclayey fines, and proposed a modification of the method of the method of Burenkova (1993).The recent findings of Crawford-Flett (2014) indicate that the “[p]lasticity in the fines compo-nent of widely-graded soils results in internally stable behaviour.”Adel et al. (1988) first imposed a controlled, variable hydraulic load to determine the seep-age velocity at which internal instability initiated. Skempton and Brogan (1994) establishedthat the of onset suffusion initiated at hydraulic gradients of approximately one order of magni-tude smaller than the hydraulic gradient at which fluidisation occurs. The tests on glass beadsof Crawford-Flett (2014) confirmed this finding and related the threshold hydraulic gradient tothe sink velocity of the fine particles. The findings of Luo et al. (2012) suggest that the rate ofparticle migration increases with increasing hydraulic gradient.Skempton and Brogan (1994) postulated that the fine particles exhibit a reduced effectivestress, quantified by a reduction factor α:α =σ ′v fσ ′v(2.19)where σ ′v f = vertical effective stress on the fine particle fraction and σ ′v = vertical effectivestress.The work of Skempton and Brogan (1994) appears to have influenced Moffat (2005), whofirst systematically investigated the influence of the effective stress and hydraulic conditions onseepage-induced internal instability. Moffat (2005), using a rigid wall permeameter, identifiedsuffosion in four gap-graded soils as a marked change in local hydraulic conductivity, typically19accompanied by a marked displacement. Although the progression of suffosion was observedto be temporally and spatially variable (Moffat et al., 2011), the onset of seepage-induced fail-ure, associated with large deformations, occurred at a critical combination of hydraulic gradientand effective stress. The concept of a hydro-mechanical failure envelope (Moffat and Fannin,2011) was introduced to quantify the conditions at failure. The findings of Li (2008), whotested soil and glass beads gradations in a rigid wall permeameter, appear to confirm the valid-ity of the hydro-mechanical failure envelope. Li and Fannin (2012) noticed an apparent scaleeffect, when comparing the conditions at the onset of instability in a small permeameter anda large permeameter. However, the scale effect (see Appendix C.1) appears the unintendedconsequence of the reporting of test results in terms of the hydraulic gradient, instead of porewater pressures, which is a more suitable variable to quantify the seepage-induced change ofeffective stress (see also Section 2.3). The influence of effective stress on seepage-inducedinternal instability in clayey sand was established by Bendahmane et al. (2008), who, using aflexible wall permeameter, observed a decreasing rate of clay erosion with increasing confin-ing stress. Chang and Zhang (2013), also using a flexible wall permeameter, investigated theprogression of seepage-induced internal instability by consolidating one gap-graded gradationof soil to varying isotropic and anisotropic stress states, prior to imposing an incrementally in-creasing seepage flow. Chang and Zhang (2013) identified four phases in the erosion process:1) a stable phase; 2) an initiation phase, in which some fine particles erode but the specimen de-formation is limited; 3) a development phase, in which a large amount of fine particles washesout of the specimen, accompanied by large deformations; and 4) a failure phase. Compari-son with the previously defined seepage-induced internal instability phenomena (see Section2.1.3), suggests that the “initiation phase” is similar to suffusion, and that the “developmentphase” is similar to suffosion. The findings of Chang and Zhang (2013) indicate that the onsetof suffusion, suffosion and failure, respectively, is dependent on the mean effective stress, thedeviatoric stress and the hydraulic conditions. In contrast, Crawford-Flett (2014), in tests ontwo glass beads gradations with a lower finer fraction content, established that the onset of suf-fusion is independent of effective stress.Ke and Takahashi (2014a) investigated the shear strength characteristics of an internallyunstable soil that exhibited a substantial seepage-induced mass loss and volume change, whichyielded a strength reduction in the drained triaxial test. This finding is in broad agreement withthe previous finding of Ke and Takahashi (2012). Xiao and Shwiyhat (2012) also reported achanged undrained shear strength in soils that exhibited seepage-induced internal instability.Finally, Crawford-Flett (2014) proposed two additional necessary conditions for suffusion,in addition to the previously discussed hydraulic threshold. The particle detachment potential,quantified by α ≈ 0, was established in a clast-supported micro-structure, based on the results of20DEM (Shire and O‘Sullivan, 2013) that in this micro-structure, α is indeed approximately zero.A transportation potential was established based on the concepts of a controlling constrictionsize Dc,D95 (after Kenney et al., 1984 and Indraratna et al., 2011), defined as the constriction sizecorresponding to the 95th percentile of the theoretical densest constriction curve for the coarsefraction, and d85, defined as the particle size corresponding to the 85th percentile of the finerfraction of a gap-graded material. A gap-graded material is deemed susceptible to suffusion if:Dc,D95d85> 1 (2.20)2.6 Reflection on research hypothesesIn light of the literature review, the incomplete knowledge of the factors governing suffusionand suffosion, which was briefly alluded to in the introduction of the research hypotheses inSection 1.1, can be examined in greater detail. Herein, the knowledge gaps leading to therespective research hypothesis are discussed.2.6.1 Hypothesis No. 1The first research hypothesis was proposed as follows: Volume change is a characteristic vari-able of seepage-induced internal instability and serves to distinguish between suffusion andsuffosion. A review of the literature on seepage-induced internal instability phenomena, identi-fied volume change as one of three variables to characterise, and distinguish between, suffusionand suffosion (see Section 2.1.3). In experimental investigations using rigid wall permeame-ters (e.g. Chapuis et al., 1996; Crawford-Flett, 2014; Li, 2008; Moffat and Fannin, 2006; Sailet al., 2011), volume change could be inferred from the measurement of axial deformation. Inthe more recent experimental investigations using a flexible wall permeameter (e.g. Chang andZhang, 2011; Ke and Takahashi, 2014b; Luo et al., 2012), volumetric deformation was deducedfrom local measurements of radial and axial strain. The disparity between the volume changeas a variable to characterise, and distinguish between, suffusion and suffosion, and the indi-rect measurement of volume change in flexible wall permeameters, is addressed by researchhypothesis No. 1.2.6.2 Hypothesis No. 2The second research hypothesis was proposed as follows: The onset of suffosion is dependenton effective stress. Crawford-Flett (2014) established that the onset of suffusion is independentof effective stress. In contrast, the findings of Moffat (2005) and Li (2008) indicated thatfailure in suffosive materials is governed by a critical combination of hydraulic gradient andeffective stress. The experiments of Moffat (2005) and Li (2008) were conducted in a rigidwall permeameter, which exhibits obvious limitations in control of effective stresses. Recently,21Chang and Zhang (2013), based on tests on one granular soil gradation using a flexible wallpermeameter, suggest that the onset of suffosion is dependent on mean effective stress. Thegeneral influence of effective stress on suffosion has thus predominantly been established usingrigid wall permeameters, with limited control of effective stress, and it was broadly confirmedby tests using a flexible wall permeameter on only one soil gradation. Hypothesis No. 2 seeksto address the incomplete knowledge of the influence of the effective stress on the onset ofsuffosion.2.6.3 Hypothesis No. 3The third research hypothesis was proposed as follows: The micro-structure of internally un-stable materials controls the phenomenological response to seepage flow. Several researchers(e.g. Kenney and Lau, 1985; Kezdi, 1979; Skempton and Brogan, 1994; Wittmann, 1978) al-luded to the concept of a clast-supported micro-structure to explain seepage-induced internalinstability. Crawford-Flett (2014), based on the work of Thevanayagam et al. (2002), examinedthe characteristics of three types of micro-structure in relation with experiments, and found thata clast-supported micro-structure is susceptible to suffusion. The influence of micro-structureon suffosion appears not to have been investigated. Hypothesis No. 3 is proposed to examinethe influence of the micro-structure on suffosion, and advance the knowledge of the influenceof the micro-structure on suffusion.2.6.4 Hypothesis No. 4Finally, the fourth research hypothesis was proposed as follows: Gap-graded gradations ofrounded particles are more susceptible to seepage-induced internal instability than identicalgap-graded gradations of sub-angular particles. Although the particle shape greatly affects theengineering response of granular soils, it appears to have received no attention in the investiga-tion of seepage-induced internal instability. Hypothesis No. 4 seeks to address this knowledgegap.2.7 SummaryA review of the literature established three distinct phenomena of soils subject to seepage flow:• Migration of fine particles from a soil, termed suffusion, which is characterised as seepage-induced mass loss, without change in volume, accompanied by an increase of hydraulicconductivity; and,• Collapse of the soil structure, termed suffosion, which is characterised as a seepage-induced mass loss, accompanied by a reduction in volume, and change in hydraulic con-ductivity; and,22• Reduction of effective stress to zero in internally stable soils, termed fluidisation, whichis characterised as a seepage-induced volumetric expansion, accompanied by an increasein hydraulic conductivity, with no mass loss.The micro-structure of a soil is defined as the arrangement of the particles, or the soil fabric,and its stability. The characteristics of the micro-structure of packings of equal-size sphericalparticles are presented first, prior to a discussion on the different micro-structures that havebeen identified in binary mixtures of discrete coarse and fine fractions. The concepts of threetypes of micro-structures have been invoked in relation to seepage-induced internal instability:a coarse-particle dominated clast supported micro-structure, a fine-particle dominated matrixsupported micro-structure and a transitional micro-structure. The role of the particle shape onthe stability of the arrangement of particles is briefly described.Following a description of the seepage flow in granular materials, the characteristics ofexperimental investigations on seepage-induced internal instability are presented. Seepage-induced internal instability is commonly quantified by a change in particle size distribution,mass loss from the specimen, a change in hydraulic conductivity, axial deformation and radialdeformation. The ability to control the effective stress has driven the recent development offlexible wall permeameters.Several empirical tools exist to assess the susceptibility of a material to seepage-inducedinternal instability. The onset of suffusion initiates at a relatively small hydraulic gradient, in-dependent of the effective stress. The occurrence of suffosion is temporally and spatially vari-able. The onset of failure in internally unstable materials, associated with large deformations,appears dependent on the hydraulic load and the effective stress. Finally, the knowledge gapsassociated with the research hypotheses are discussed in light of the review of the literature.23Table 2.1: Descriptions of seepage-induced internal instability phenomena (source: Fannin and Slangen (2014), with permission from ICEPublishing).Study Phenomenon DescriptionUSACE (1953) Inherent stability “Inherent stability was determined by the degree to which the gradation curvesof the sections compared with those of the original material”Wittmann (1978) Suffosion “[..] mixtures with a gravel skeleton [..] permit transport of sand particles into,inside and out of the skeleton. These phenomena are known as suffosion (trans-port out of the skeleton) and colmatation or sluicing (transport into the skeleton).These processes influence the permeability of the mixture [..] The skeleton re-mains stable.”Erosion “For pure sand or only small amounts of the gravel component this [deforma-tion] phenomenon is known as piping by heave, whereas mixtures with a higheramount of coarser particles show the phenomenon of erosion piping. Erosionleads to collapse of the mixture, with the finer particles drawn away by the seep-ing water.”Kezdi (1979) Suffusion “Suffusion is a phenomenon where water, while seeping through the pores, car-ries along the fine particles without destroying the soil structure.”Erosion “Erosion [..] destroys the soil structure. Not only are single grains or fractionsdisrupted, but the whole soil structure is progressively destroyed and tubelikecavities are formed.”Continued on next page24Table 2.1 – Continued from previous pageStudy Phenomenon DescriptionMolenkamp et al.(1979)Suffosion “[..] migration of the finer material of the very filter layer [..]”Internal instability “The internal stability of the filter material could possibly be checked by com-paring the grain size distributions of the top and bottom part of the column”Sherard (1979) Internal erosion stabil-ity“[..] a concentrated leak developed through the core which caused a type ofinternal erosion in which the soil fines are eroded selectively and carried out ofthe core, leaving the coarse sand and gravel particles behind to act as a perviousdrain. The volume of fine material eroded was larger than the volume of thevoid spaces between the coarser soil particles causing progressive collapse ofthe material above the initial leakage channel, which action finally reaches thedam surface as manifested by the sinkhole or crater.”Jones (1981) Suffosion “To Pavlov (1898) and Savarensky (1940) [..] suffosion meant mechanical re-moval of loose particles [..] Both mechanical and chemical forms of suffosionprocess were described by Russian, Polish and French workers [..]”Kovacs (1981) Suffusion “Redistribution of fine grains within the layer [..] when the solid volume of thelayer is not changed only the local permeability is altered”Internal suffusion “[..] when the solid volume of the layer is not changed only the local permeabil-ity is altered”External suffusion “[..] when the volume of the solid matrix is reduced, accompanied by an increasein permeability”Continued on next page25Table 2.1 – Continued from previous pageStudy Phenomenon DescriptionDestruction of theskeleton“Subsidence of the layer, when some of the coarse grains are removed from thesolid matrix, and thus the load of overlying layers causes the total volume of thelayer to decrease”Kenney and Lau(1985)Internal instability “Because of its wide grain size distribution, the small grains can easily be washedout, through the skeleton of the large grains.”Lafleur et al. (1989) Internally unstable “[..] substantial migration took place within the layers [..] This trend was sup-ported by the permeability curves [..]”A˚ berg (1993) Grading instability “[..] the material has loose grains, which are so small that they can pass throughthe constrictions between fixed grains [..] loose grains should be consideredpotentially unstable”Internal filter forma-tion process or self-filtration“[..] when this process proceeds in a [..] upward direction [..] shrinkage ofthe material during washout causes movement in the overlying material [..] andthereby also make washout of fixed grains possible”Burenkova (1993) Inner suffosion “Inner suffosion takes place, when fine particles are transported in the soil struc-ture [..] This process can lead to an increase of the permeability of the soil [..]”Outer suffosion “Outer suffosion means transportation of fine fractions out of the soil [..] Thisleads to an increase of the void ratio, the permeability and [..] to instability ofthe wh[o]le structure.”Internal instability “The tests showed particle movements along the interface of the test apparatusand the soil specimen under seepage flow. The readings of the piezometers al-tered during the tests and particles eroded.”Continued on next page26Table 2.1 – Continued from previous pageStudy Phenomenon DescriptionSkempton and Brogan(1994)Internal instability “[..] the sand can migrate within the interstices of a framework of primary fabricformed predominantly of the gravel particles and can be washed out [..]”Segregation piping “[..] failure took the form of piping of the fines whereas the gravel particlesremained practically undisturbed [..] In the stable materials this occurred at ap-proximately the critical gradient given by piping theory [..], but in the unstablematerials migration and strong piping of fines took place at gradients of aboutone fifth to one third of the theoretical value [..]”Onset of instability “A further increase in gradient causes a disproportionate increase in flow, leadingtowards failure either by piping [..] or by the opening of a horizontal crack [..]which then works its way upwards until piping occurs throughout.”Chapuis et al. (1996) Suffossion “[..] migration of fine particles of a soil within its own pore space.”Internal instability “If a 0-20 mm base has an internal instability problem [..] creation of smalllayers with high capillarity and low permeability [..] are likely to develop withtime.”Internal segregation “The internal segregation of particles was evaluated by wet sieve analysis” and“[..] changes [in permeability] appeared with [internally unstable] gradation 2.”Chapuis and Aubertin(2004)Suffusion “[..] a permeating process, often a fluid movement towards a surface or over asurface; thus, using it for internal erosion would be incorrect [..]”Suffosion “[..] this word is not found in English and French dictionaries”Suffossion “[..] correctly represents the phenomenon of internal erosion. It is unfortunatethat it has been forgotten after having been used a long time ago by militaryengineers as an undermining technique.”Continued on next page27Table 2.1 – Continued from previous pageStudy Phenomenon DescriptionHonjo et al. (1996) Self-filtration “As flow starts, the finer base particles pass through the filter but the coarserparticles are caught at the base soil-filter interface quickly forming a thin denselayer [which] consists of the coarser particles. The rest of the base soil is thenprotected by this layer [..]”Stable “The loss of base soil is observed for a limited time and the development of [a]self-filtration layer is prominent as observed from the after test sieve analysis”Unstable “Loss of base soil is observed to be continuous and the whole sample is sub-ject to disturbance”; “self-filtration layer is absent in this case”, also related to“progressive settlement”Garner and Sobkowicz(2002)Suffosion “[..] the mass movement of the fine fraction within the skeleton of a dispersed,potentially unstable coarse fraction”, associated with an increase of permeabilitySuffusion “[..] the redistribution of fine grains within a stable densely packed skeleton” andassociated reduction in permeability, referring to the term “internal suffusion”used by Kovacs (1981)Wan and Fell (2004) Suffusion or internalinstability“[..] an internal erosion process which involves selective erosion of fine particlesfrom the matrix of a soil made up of coarse particles”Unstable “[..] a change in the grain-size distribution of the test sample after the test”; “[..]signs of erosion [..] are observed as the hydraulic gradient across the test sampleincreases”; and “The colour of the effluent provided an indication of whether ornot erosion was taking place [..]”Continued on next page28Table 2.1 – Continued from previous pageStudy Phenomenon DescriptionStable “[..] the grain-size distribution of the test sample will remain unchanged afterthe test”Richards and Reddy(2007)Suffosion “This process can result in a loose framework of granular material with relativelyhigh seepage flows that leads to collapse of the soil skeleton (McCook, 2004). Innon-cohesive materials suffosion leads to zones of high permeability (and watertransmission), potential outbreaks of increased seepage, increased erosive forcesand potential collapse of the skeletal soil structure.”Suffusion “Gradual loss of finer matrix materials in a soil supported by a coarser grainedskeleton is termed suffusion, which may lead to a more general collapse and lossof soil structure, termed suffosion (Kezdi 1979; McCook 2004).”Fell and Fry (2007) Suffusion or internalinstability“[..] selective erosion of fine particles from the matrix of coarser particles [..]by seepage flow, leaving behind an intact soil skeleton formed by the coarserparticles”Li (2008) Internal suffosion “[..] refers to the redistribution of finer particles within layers that is accompa-nied by a change in local hydraulic conductivity”. “No displacement or massloss was observed.”External suffosion “[..] refers to the scouring of finer particles that is associated with an overallincrease in hydraulic conductivity”. “A contractive displacement and mass losswas observed”Bendahmane et al.(2008)Suffusion “[..] the permeability [..] decreased by a factor of 10 during the tests whereerosion was initiated [..] characterized by some diffuse mass losses”Continued on next page29Table 2.1 – Continued from previous pageStudy Phenomenon DescriptionBonelli and Marot(2011)“Suffusion (or suffos-ion)”“[..] an internal erosion process by which finer soil particles are detached fromthe solid matrix and transported through pore constrictions by seepage flow.”Moffat et al. (2011) Suffusion “[..] the finer fraction of an internally unstable soil moves within the coarserfraction without any loss of matrix integrity or change in total volume [..] Thevisual observations, and the companion spatial and termporal variations of localhydraulic gradient, reveal a transport of finer particles from the soil with eachincrement of hydraulic gradient that yields a relatively slow and small change inlocal permeability, but no change in volume.”Suffosion “[..] particle migration yields a reduction in total volume and a consequent po-tential for collapse of the soil matrix (Richards and Reddy, 2007).” “Visual ob-servations and the companion spatial and temporal variations of local hydraulicgradient, reveal a particle loss that yields a relatively large and rapid change inlocal permeability and a companion change in specimen volume.”30Table 2.2: Overview of selected studies on seepage-induced internal instability using a flexible wall permeameter.Measurements 1Publication Material Specimensize (cm)2Direction ofseepage flow3Eroded mass Axial defor-mationsVolumetricdeformationB-valueMolenkamp et al. (1979)Sand and gravel D = 45l = 100DF no no no noSanchez et al. (1983)Silt and clay D = 7l = 15.5DF yes no no noSun (1989)Sand and clayeysiltD = 7l = 2.5DF, UF yes no no noBendahmane et al. (2008)Clayey sand D = 5l = 5DF yes no no yesChang and Zhang (2011)Sand and gravel D = 10l = 10UF, DF yes yes yes4 ≈ 0.85Luo et al. (2012)Sandy gravel D = 10l = 10DF yes yes yes5 noXiao and Shwiyhat (2012)Clayey sand D = 5.1l = 10.2DF yes yes yes6 > 0.95Ke and Takahashi (2014b)Sand D = 7l = 15DF yes yes yes5 > 0.95This study (2014) Glass beads andsoilsD = 10l = 10UF no yes yes7 > 0.95Notes:1 All studies measured flow rate and hydraulic pressure difference across the specimen2 D, specimen diameter and l, specimen length3 UF = Upward flow, DF = Downward flow4 Using a photographic method;5 Derived from strain gauges6 By means of monitoring the volume of the inflow and outflow;6 Monitoring of volume change of cell fluid31Figure 2.1: Non-destructive seepage-induced internal instability in a test conducted bySkempton and Brogan (1994) (source: Fannin and Slangen (2014), with permissionfrom ICE Publishing).Figure 2.2: Destructive seepage-induced internal instability in a test conducted by Li(2008) (source: Fannin and Slangen (2014), with permission from ICE Publishing).32Figure 2.3: Seepage-induced instability in a test conducted by Li (2008) (source: Fanninand Slangen (2014), with permission from ICE Publishing).33Onset(of(seepage-induced(instabilityΔm(>(0ΔV/V(<(0(ΔV/V(>(0Δm(>(0SuffusionSuffosionFluidisationΔk(>(0Δk(>(0(or(Δk(<(0(Δk(>(0ΔV/V(=(0(Δm(=(0Prior(to(onset(of(seepage-induced(instability(a)(b)(c)Figure 2.4: Schematic illustration of seepage-induced instability phenomena: (a) suffu-sion; (b) suffosion; and (c) fluidisation (source: Fannin and Slangen (2014), withpermission from ICE Publishing).34Mass loss(Δm)Change in hydraulic conductivity(Δk)Volume change(ΔV)SuffusionSuffosionFluidisationSeepage-induced internal instabilityFigure 2.5: Conceptual framework for seepage-induced instability phenomena (source:Fannin and Slangen (2014), with permission from ICE Publishing).Total voidseFine-grain solidsSfCoarse-grain solids1-SfApparent voidseFine-grain solidsSfApparent voidse + SfCoarse-grain solids1-Sfe = void ratio ef = inter-fine void ratioes = inter-granular void ratioef = e / Sf es = (e + Sf) / (1 - Sf)Figure 2.6: Inter-coarse and inter-fine void ratios (adapted from Thevanayagam, 1998,with permission from ASCE).35Figure 2.7: Micro-structure cases (source: Thevanayagam et al., 2002, with permission from ASCE).36Figure 2.8: Definitions of particle size and shape (source: Altufi et al., 2013, with permis-sion from ASCE).Figure 2.9: Plot of sphericity versus convexity for determination of particle shape (source:Altufi et al., 2013, with permission from ASCE).37Chapter 3Apparatus, materials and testingprogramThis Chapter addresses the design of a new flexible wall permeameter (research objective No.2), that was developed for this experimental investigation, the test procedure, the selectionof the materials, and the test program. The test device, with its novel measurement of totalvolume change, is described in detail first (see Section 3.1), followed by a characterisation ofthe materials, comprising glass beads and soils, in Section 3.2. The test procedure is presentedin Section 3.3, followed by the presentation of the test program in Section 3.4. A summary ofthis Chapter is provided in Section 3.5.3.1 Flexible wall permeameterA flexible wall permeameter was designed by the author and was custom-made at the CivilEngineering workshop at UBC. The general configuration of the flexible wall permeameter isillustrated in Fig. 3.1. The term “flexible wall permeameter” is selected in agreement with theterminology used by ASTM D5084-10 (ASTM, 2010). The device comprises a double-walledtriaxial cell, which is assembled in a large water bath; a seepage control system, through whichunidirectional upward multi-stage seepage flow can be imposed; and instrumentation.3.1.1 Double-walled triaxial cellThe concept of a double-walled triaxial cell was originally introduced by Bishop and Donald(1961) to determine volume change during shear of an unsaturated specimen: the double-walledarrangement enables measurement of volume change of the cell fluid, from which the volumechange of the specimen is deduced, independent of changes in cell pressure. The elegant de-sign of the double-walled triaxial cell reported by Wheeler (1986), with its improved accuracyin the measurement of volume change, forms the basis for the triaxial cell developed for thisinvestigation.38The use of an inner acrylic tube (see Fig. 3.2), with internal diameter of 206 mm and lengthof 440 mm, and an outer acrylic tube, with internal diameter of 236 mm and length of 440mm, creates two chambers: an inner chamber, in which the specimen is located, and an outerchamber, between the two tubes. The use of acrylic tubes permits visual observations duringtesting, which proved to be very useful when interpreting the test results. The inner and outerchamber ports are connected to a common pressure source (see Fig. 3.3), through a parallelarrangement of two air/water interface measurement burettes. By means of pressurising bothchambers to an equal cell pressure (see Fig. 3.4), the differential pressure across the inneracrylic tube is reduced to zero, thus eliminating any cell pressure dependent volume change ofthe inner chamber.For this study, all specimens were reconstituted and subject to isotropic consolidation (seeSection 3.3). The axial loading system consists of a rod through which loading is applied tocompensate for the combined effect of buoyancy force from the cell pressure, and friction inthe two sets of linear ball bearings mounted on the top plate. A U-cup provides a water tightseal between the inner chamber and the loading ram.3.1.2 Membrane, base pedestal and top capThe double-walled triaxial cell accommodates a cylindrical test specimen with diameter of 100mm. The specimen is reconstituted to a height of approximately 100 mm, yielding a volumeof approximately 780 cm3, which appears a typical size compared to other flexible wall perme-ameters used in experimental investigations of seepage-induced internal instability (see Table2.2). The specimen is enclosed by a rubber membrane with dimensions 10.2 cm x 0.03 cm x30.5 cm, which is sealed to the base pedestal and top cap, respectively, using O-rings. Fig. 3.5shows a dummy specimen in the cell.In contrast to many other flexible wall permeameters (see Table 2.2), this device is config-ured for upward seepage flow in order to physically separate the basal support of the specimenfrom its outflow boundary, at the bottom and top of the specimen, respectively. The upwardseepage flow configuration allows for selection of relatively fine wire meshes for the basalsupport of the specimen, which prevent mass loss during specimen reconstitution. The disad-vantage that mass loss during multi-stage seepage flow cannot be measured in this configurationis mitigated by obtaining forensic observations of mass loss at the end of a test.The conceptual design of the hollow base pedestal and the hollow top cap are similar tothe designs of Sun (1989) and Chang and Zhang (2011): they are designed to allow unimpededupward seepage flow. The cavity in the base pedestal is shaped as a inversed cone (see Fig.393.6): the inner diameter increases from 10 mm at the bottom, where the opening connects tothe inlet port in the base frame, to 90 mm at the top. A small port in the base pedestal connectsto measurement port (MP) #1 in the base frame, to which the base pedestal is attached with anair-tight seal using bolts and O-rings. A narrow rim at the top of the base pedestal providessupport for a perforated plate with 5 mm diameter holes, spaced 10 mm centre-to-centre. Twobottom wire meshes are placed on the perforated plate: the primary bottom wire mesh has asmall opening size, which prevents particle loss during specimen reconstitution; a secondary,coarser and stiffer bottom wire mesh keeps the primary wire mesh in place. A gasket of PVCmaterial seals the interface between the bottom wire meshes and the rim of the base pedestal.Two wire meshes are placed against a perforated plate mounted in the top cap, which sits onthe top surface of the reconstituted specimen. The primary top wire mesh retains the coarseparticles and the secondary, coarser wire mesh provides spacing for migration of washed-outparticles between the primary top wire mesh and the perforated plate. A gasket of PVC materialseals the interface between the top wire meshes and the top cap. The selection of the openingsizes of the wire meshes is dependent on the particle size distribution of the material( see Sec-tion 3.4).The top cap with an outer diameter of 100 mm sits on the top surface of the reconstitutedspecimen. The top cap has a cylindrical cavity with inner diameter of 90 mm and height of50 mm, which serves as a settling reservoir for any washed-out particles. Three ports (see Fig.3.6) connect to different ports in the base frame using flexible nylon tubes: two ports connectto the outlet port, whereas the third port connects to MP # 2 (see Fig. 3.3). A small perforatedplate and a fine tertiary wire mesh, mounted in the top cap (see Fig. 3.6), prevent washed-outparticles from entering the seepage control system.3.1.3 Seepage control systemThe seepage control system is a recirculation circuit based on the principle of multi-stage headcontrol. The principle of head control is commonly applied in permeameter testing (e.g. Ben-dahmane et al., 2008; Chang and Zhang, 2011; Luo et al., 2012; Moffat and Fannin, 2006),and is conceptually different from the principle of flow control adopted by Ke and Takahashi(2014b). The total dynamic head (TDH) (see Fig. 3.1) between the free water surface in the in-flow constant-head device (I-CHD), which connects to the inlet port in the base frame (see Fig.3.3), and the free water surface in the outflow constant-head device (O-CHD), which connectsto the outlet port in the base frame, yields upward seepage flow through the specimen. Theelevation of the wall-mounted I-CHD is changed by a manually controlled winch. In contrast,the elevation of the O-CHD is fixed. The maximum total dynamic head, TDHmax that can beimposed in this manner is limited to 165 cm by the geometry of the apparatus and the height ofthe laboratory ceiling.40A supply of filtered, de-aired water is prepared from tap water in the manner described byMoffat and Fannin (2006): tap water is passed through a sand filter and a carbon filter (see Fig.3.1), removing suspended solids larger than 10µm and 3µm, respectively. The filtered water isthen dripped into a de-airing tank where it is stored under a vacuum of approximately -70 kPafor a minimum of 8 hours, which experience has shown yields a dissolved oxygen content lessthan 2.5 mg/L. At the beginning of a test, the storage reservoir (see Fig. 3.1) is filled with thefiltered de-aired water. During multi-stage seepage flow, this water is recirculated to the I-CHDusing a peristaltic pump.3.1.4 InstrumentationAn overview of the instrumentation scheme is presented in Fig. 3.3. One transducer is usedto control the cell pressure and five additional transducers are used to monitor the specimenresponse during consolidation and multi-stage seepage flow. The output voltages are auto-matically recorded at a frequency of 20 Hz and time-averaged over 1 s intervals for improvedprecision using a computer connected to a data acquisition system (DAQ) (see Fig. 3.1).The accuracy, precision and resolution are discussed in detail in Appendix A. In summary,the accuracy, a measure for the error associated with a measured value, is defined as the standarddeviation of the calibration data around the linear regression line; the precision, associated withrepeatability, is defined as the standard deviation of the measured values around the mean ofthe measured values; and the resolution, or the smallest significant change that can be detected,is assumed as four times the precision. Importantly, the accuracy of the derived variables is afunction of the value of the variable (see Table A.3).3.1.4.1 Cell pressureThe cell pressure is measured using a total pressure transducer (TPT) #1, which connects to theinner chamber port (see IC-P on Fig. 3.3). TPT # 1 is a model PDCR 130/w/c manufacturedby Druck, with a pressure range of 0 to 700 kPa, accuracy of +/- 0.2 kPa, precision of 0.02 kPa,and resolution of 0.08 kPa.The air pressure to the air/water interface measurement burettes #2, #3 and #4 is regulatedusing two manual regulators, the second of which has a smaller working range. The regulatorsare connected in series to facilitate sensitive control of the cell pressure.413.1.4.2 Pore water pressure and differential pore water pressureThe pore water pressure during consolidation of the specimen is measured using TPT #2, whichconnects to the base pedestal via measurement port MP #1. TPT #2 is a model 111 manufac-tured by GP:50, with a pressure range of 0 to 1050 kPa, accuracy of +/- 0.5 kPa, precision of0.1 kPa, and resolution of 0.4 kPa.The differential pore water pressure across the specimen, ∆u, is measured using a differen-tial pressure transducer (DPT) #1: the low and high pressure ends are connected to the top cap(see Figs. 3.3 and 3.6), via MP #2, and to the base pedestal, via MP #1, respectively. DPT #1 isa model PDW/E972-05-01, manufactured by Sensotec, with a range of +/- 70 kPa, accuracy of+/- 0.1 kPa, precision of 0.003 kPa, and resolution of 0.012 kPa. A value of the mean effectivestress, p′, is derived from measurement of the cell pressure and mean pore water pressure in thespecimen.3.1.4.3 Flow rateFlow rate through the specimen is measured manually by collecting the mass of the water thatflows from the O-CHD (see Fig. 3.1) over an increment of time. The accuracy of the manualmethod is +/- 0.0004 cm3/s, which is derived from the uncertainties of the measured quantitiesof mass of collected water and elapsed time (see Appendix A). For the precision and resolutionof the flow rate measurement, conservative values equal to the accuracy are assumed. Thespecific discharge, v, is derived from the flow rate measurement assuming a constant cross-sectional area of the specimen.3.1.4.4 Axial displacementThe specimen height during reconstitution, and change of height during consolidation, is mon-itored using a dial gauge with an accuracy of +/- 0.01 mm. The axial deformation duringmulti-stage seepage flow, from which the axial strain εa is derived, is measured using a lin-ear variable differential transformer (LVDT), mounted on the loading ram (see Fig. 3.2). TheLVDT is a model TS 25, manufactured by Novotechnik. with a range of 0 to 25 mm, accuracyof +/- 0.01 mm, precision of 0.0008 mm, and resolution of 0.0032 mm.3.1.4.5 Volume changeThe volume change of the test specimen during consolidation is monitored in measurementburette #1. The measurement is corrected for membrane compliance using the single specimenmethod proposed by Vaid and Negussey (1984), see Appendix B.A novel feature of the device is the measurement of the total volume change during multi-42stage seepage flow. The arrangement of the double-walled triaxial cell eliminates the pressure-dependent volume change of the inner chamber. Thus, volume change of the inner chamberis only caused by: 1) tiny pockets of air trapped underneath the top plate, which dissolve inwater over time; 2) absorption of water into the acrylic tube with time; 3) volume change ofthe specimen; and 4) movement of the loading ram (Wheeler, 1986). The presence of air inthe inner chamber is significantly reduced by assembling the double-walled triaxial cell in alarge bath filled with de-aired water (see Fig. 3.1). Absorption of water into the acrylic tubesis minimized by maintaining a high degree of saturation of the acrylic tubes by storing them inwater when they are not used for testing.Calibration tests indicate that the rate of volume change attributed to the actions of dissolv-ing air and absorption of water is very small and fairly repeatable: the rate of volume changereduces to 0.036 +/- 0.025 cm3/h after an elapsed time of approximately 16 h (see Fig. 3.7).Therefore, upon completion of consolidation, the assembled apparatus is left standby for a min-imum of 16 h prior to imposing the first stage of seepage flow. The volume change of the innerchamber is derived from the measurement of the water level in measurement burette #2 (seeFig. 3.3), using DPT #2, similar to the technique used by Tatsuoka (1981): the low-pressureend of DPT #2 is connected to the air pressure supply, while the high-pressure end of DPT#2 is connected to the bottom of measurement burette #2. The working range of measurementburette #2 is 20 cm3 and it can be recharged from the larger measurement burette #3. DPT #2is a model 316, manufactured by GP:50, with a range of 0-14 kPa, accuracy of +/-0.02 kPa,precision of 0.004 kPa, and resolution of 0.012 kPa. The resolution of DPT #2 and the innersurface area of measurement burette # 2 (0.47 cm2), govern the resolution of the volume changemeasurement, which is 0.06 cm3 or approximately 0.01% volumetric strain.The total volume change of the specimen is determined by monitoring the volume changeof the fluid in the inner chamber and correcting it for: the intrusion of the loading ram; mem-brane penetration; and the calibrated, time-dependent volume change arising from the actionsof dissolving air and and absorption of water. The accuracy of the volume change measurementsystem is derived from the uncertainties of the measured quantities (see Appendix A); the accu-racy is +/- 0.5 cm3, or approximately +/- 0.07% volumetric strain, εv, for a multi-stage seepageduration of less than 8 hours.3.2 MaterialsThe materials tested in the laboratory investigation were selected to investigate the influence ofthe particle size distribution and the particle shape on seepage-induced internal instability. Ninegap-graded mixtures of glass beads and ten gap-graded gradations of angular soil particles withvarying finer fraction content, S f , and D′15/d′85-ratios were produced from uniformly graded43component-fractions. The code used for the gap-graded gradations are generated as follows:the first two numbers are the particle size ratio of the coarse and the fine fractions, D′15/d′85, fol-lowed by two letters indicating the type of material (GB for glass beads or BT for soil particles);the last two numbers denote the percentage fine fraction. For example, gradation 4.8GB20 is agap-graded gradation of glass beads with D′15/d′85 = 4.8 and S f = 0.20.3.2.1 Glass beadsGlass beads (GB) were selected for the main program of testing, because of their shape and rel-atively easy handling. Spherical particles have the benefit that extensive experimental studies(e.g. McGeary, 1961; Scott, 1960; Scott and Kilgour, 1969) and DEM studies (e.g. Shire andO‘Sullivan, 2013) have been conducted on packing arrangements of such particles.The glass beads, manufactured by Potters Industries Inc., are specified as having a soda-limesilica glass composition, with density ρ = 2.5 g/cm3, coefficient of static friction µ = 0.9 to 1.0and roundness R = 0.8 to 0.9. The specific gravity of the glass beads Gs = 2.49 was determinedfollowing ASTM D854-14 (ASTM, 2014). The glass beads components used to prepare gap-graded mixtures are fine component GB-F and coarse components, GB-C1, GB-C2, GB-C3and GB-C4 (in order of increasing particle size, see Fig. 3.8). The particle size distributionand characteristics of the particle shape of all components except GB-C1 (see Table 3.1 andFig. 3.9), were quantified by the author using the QicPic apparatus (Sympatec, 2008) duringa research visit to Imperial College London. All components were uniform, with Cu rangingfrom 1.1 to 1.25, and Cc ranging from 0.99 to 1.03. The particle shapes (see Section 2.2.2.1)of the three coarse components GB-C2, GB-C3 and GB-C4 were nearly identical with AR =0.95 to 0.97, Cx = 0.98, and SQP = 0.94. The particles of these components are almost perfectlyspherical, which can be observed on the microscopic image of GB-C4 (see Fig. 3.10). Theparticle shape of the fine component GB-F is slightly less perfectly spherical (see Fig. 3.11),with AR = 0.91, Cx = 0.94, and SQP = 0.94. Based on the indices of particle shape, all glassbeads components would qualify as rounded (see Fig. 2.9). The minimum index void ratioemin = 0.57 and maximum index void ratio emax = 0.67 of component GB-C1 were determinedusing ASTM D4253-14 (ASTM, 2015b) and ASTM D4254-14 (ASTM, 2015a), respectively.These values are also assumed for the other coarse GB components, given the similar particleshape and uniformity of the particle size distribution. Nine gap-graded glass gradations werefabricated (see Table 3.2 and Figs. 3.12 to 3.15). Additionally, component GB-F was testeddirectly as gradation GB-F to commission the device, and the results serve as a benchmark forcomparative analysis: it is uniformly graded with d0 = 0.12 mm, d100 = 0.20 mm and d85 =0.19 mm.443.2.2 SoilsIn order to investigate the influence of the particle shape, gradations of sub-angular soil par-ticles were fabricated with particle size distributions that were nearly identical to some of theglass beads mixtures. The sub-angular soil particles (BT) originate from a filter in a zonedembankment dam. The bulk sample of filter material was separated into uniform componentsby means of mechanical sieving.The specific gravity of the angular soil particles Gs = 2.70 was determined following ASTMD854-14 (ASTM, 2014). The BT components used to prepare the gap-graded mixtures are finecomponent BT-F and coarse components BT-C1, BT-C2, BT-C3, BT-C4 and BT-C5 (see Fig.3.16). The particle size distribution and characteristics of the particle shape of the individualcomponents, were quantified using the same QicPic apparatus at Imperial College London (seeTable 3.1 and Fig. 3.17). All fine and coarse components were uniform, with Cu ranging from1.27 to 1.38, and Cc ranging from 1.00 to 1.03. The indices of particle shape of the coarsecomponents fall within a very narrow range with AR = 0.73 to 0.74, Cx = 0.96 to 0.97 , andSQP = 0.87 to 0.88. The particle shape of the fine component is very similar, with AR = 0.76,Cx = 0.97, and SQP = 0.88. Microscopic images of BT-F and BT-C3 are presented in Figs. 3.18and 3.19, respectively. Based on the measures of particle shape, all BT components qualify assub-angular (see Fig. 2.9). The minimum index void ratio emin = 0.63 and maximum indexvoid ratio emax = 0.80 of component BT-C1 were determined using ASTM D4253-14 (ASTM,2015b) and ASTM D4254-14 (ASTM, 2015a), respectively. These values are assumed for theother coarse BT components, given the similar particle shape and uniformity of the particle sizedistribution. Ten gap-graded gradations of sub-angular soil particles were fabricated (see Table3.2 and Figs. 3.20 to 3.24).3.3 Test procedureFollowing preparation of the specimen, it is consolidated and then subject to multi-stage seep-age flow. Each step is described in detail herein.3.3.1 Specimen preparationThe objective of the preparation method is to produce a saturated, homogeneous specimen. Themethod adopted for this study involves: 1) thoroughly mixing the coarse and fine componentsof glass beads or soil, and saturating the mixture by boiling it in de-aired water for a minimumof 30 min and, after cooling to room temperature, storing it in a vacuum desiccator for at least12 hrs; 2) reconstituting the specimen in a forming mold that supports the triaxial membraneon the base pedestal; 3) placing the hollow top cap on the leveled top surface of the specimen;4) sealing the membrane against the top cap with a rubber O-ring, and applying a vacuum pres-45sure of approximately -20 kPa to the specimen through the air/water interface in measurementburette #1 (see Fig. 3.3); 5) removing the forming mold and assembling the flexible wall per-meameter; and 6) applying a cell pressure of approximately 20 kPa and subsequently releasingthe vacuum pressure.More specifically, on the matter of specimen reconstitution, the uniformly graded GB-Fmaterial is reconstituted using the water pluviation technique (Fannin et al., 1994; Vaid andNegussey, 1988). In contrast, the gap-graded glass beads and soils are reconstituted usingthe modified slurry deposition technique (Moffat and Fannin, 2006). The ability of the slurrydeposition technique to produce saturated, homogeneous specimens of angular tailings sandhas been demonstrated by Kuerbis and Vaid (1988). The free water surface in the I-CHD and inthe O-CHD (see Section 3.1.3) does not permit the application of a back pressure during multi-stage seepage flow. Therefore, the B value was determined without application of additionalback pressure following step six of the specimen preparation method. All gradations testedin this study exhibited B-values greater than 0.95 (see Tables 4.1 and 4.3), indicating fullysaturated specimens (ASTM, 2011). The homogeneity of the slurry deposited gap-graded glassbeads materials was evaluated by excavating, in five layers of equal thickness, trial specimensreconstituted solely for this purpose. The homogeneity of the materials 4.8GB20, 6.0GB35,6.5GB25 and 6.5GB35 was achieved with a variation in the finer fraction content of S f = +/-0.03 (see Figs. 3.25, 3.26 and 3.27, respectively). The relatively large spatial variations of S f= +/- 0.08 in two trial specimens of gradation 6.0GB20 (see Figs. 3.28 and 3.29) demonstratesthat the homogeneity of this material, with its relatively low fines content and relatively largeratio of particle sizes, cannot be guaranteed. Accordingly, the findings suggest that the modifiedslurry deposition technique is capable of producing a homogeneous saturated specimen of gap-graded glass beads gradations with: S f = 0.25 to 0.35 and D′15/d′85 ≤ 6.5; and S f = 0.20 andD′15/d′85 ≤ 4.8. As the upper limit for a reasonably homogeneous specimen of glass beads withS f = 0.20, D′15/d′85 = 6.0 is considered appropriate.3.3.2 Stage 1: ConsolidationTest specimens are consolidated isotropically at a cell pressure of 50, 100 or 150 kPa. The re-sulting excess pore water pressure dissipates under a condition of single drainage in the upwarddirection.3.3.3 Stage 2: Multi-stage seepage flowHydraulic head is applied across the specimen by incrementally raising the elevation of the I-CHD (see Fig. 3.1) in a multi-stage test procedure, to impose upward seepage flow through thespecimen. Energy losses occur in the fittings and tubes, which reduce the TDH to the differen-tial head across the specimen (∆u / γw). Accordingly, the mode of operation of the flexible wall46permeameter is one of TDH-control, rather than control of the differential pore water pressureacross the specimen ∆u. Relatively small increments of ∆TDH = 1 cm were imposed in theinitial 4 to 5 stages of a test, in order to quantify the pre-critical response; a test is terminatedat TDHmax =165 cm, or when the range of the volume change measurement is exceeded. In theintermediate stages, ∆TDH = 5 to 20 cm, depending on the hydraulic conductivity of the speci-men. The duration of each stage ∆t is typically 10 to 60 min to permit at least two independentmeasurements of flow rate.The variation of effective stress of the specimen, isotropically consolidated and subse-quently subject to upward multi-stage seepage flow, is discussed in Appendix C. The effectivestress of a specimen diminishes with subsequent stages of seepage flow.3.4 Test programAn overview of the test program is presented in Fig. 3.30. The test codes are constructed asfollows: “gradation” + “target cell pressure” + “(R)” (optional). The designation “(R)” is usedwhen a test is repeated. For example, test code 4.8GB20-50(R) refers to the repeated test ongradation 4.8GB20 consolidated at a target cell pressure 50 kPa.Commissioning tests were conducted on uniform gradation GB-F and gap-graded gradation6.5GB25, isotropically consolidated at cell pressures of 100 kPa, respectively. The objectivesof these commissioning tests were to:• verify the reliability of the measurements in the flexible wall permemaeter, especially thevolume change measurement.• determine the seepage regime, including hydraulic conductivity, of the fine componentof the gap-graded glass beads mixtures.In the main test program, all specimens were isotropically consolidated and then subject toupward multi-stage seepage flow. The objectives of the main test program were to:• assess the repeatability of the test results.• gather experimental data to investigate the influence of the effective stress, particle sizedistribution, and particle shape on suffusion and suffosion (see research hypotheses Nos.1 to 4 in Section 1.1).The 23 tests on glass beads mixtures comprise the majority of the test program. The testson gradations 4.8GB20 were conducted to examine the influence of the effective stress on gra-dations with S f = 0.20, which were anticipated to yield a clast-supported micro-structure. Test4.8GB20-50(R) was conducted to examine the repeatability of the test results on gradations47with S f = 0.20 Additionally, one test was conducted on 3.3GB20 in an attempt to establish, atS f = 0.20, the limit of D′15/d′85 at which internal instability could occur. The tests on gradations6.0GB25, 6.0GB30 and 6.0GB35 were conducted to investigate the influence of increasing S fwith constant D′15/d′85. Test 6.0GB35-100(R) was conducted to examine the repeatability ofthe test results at the upper boundary of S f = 0.35. Tests on gradations 4.8GB35 and 6.5GB35were conducted to investigate the influence of D′15/d′85 at S f = 0.35.The 16 tests on BT mixtures were conducted to investigate the influence of particle shape.At the lower boundary of S f = 0.20 three tests were conducted on 5.7BT20, isotropically con-solidated at cell pressures of 50, 100 ad 150 kPa. Since the results indicated that the instabilityat the lower boundary was not governed by effective stress, gradations 5.1BT20 and 7.0BT20were only tested at cell pressures of 50 and 150 kPa. The test on 8.6BT20 was only conductedat 50 kPa. The tests 5.7BT35-100, 7.0BT35-50 and 8.6BT35-50 exhibited no internal insta-bility. Therefore, these gradations were not tested at different pressures. Tests on 10.4BT35did exhibit an internally unstable response and hence this gradation was tested at different cellpressures. Test 10.4BT35-50(R) was conducted to examine the repeatability of the test resultson BT gradations at S f = 0.35. Tests 10.4BT25-50 and 10.4BT30-50 were conducted to gainan appreciation of the response of soils at intermediate finer fraction contents.For all gradations tested in this study, the same bottom wire meshes were used: the openingsize of the primary bottom wire mesh (see Fig. 3.6) is equal to 0.033 mm, in order to preventparticle loss during specimen reconstitution; the secondary bottom wire mesh, with openingsize of 0.6 mm, serves to keep the fine wire mesh in place. The opening size of the primary topwire mesh is equal to D0 for the gap-graded specimens and equal to 0.033 mm for gradationGB-F. The opening size of the secondary top wire mesh and tertiary top wire mesh are 2.8 mmand 0.033 mm, respectively, for all gradations.3.5 SummaryA new flexible wall permeameter has been developed at the University of British Columbia toinvestigate seepage-induced internal instability in gap-graded materials. The device comprisesa double-walled triaxial cell; a seepage control system, through which unidirectional multi-stage seepage flow can be imposed; and instrumentation. A novel feature in seepage-inducedinternal instability testing is the measurement of total volume change. It is derived from mon-itoring the volume change of the cell fluid in the inner chamber of the double-walled triaxialcell, that, with correction for the intrusion of the loading ram, membrane penetration, and smallcalibrated volume changes of dissolving air and absorption of water, yields an accurate mea-surement of total volume change. In addition, the flow rate, differential pore water pressureacross the specimen, the pore water pressure and the axial deformation are measured. The main48limitations of the apparatus relate to the control and working range of the total dynamic headand the working range of the volume change measurement.Gap-graded mixtures of glass beads and soils are prepared by combining, in various ratios,the respective fine components with different coarse components. Assessment of trial speci-mens suggest that the modified slurry deposition technique is capable of producing saturatedspecimens of gap-graded glass bead mixtures with a small variation of S f = +/- 0.03 with:S f = 0.25 to 0.35 and D′15/d′85 ≤ 6.5; and S f = 0.20 and D′15/d′85 ≤ 4.8. The upper limitfor a reasonably homogeneous specimen (S f = +/- 0.08) of glass beads with S f = 0.20, wasfound to occur at D′15/d′85 = 6.0. The cylindrical specimens, with a diameter of 100 mm, arereconstituted to a height of approximately 100 mm, subsequently isotropically consolidated toa target cell pressure of 50, 100 or 150 kPa and then subject to multi-stage upward seepage flow.The testing program has been developed to: commission the apparatus; assess the repeata-bility of the test results; and generate data to investigate the influence on seepage-induced inter-nal instability in gap-graded materials of effective stress, particle size distribution and particleshape. The commissioning program comprised two tests conducted on glass beads mixtures.The main testing program comprises 23 tests on glass beads mixtures, of which two are re-peatability tests, and 16 test on soil mixtures, of which one is a repeatability test.49Table 3.1: Characteristics of test materials.Component1 d0 d15 d85 d100 Cu Cc AR50 Cx50 SQP50(mm) (mm) (mm) (mm) (-) (-) (-) (-) (-)GB-F 0.12 0.14 0.19 0.20 1.25 1.03 0.91 0.94 0.94GB-C12 0.60 0.63 0.81 0.85 - - - - -GB-C2 0.80 0.90 1.03 1.20 1.10 1.00 0.97 0.98 0.94GB-C3 1.00 1.13 1.29 1.40 1.11 1.00 0.97 0.98 0.94GB-C4 1.10 1.23 1.50 1.80 1.15 0.99 0.95 0.98 0.94BT-F 0.09 0.12 0.17 0.21 1.38 1.03 0.76 0.92 0.91BT-C1 0.70 0.89 1.19 1.5 1.27 1.01 0.73 0.96 0.88BT-C2 0.80 0.98 1.46 2.0 1.33 1.00 0.73 0.96 0.87BT-C3 0.90 1.21 1.76 2.2 1.34 1.00 0.73 0.97 0.88BT-C4 1.10 1.50 2.14 2.6 1.32 1.00 0.74 0.97 0.88BT-C5 1.40 1.79 2.49 3.0 1.30 1.01 0.74 0.97 0.88Notes:1 GB = Glass beads, BT = sub-angular soils, F = Fine component, C = Coarsecomponent2 Material GB-C1 was not tested in the QicPic deviceTable 3.2: Gradations of the test specimens.R1 Gradation Fine fraction Coarse fraction S f (-) D′15/d′85 (-)R 3.3GB20 GB-F GB-C1 0.20 3.3R 4.8GB20 GB-F GB-C2 0.20 4.8R 4.8GB35 GB-F GB-C2 0.35 4.8R 6.0GB20 GB-F GB-C3 0.20 6.0R 6.0GB25 GB-F GB-C3 0.25 6.0R 6.0GB30 GB-F GB-C3 0.30 6.0R 6.0GB35 GB-F GB-C3 0.35 6.0R 6.5GB25 GB-F GB-C4 0.35 6.5R 6.5GB35 GB-F GB-C4 0.35 6.5SA 5.1BT20 BT-F BT-C1 0.20 5.1SA 5.7BT20 BT-F BT-C2 0.20 5.7SA 5.7BT35 BT-F BT-C2 0.35 5.7SA 7.0BT20 BT-F BT-C3 0.20 7.0SA 7.0BT35 BT-F BT-C3 0.35 7.0SA 8.6BT20 BT-F BT-C4 0.20 8.6SA 8.6BT35 BT-F BT-C4 0.35 8.6SA 10.4BT25 BT-F BT-C5 0.25 10.4SA 10.4BT30 BT-F BT-C5 0.30 10.4SA 10.4BT35 BT-F BT-C5 0.35 10.4Note:1 R = Rounded; SA = Sub-angular.50ComputerDAQInstrumentationvpanelDouble-walledvtriaxialvcellWatervbathWatervsupplySandvfilterCarbonvfilterDe-airingvvtank StoragevreservoirPeristalticvpumpInflowvconstant-vheadvdevicevOutflowvconstand-headvdeviceTotalvdynamicvheadFigure 3.1: General configuration of flexible wall permeameter arrangement.LinearSvariableSdifferentialStransformerLoadingSramInnerSacryclicStubeInnerSchamberTopScapFlexibleStubesTieSrodsOuletSportMeasurementBaseSframeBaseSpedestalInletSportInnerSchamberSportMeasurementSportS,2OuterSchamberSportSpeci-menO-ringPerforatedSplate,topSwireSmeshesSandSgasketO-ringOuterSacrylicStubeOuterSchamberTopSplateVentingSportsportS,1U-cupPerforatedSplate,SbottomSwireSmeshesSandSgasketFigure 3.2: Double-walled triaxial cell.51Manualiregulatoriw1BackipressureManualiregulatoriw2Measurementiburetteiw2Measurementiburetteiw3Buretteiw4DPTiw2TPTiw1 PVolume change: seepage stageDPTiw1 TPTiw2I-CPMPiw1OutletiportFiltered,ide-airediwaterifromiI-CHDInletiportO-CPMPiw2VacuumiregulatorVacuumisupplyLegendDPTi=iDifferentialiPressureiTransducerTPTi=iTotaliPressureiTransducerI-CHDi=iInflowiConstantiHeadiDeviceO-CHDi=iOutflowiConstantiHeadiDeviceiMPi=iMeasurementiPortI-CPi=iInneriChamperiPortO-CPi=iOuteriChamperiPortSeepageiflowitoiO-CHDTieirodInneriacrylicitubeOuteriacrylicitubeBaseipedestalMeasure-mentiburetteiw1PVolume change: reconstitution stageFigure 3.3: Schematic plan view of double-walled triaxial cell, including volume changemeasurement systems.Cell pressure, σcAtmospheric pressureσcInner chamber Outer chamberFigure 3.4: Pressures in the double-walled triaxial cell.52Figure 3.5: Dummy specimen mounted between base pedestal and top cap in double-walled triaxial cell (without acrylic tubes).53PerforatedgplateTertiarygtopgwiregmeshPVCgGasketPerforatedgplateSecondarygtopgwiregmeshPrimarygtopgwiregmeshPVCgGasketPVCgGasketSecondarygbottomgwiregmeshPrimarygbottomgwiregmeshPerforatedgplateSpecimenTopgcapBasegpedestalInnergacryclicgtubeOutergacrylicgtubeTiegrodsO-ringsOutergchambergportInnergchambergportMeasurementgportg22IntletgportBasegframeOutletgportMeasurementgportg21 BoltsPipegfittingsPipegfittingsPortgingbasegpedestalFlexiblegtubesFigure 3.6: Exploded view of connections between base frame, base pedestal, specimen,and top cap.54Figure 3.7: Relation between rate of volume change of inner chamber and cell pressurefrom calibration testing (t from 16 to 22 h after application of cell pressure).Figure 3.8: Particle size distribution of the glass bead components using QicPic dFmin .55(a)(b)(c)Figure 3.9: Shape attributes of the glass bead particles.56Figure 3.10: Microscope image of component GB-C4 of glass beads.Figure 3.11: Microscope image of component GB-F of glass beads.57Figure 3.12: Particle size distribution of the 3.3GB gradation.Figure 3.13: Particle size distribution of the 4.8GB gradations.Figure 3.14: Particle size distribution of the 6.0GB gradations.58Figure 3.15: Particle size distribution of the 6.5GB gradations.Figure 3.16: Particle size distribution of the components of soil.59(a)(b)(c)Figure 3.17: Shape attributes of the soil particles.60Figure 3.18: Microscope image of the component BT-F of soil.Figure 3.19: Microscope image of the component BT-C3 of soil.61Figure 3.20: Particle size distribution of the 5.1BT gradation.Figure 3.21: Particle size distribution of the 5.7BT gradations.Figure 3.22: Particle size distribution of the 7.0BT gradations.62Figure 3.23: Particle size distribution of the 8.6BT gradations.Figure 3.24: Particle size distribution of the 10.4BT gradations.Figure 3.25: Variation of percentage finer fraction content per layer of a trial specimen ofgradation 4.8GB20.63Figure 3.26: Variation of percentage finer fraction content per layer of a trial specimen ofgradation 6.5GB25.Figure 3.27: Variation of percentage finer fraction content per layer of a trial specimen ofgradation 6.5GB35.Figure 3.28: Variation of percentage finer fraction content per layer of a trial specimen ofgradation 6.0GB20.64Figure 3.29: Variation of percentage finer fraction content per layer of a second trial spec-imen of gradation 6.0GB20.65Figure 3.30: Test program.66Chapter 4ResultsLaboratory experiments were performed to commission the flexible wall permeameter, assessthe repeatability of the test results (research objective No. 3) and generate experimental data.The results of these experiments are presented in this Chapter: the results and findings of thecommissioning tests are described in Section 4.1, followed by descriptions of the results oftests on glass beads gradations in Section 4.2 and descriptions of the results of tests on soilgradations in Section 4.3. Upward multi-stage seepage flow, with head control, was imposedin all tests. The results are reported in terms of the response variables ∆u 1, v, εa and εv (seeFig. 4.1); contractive strains are positive. In addition, visual observations and post-test particlesize analyses are reported for select tests in Appendix D. The repeatability of the test resultsis examined in Section 4.4. A synthesis of the findings of the experiments, including a shortdescription of typical responses, is provided in Section 4.5.4.1 Commissioning testsTwo commissioning tests were performed on two glass beads gradations: one test was con-ducted on the uniform gradation GB-F and the other on the gap-graded gradation 6.5GB25.4.1.1 Test GB-F-100Gradation GB-F was reconstituted by water pluviation and isotropically consolidated to p′c =102 kPa, l = 97 mm and ec = 0.65 (see Table 4.1). Twenty-one stages of seepage flow wereapplied; the test commenced with TDH = 2 cm for ∆t = 40 min, yielding ∆u = 0.2 kPa and v= 0.002 cm/s; it was terminated at TDHmax after t = 280 min, yielding ∆u = 8.1 kPa, and v =0.117 cm/s (see Table 4.2).1The results are reported in terms of ∆u, instead of hydraulic gradient, as the former is more suitable to quantifythe seepage-induced change of effective stress (see Section 2.3). Considering the limited extend of axial defor-mation, typically εa < 2%, and typical length of the specimen l = 100mm, the value of hydraulic gradient isapproximately equal to the value of ∆u.67The specific discharge, v, was directly proportional to the differential pore water pressureacross the specimen, ∆u, throughout the test (see Fig. 4.2a). A very small axial strain of εa= 0.04 % (see Fig. 4.2b), and a small volumetric strain of εv = 0.18 % (see Fig. 4.2c), wererecorded at the end of the test.4.1.2 Test 6.5GB25-100Gradation 6.5GB25 was reconstituted by slurry deposition and isotropically consolidated to p′c= 102 kPa, l = 96 mm and ec = 0.36 (see Table 4.1). Twelve stages of seepage flow wereapplied; the test commenced with TDH = 1 cm for ∆t = 20 min, yielding ∆u = 0.1 kPa and v= 0.004 cm/s; it was terminated at TDH = 35 cm after t = 270 min, when the volume changemeasurement device approached the limit of its working range, yielding ∆u = 1.6 kPa and v =0.077 cm/s (see Table 4.2).The variation of v:∆u exhibited an overall curvilinear response in three sequences (see Fig.4.3a): (i) it commenced with an approximately directly proportional relation to v = 0.039 cm/sand ∆u = 1.1 kPa; (ii) it was followed by a marked increase to v = 0.048 cm/s at ∆u = 1.1 kPa;(iii) it was then followed by an approximately proportional relation for the remainder of thetest. The variation of εa:∆u exhibited an overall curvilinear response in two sequences (see Fig.4.3b): (i) there was no axial strain to ∆u = 1.1 kPa; (ii) the axial strain subsequently increased toa very small value εa = 0.08 % at the end of the test. The variation of εv:∆u exhibited an overallcurvilinear response in two sequences (see Fig. 4.3c): (i) the relation was directly proportionalto a very small value of εv = 0.06 % at ∆u = 1.1 kPa; (ii) it was followed by a series of discreteincrements of increasing volumetric strain to εv = 1.65 % at the end of the test.4.1.3 Findings of the commissioning testsTwo commissioning tests were performed with the primary objective of assessing the measure-ment techniques of the apparatus. In test GB-F-100, the small end-of-test values of axial andvolumetric strains εa = 0.04 % and εv = 0.18 %, respectively, and the directly proportionalresponse of v:∆u, indicate that no substantial change of the specimen fabric occurred. Accord-ingly, the variations of v:∆u, εa:∆u and εv:∆u in test GB-F-100 are consistent, and indicate thatno change transpired in the specimen during the test.The second commissioning test, 6.5GB25-100, was conducted to assess the measurementtechniques of the apparatus on a specimen where substantial change was anticipated. Thenature of the v:∆u relation, and the absence of significant axial and volumetric deformations,indicate no change occurred to ∆u = 1.1 kPa. The marked increase to v = 0.048 cm/s at ∆u = 1.1kPa indicates a change occurred in the fabric of the specimen, which is also apparent from themarked development of volumetric strain at this stage. Accordingly, the variations of v:∆u and68εv:∆u in test 6.5GB25-100 appear to be consistent and indicate that substantial change occurredduring the test. The variation of εa:∆u does not indicate substantial change, which is attributedto the localised nature of the volume change. On the matter of accuracy, the end of the test εv= 1.65 % is more than one order of magnitude greater than the accuracy of the total volumechange measurement of εv = +/- 0.07%. On the matter of resolution, the first increment ofvolumetric strain at the onset of deformation ∆εv = 0.5%, is more than one order of magnitudegreater than the resolution of the volume change measurement of +/- 0.01% volumetric strain.The resolution and accuracy of the volume change measurement technique are thus believedsufficient to quantify the onset and progression of deformation in specimens with volume ofapproximately 780 cm3.4.2 Tests on glass beadsThe initial and the end-of-test conditions of each test on glass beads are tabulated in Tables 4.1and 4.2, respectively. The response to seepage flow measured in each test is described herein.A summary of typical responses in tests on glass beads gradations is provided in Section 4.5.4.2.1 Gradation 3.3GB20One test was conducted on gradation 3.3GB20, which has the smallest gap-ratio of all grada-tions tested. The specimen was reconstituted by slurry deposition and isotropically consolidatedto a target cell pressure of 50 kPa. The tests code is 3.3GB20-50.4.2.1.1 Test 3.3GB20-50The specimen was isotropically consolidated to p′c = 54 kPa, l = 102 mm and ec = 0.53 (seeTable 4.1). It was subject to 26 stages of seepage flow: the test commenced at TDH = 1 cm for∆t = 30 min, yielding ∆u = 0.1 kPa and v = 0.003 cm/s; it was terminated at TDHmax after t =280 min, yielding ∆u = 3.6 kPa and v = 0.127 cm/s (see Table 4.2).The variation of v:∆u exhibited a curvilinear response in two sequences (see Fig. 4.4a):(i) the relation was proportional during the first eleven stages of the test to ∆u = 1.0 kPa andv = 0.043 cm/s, with an intercept of v = 0.005 cm/s at ∆u = 0 kPa; (ii) it was followed by anapproximately proportional response, at a slightly smaller rate, during the remaining stages.There was no axial deformation during the test (see Fig. 4.4b). A very small end-of-test valueεv = 0.05 % was measured (see Fig. 4.4c). A post-test particle size analysis of the specimenshowed that the finer fraction content was equal to S f = 0.20 in all five layers (see Fig. D.23).694.2.2 Gradation 4.8GB20Four tests were conducted on gradation 4.8GB20. The specimens were reconstituted by slurrydeposition and isotropically consolidated to target cell pressures of 50 kPa (in two sequen-tial tests), 100 kPa and 150 kPa, respectively. The corresponding test codes are 4.8GB20-50,4.8GB20-50(R), 4.8GB20-100, and 4.8GB20-150.4.2.2.1 Test 4.8GB20-50The specimen was isotropically consolidated to p′c = 54 kPa, l = 103 mm and ec = 0.48 (seeTable 4.1). It was subject to 20 stages of seepage flow: the test commenced at TDH = 1 cm for∆t = 40 min, yielding ∆u = 0.1 kPa and v = 0.002 cm/s; it was terminated at TDHmax after t =310 min, yielding ∆u = 5.3 kPa and v = 0.148 cm/s (see Table 4.2).The variation of v:∆u exhibited a curvilinear response in three sequences (see Fig. 4.5a): (i)the relation was directly proportional during the first two stages of the test to ∆u = 0.2 kPa andv = 0.004 cm/s; (ii) it was followed by an increasing specific discharge at an initially increasingrate and subsequently decreasing rate, to ∆u = 2.9 kPa and v = 0.090 cm/s; (iii) it was thenfollowed by an approximately proportional relation during the remaining stages. There was noaxial deformation (see Fig. 4.5b). A small end-of-test value εv = 0.12 % (see Fig. 4.5c) wasrecorded.4.2.2.2 Test 4.8GB20-50(R)The specimen was isotropically consolidated to p′c = 53 kPa, l = 102 mm and ec = 0.45 (seeTable 4.1). It was subject to 23 stages of seepage flow: the test commenced at TDH = 1 cm for∆t = 40 min, yielding ∆u = 0.1 kPa and v = 0.002 cm/s; it was terminated at TDHmax after t =370 min, yielding ∆u = 6.7 kPa and v = 0.138 cm/s (see Table 4.2).The variation of v:∆u also exhibited a curvilinear response comprising three sequences (seeFig. 4.5a): (i) the relation was directly proportional during the first two stages of the test to∆u = 0.2 kPa and v = 0.005 cm/s; (ii) it was followed by an increasing specific discharge at aninitially increasing rate and subsequently decreasing rate to ∆u = 2.3 kPa and v = 0.055 cm/s;(iii) it was then followed by an approximately proportional relation during the remaining stages.The negligible end-of-test values εa = - 0.01 % and εv = 0.03 % indicate that there was no axialand volumetric deformation (see Figs. 4.5b and 4.5c).4.2.2.3 Test 4.8GB20-100The specimen was isotropically consolidated to p′c = 104 kPa, l = 102 mm and ec = 0.47 (seeTable 4.1). It was subject to 21 stages of seepage flow: the test commenced at TDH = 1 cm for70∆t = 70 min, yielding ∆u = 0.1 kPa and v = 0.001 cm/s; it was terminated at TDHmax after t =310 min, yielding ∆u = 6.0 kPa and v = 0.142 cm/s (see Table 4.2).The variation of v:∆u exhibited a curvilinear response comprising three sequences (see Fig.4.5a): (i) the relation was directly proportional during the first two stages of the test to ∆u = 0.2kPa and v = 0.004 cm/s; (ii) it was followed by an increasing specific discharge, at an initiallyslightly increasing rate and subsequently slightly decreasing rate, to ∆u = 2.1 kPa and v = 0.050cm/s in stage 11; (iii) it was then followed by an approximately proportional relation duringthe remaining stages. A very small end-of-test value εa = 0.02 % (see Fig. 4.5b) was recorded.The variation of εv:∆u exhibited a curvilinear response in three sequences (see Fig. 4.5c): (i)it commenced with an increasing volumetric strain to a very small value εv = 0.05 % duringstages one to nine; (ii) it was followed by an isolated increase to εv = 0.50 % at ∆u = 1.8 kPain stage ten; (iii) it was then followed by a sequence of constant volumetric strain through theremainder of the test.4.2.2.4 Test 4.8GB20-150The specimen was isotropically consolidated to p′c = 153 kPa, l = 101 mm and ec = 0.44 (seeTable 4.1). It was subject to 21 stages of seepage flow: the test commenced at TDH = 1 cm for∆t = 70 min, yielding ∆u = 0.1 kPa and v = 0.038 cm/s; it was terminated at TDHmax after T =410 min, yielding ∆u = 6.0 kPa and v = 0.142 cm/s (see Table 4.2).The variation of v:∆u exhibited a curvilinear response comprising three sequences (see Fig.4.5a): (i) it commenced with a directly proportional relation during the first three stages of thetest to ∆u = 0.2 kPa and v = 0.003 cm/s; (ii) it was followed by an increasing v at an increas-ing rate to ∆u = 1.5 kPa and v = 0.021 cm/s; (iii) it was then followed by an approximatelyproportional relation to the end of the test. A very small end-of-test value εa = 0.04 % (seeFig. 4.5b) indicates that there was no substantial axial deformation. The variation of εv:∆uexhibited a curvilinear response in three sequences (see Fig. 4.5c): (i) it commenced with anincreasing volumetric strain to a very small value εv = 0.05 % during stages one and two; (ii) itwas followed by an isolated increase to εv = 0.44 % at ∆u = 0.25 kPa in stage three; (iii) it wasthen followed by a very small increase to εv = 0.48 % at the end of the test. A post-test particlesize analysis of the specimen showed that the finer fraction content of S f = 0.15 in the top layerwas markedly lower than the finer fraction content of S f = 0.24 to 0.28 in the other four layers(see Fig. D.24).4.2.3 Gradation 4.8GB35Three tests were conducted on gradation 4.8GB35. The specimens were reconstituted by slurrydeposition and isotropically consolidated to target cell pressures of 50 kPa, 100 kPa and 15071kPa, respectively. The corresponding test codes are 4.8GB35-50, 4.8GB35-100, and 4.8GB35-150.4.2.3.1 Test 4.8GB35-50The specimen was isotropically consolidated to p′c = 57 kPa, l = 103 mm and ec = 0.36 (seeTable 4.1). It was subject to 28 stages of seepage flow: the test commenced at TDH = 1 cm for∆t = 40 min, yielding ∆u = 0.1 kPa and v = 0.001 cm/s; it was terminated at TDHmax after t =360 min, yielding ∆u = 8.7 kPa and v = 0.114 cm/s (see Table 4.2).The variation of v:∆u exhibited a curvilinear response in two sequences (see Fig. 4.6a): (i) itcommenced with a directly proportional relation to ∆u = 2.1 kPa and v = 0.003 cm/s; (ii) it wasfollowed by an increasing specific discharge, at initially slightly decreasing and subsequentlyslightly increasing rate, to the end of the test. The variation of εa:∆u exhibited a curvilinearresponse in two sequences (see Fig. 4.6b): (i) it commenced with an increasing axial strainto a very small value εa = 0.04 % at ∆u = 2.7 kPa; (ii) it was followed by a series of discreteincrements of increasing axial strain to εa = 1.15 % at the end of the test. The variation of εv:∆ualso exhibited a curvilinear response in two sequences (see Fig. 4.6b): (i) it commenced withan increasing volumetric strain to a very small value εv = 0.07 % at ∆u = 1.5 kPa; (ii) it wasfollowed by a series of discrete increments of increasing volumetric strain to εv = 1.60 % at theend of the test. Visual observations (see Fig. D.1) suggest that the volumetric deformation wasrestricted to the top half of the specimen.4.2.3.2 Test 4.8GB35-100The specimen was isotropically consolidated to p′c = 103 kPa, l = 102 mm and ec = 0.37 (seeTable 4.1). It was subject to 29 stages of seepage flow: the test commenced at TDH = 1 cm for∆t = 30 min, yielding ∆u = 0.1 kPa and v = 0.001 cm/s; it was terminated at TDHmax after t =360 min, yielding ∆u = 7.3 kPa and v = 0.128 cm/s (see Table 4.2).The specific discharge v was approximately directly proportional to ∆u throughout the test(see Fig. 4.6a). The variation of εa:∆u exhibited a curvilinear response in three sequences (seeFig. 4.6b): (i) it commenced with a very small value εa = 0.02 % to ∆u = 1.6 kPa; (ii) it wasfollowed by a series of discrete increments of increasing axial strain to εa = 0.79% at ∆u = 4.3kPa; (iii) it was then followed by a constant axial strain to the end of the test. The variationof εv:∆u exhibited a curvilinear response in three sequences (see Fig. 4.6c): (i) it commencedwith an increasing volumetric strain to a very small value εv = 0.07 % at ∆u = 1.6 kPa; (ii) itwas followed by a series of discrete increments of increasing volumetric strain to εv = 1.13% at∆u = 4.3 kPa; (iii) it was then followed by a slightly increasing volumetric strain to εv = 1.20% at the end of the test. A post-test particle size analysis of the specimen showed that the finer72fraction content S f = 0.37 was identical in all five layers (see Fig. D.25).4.2.3.3 Test 4.8GB35-150The specimen was isotropically consolidated to p′c = 152 kPa, l = 103 mm and ec = 0.37 (seeTable 4.1). It was subject to 29 stages of seepage flow: the test commenced at TDH = 1 cm for∆t = 30 min, yielding ∆u = 0.1 kPa and v = 0.002 cm/s; it was terminated at TDHmax after t =390 min, yielding ∆u = 8.6 kPa and v = 0.114 cm/s (see Table 4.2).The variation of v:∆u exhibited a curvilinear response in two sequences (see Fig. 4.6a):(i) it commenced with a directly proportional increase during the first twelve stages to ∆u =2.8 kPa and v = 0.045 cm/s; (ii) it was followed by an increasing v at initially decreasing andsubsequently slightly varying rate to the end of the test. A very small end-of-test value εa =0.05 % (see Fig. 4.6b) was recorded. In contrast, the variation of εv:∆u exhibited a curvilinearresponse in two sequences (see Fig. 4.6b): (i) it commenced with an increasing volumetricstrain to a very small value εv = 0.08 kPa at ∆u = 2.8 kPa; (ii) it was followed by a series ofdiscrete increments of increasing volumetric strain to εv = 0.78 % at the end of the test. A post-test particle size analysis of the specimen showed that the finer fraction content varied from S f= 0.36 to 0.39 throughout the specimen (see Fig. D.26).4.2.4 Gradation 6.0GB20Three tests were conducted on gradation 6.0GB20, which proved challenging to reconstitute ina homogeneous manner (see Figs. 3.28 to 3.29), owing to the combination of the spherical par-ticles, the relatively low fines content and the relatively large value of D′15/d′85. The specimenswere reconstituted by slurry deposition and isotropically consolidated to target cell pressuresof 50 kPa, 100 kPa and 150 kPa, respectively. The corresponding tests codes are 6.0GB20-50,6.0GB20-100 and 6.0GB20-150.4.2.4.1 Test 6.0GB20-50The specimen was isotropically consolidated to p′c = 54 kPa, l = 103 mm and ec = 0.45 (seeTable 4.1). It was subject to 14 stages of seepage flow: the test commenced at TDH = 1 cm for∆t = 50 min, yielding ∆u = 0.1 kPa and v = 0.004 cm/s; it was terminated at TDH = 70 cm aftert = 260 min, yielding ∆u = 3.9 kPa and v = 0.098 cm/s (see Table 4.2).The variation of v:∆u exhibited a curvilinear response in three sequences (see Fig. 4.7a): (i)it commenced with a constant specific discharge during the first stage; (ii) it was followed by agreater, but subsequently decreasing specific discharge during the second stage from v = 0.011cm/s to v = 0.010 cm/s at ∆u = 0.2 kPa; (iii) it was then followed by a sequence of increasing73specific discharge at initially increasing and subsequently decreasing rate to the end of the test.There was no axial deformation and the volumetric strain increased to a very small end-of-testvalue εv = 0.09 % (see Fig. 4.7c). A post-test particle size analysis of the specimen showedthat the finer fraction content of S f = 0.14 in the top layer was markedly lower than the finerfraction content of S f = 0.21 to 0.28 in the other four layers (see Fig. D.27).4.2.4.2 Test 6.0GB20-100The specimen was isotropically consolidated to p′c = 102 kPa, l = 102 mm and ec = 0.41 (seeTable 4.1). It was subject to 27 stages of seepage flow: the test commenced at TDH = 1 cm for∆t = 30 min, yielding ∆u = 0.1 kPa and v = 0.004 cm/s; it was terminated at TDHmax after t =370 min, yielding ∆u = 6.9 kPa and v = 0.133 cm/s (see Table 4.2).The variation of v:∆u exhibited a curvilinear response comprising three sequences (see Fig.4.7a): (i) the relation was directly proportional to ∆u = 0.3 kPa and v = 0.011 cm/s during thefirst four stages; (ii) it was followed by an increase of v at an initially smaller rate (during thefifth stage); (iii) it was then followed by an increase of v, at an initially constant and subse-quently decreasing rate, to the end of the test. The variation of εa:∆u exhibited a curvilinearresponse comprising three sequences (see Fig. 4.7b): (i) there was no axial deformation duringthe first stage; (ii) it was followed by an increase of axial strain, which initiated at ∆u = 0.2kPa to εa = 0.28 % at ∆u = 0.4 kPa; (iii) it was then followed by a very small increase to εa =0.30 % at the end of the test. The variation of εv:∆u exhibited a curvilinear response in threesequences (see Fig. 4.7c): (i) there was no volumetric deformation during the first stage; (ii)it was followed by an increasing volumetric strain in two increments to εv = 0.76 %, whichinitiated at ∆u = 0.2 kPa; (iii) it was then followed by a very small increase of volumetric strainto εv = 0.82 % at the end of the test. A post-test particle size analysis of the specimen showedthat the finer fraction content of S f = 0.13 in the top layer was markedly lower than the finerfraction content of S f = 0.23 to 0.27 in the other four layers (see Fig. D.28).4.2.4.3 Test 6.0GB20-150The specimen was isotropically consolidated to p′c = 151 kPa, l = 101 mm and ec = 0.40 (seeTable 4.1). It was subject to twelve stages of seepage flow;: the test commenced at TDH = 5cm for ∆t = 60 min, yielding ∆u = 0.4 kPa and v = 0.004 cm/s; it was terminated at TDH = 70cm after t = 240 min, yielding ∆u = 3.9 kPa and v = 0.067 cm/s (see Table 4.2).The variation of v:∆u exhibited a curvilinear response in three sequences (see Fig. 4.7a):(i) it commenced with a decreasing specific discharge, from v = 0.0042 cm/s to v = 0.0038 cm/sduring the first stage; (i) it was followed by a greater, but decreasing specific discharge, from v= 0.007 cm/s to v = 0.006 cm/s, at ∆u = 0.7 kPa in the second stage; (iii) it was then followed74by an increasing specific discharge, at an initially increasing and subsequently decreasing rate,to the end of the test. An insignificant axial strain εa = 0.01 % and a very small volumetricstrain εv = 0.09 % were measured at the end of the test (see Figs. 4.7b and 4.7c, respectively).A post-test particle size analysis of the specimen showed that the finer fraction content of S f =0.05 in the top layer was markedly lower than the finer fraction content of S f = 0.19 to 0.29 inthe other four layers (see Fig. D.29).4.2.5 Gradation 6.0GB25Three tests were conducted on gradation 6.0GB25, which experience showed could be reconsti-tuted with a much smaller spatial variation of the finer fraction content than gradation 6.0GB20(see Section 3.3.1). The specimens were reconstituted by slurry deposition and were isotrop-ically consolidated to target cell pressures of 50 kPa, 100 kPa and 150 kPa, respectively. Thetest codes are 6.0GB25-50, 6.0GB25-100 and 6.0GB25-150.4.2.5.1 Test 6.0GB25-50The specimen was isotropically consolidated to p′c = 53 kPa, l = 100 mm and ec = 0.38 (seeTable 4.1). It was subject to 21 stages of seepage flow: the test commenced at TDH = 1 cm for∆t = 60 min, yielding ∆u = 0.1 kPa and v = 0.003 cm/s; it was terminated at TDHmax after t =380 min, yielding ∆u = 6.2 kPa and v = 0.138 cm/s (see Table 4.2).The variation of v:∆u exhibited a curvilinear response in two sequences (see Fig. 4.8a): (i)the relation was directly proportional to ∆u = 0.7 kPa and v = 0.016 cm/s; (ii) it was followedby a sequence of increasing specific discharge at varying rates. The variation of εa:∆u exhibiteda curvilinear response in three sequences (see Fig. 4.8b): (i) there was no axial strain to ∆u =0.7 kPa; (ii) it was followed by a small increase to εa = 0.12 % at ∆u = 2.1 kPa; (iii) it was thenfollowed by a series of discrete increments of increasing axial strain to an end-of-test value εa= 0.68 %. The variation of εv:∆u exhibited a curvilinear response in two sequences (see Fig.4.8c): (i) the relation was directly proportional to a very small value εv = 0.05 % at ∆u = 0.7kPa; (ii) it was followed by a series of discrete increments of increasing volumetric strain to anend-of-test value εv = 1.61 %. Visual observations (see Fig. D.2) established that the volumetricdeformation did not develop uniformly across the specimen, but was rather restricted to one sideof the specimen.4.2.5.2 Test 6.0GB25-100The specimen was isotropically consolidated to p′c = 103 kPa, l = 100 mm and ec = 0.37 (seeTable 4.1). It was subject to eleven stages of seepage flow: the test commenced at TDH = 1 cmfor ∆t =20 min, yielding ∆u = 0.1 kPa and v = 0.004 cm/s; it was terminated at TDH = 60 cm75after t = 140 min, yielding ∆u = 2.5 kPa and v = 0.057 cm/s (see Table 4.2).The variation of v:∆u exhibited a curvilinear response in two sequences (see Fig. 4.8b): (i)it commenced with an increasing specific discharge at slightly decreasing rate to v = 0.016 cm/sat ∆u = 0.7 kPa; (ii) it was followed by a proportional relation to the end of the test. There wasno significant axial strain or volumetric strain during the test (see Figs. 4.8b and 4.8c).4.2.5.3 Test 6.0GB25-150The specimen was isotropically consolidated to p′c = 151 kPa, l = 101 mm and ec = 0.36 (seeTable 4.1). It was subject to 23 stages of seepage flow: the test commenced at TDH = 4 cm for∆t =50 min, yielding ∆u = 0.3 kPa and v = 0.008 cm/s; it was terminated at TDHmax after t =320 min, yielding ∆u = 8.0 kPa and v = 0.120 cm/s (see Table 4.2).The variation of v:∆u exhibited a curvilinear response in three sequences (see Fig. 4.8a): (i)the relation was directly proportional to ∆u = 1.4 kPa and v = 0.034 cm/s; (ii) it was followed bya constant specific discharge to ∆u = 1.9 kPa and v = 0.034 cm/s, and subsequently proportionalincrease to ∆u = 4.8 kPa and v = 0.082 cm/s; (iii) it was then followed by a initially constantspecific discharge to ∆u = 5.9 kPa and v = 0.082 cm/s, and a subsequently proportional increaseto the end of the test. A very small axial strain εa = 0.05 % (see Fig. 4.8b) was measured atthe end of the test.. The variation of εv:∆u exhibited a curvilinear response in two sequences(see Fig. 4.8c): (i) it commenced with an increasing volumetric strain to a very small value εv= 0.05 % at ∆u = 1.4 kPa; (ii) it was followed by a series of discrete increments of increasingvolumetric strain to an end-of-test value εv = 1.58 %. A post-test particle size analysis of thespecimen showed that the finer fraction content of S f = 0.18 in the top layer was substantiallylower than the finer fraction content of S f = 0.25 to 0.30 in the other four layers (see Fig. D.30).4.2.6 Gradation 6.0GB30Three tests were conducted on gradation 6.0GB30. The specimens were reconstituted by slurrydeposition and isotropically consolidated to target cell pressures of 50 kPa, 100 kPa and 150kPa, respectively. The corresponding test codes are 6.0GB25-50, 6.0GB25-100 and 6.0GB25-150.4.2.6.1 Test 6.0GB30-50The specimen was isotropically consolidated to p′c = 53 kPa, l = 100 mm and ec = 0.36 (seeTable 4.1). It was subject to 21 stages of seepage flow: the test commenced at TDH = 2 cm for∆t = 70 min, yielding ∆u = 0.2 kPa and v = 0.001 cm/s; it was terminated at TDH = 120 cm,when the volume change measurement device approached the limit of its working range, after76t = 360 min, yielding ∆u = 5.9 kPa and v = 0.073 cm/s (see Table 4.2).The variation of v:∆u exhibited a curvilinear response in three sequences (see Fig. 4.9a): (i)the relation was directly proportional to v = 0.005 cm/s at ∆u = 0.6 kPa; (ii) it was followed byan increasing specific discharge, at an initially increasing rate and subsequently constant rate,to v = 0.047 cm/s at ∆u = 3.7 kPa; (iii) it was then followed by a decreasing specific discharge tov = 0.043 cm/s at ∆u = 3.9 kPa and a subsequently increasing specific discharge at varying ratesto the end of the test. The variation of εa:∆u exhibited a curvilinear response in two sequences(see Fig. 4.9b): (i) there was no axial deformation to ∆u = 3.4 kPa; (ii) it was followed by aseries of discrete increments of increasing axial strain to an end-of-test value εa = 0.44 %. Thevariation of εv:∆u also exhibited a curvilinear response in two sequences (see Fig. 4.9c): (i)it commenced with an increasing volumetric strain to a small value εv = 0.13 % at ∆u = 3.4kPa; (ii) it was followed by a series of discrete increments of increasing volumetric strain to anen-of-test value of εv = 1.90 %. Visual observations showed signs of local distress, which wererestricted to one side of the specimen and more pronounced near the top than near the bottom(see Fig. D.3).4.2.6.2 Test 6.0GB30-100The specimen was isotropically consolidated to p′c = 102 kPa, l = 98 mm and ec = 0.34 (seeTable 4.1). It was subject to 22 stages of seepage flow: the test commenced at TDH = 1 cm for∆t =50 min, yielding ∆u = 0.1 kPa and v = 0.003 cm/s; it was terminated at TDH = 140 cm,when the volume change measurement device approached the limit of its working range, aftert = 350 min, yielding ∆u = 7.5 kPa and v = 0.065 cm/s (see Table 4.2).The variation of v:∆u exhibited a curvilinear response in three sequences (see Fig. 4.9a):(i) it commenced with an increasing specific discharge at diminishing rate to v = 0.046 cm/s at∆u = 3.7 kPa; (ii) it was followed by a decreasing specific discharge at increasing rate to a localminimum of v = 0.034 cm/s at ∆u = 4.9 kPa; (iii) it was then followed by a typically increasingspecific discharge, at varying rates, to the end of the test (noteworthy during the sequence isa small decrease from v = 0.044 cm/s at ∆u = 5.4 kPa, to v = 0.042 cm/s at ∆u = 5.9 kPa).The variation of εa:∆u exhibited a curvilinear response in two sequences (see Fig. 4.9b): (i)there was no axial deformation to to ∆u = 3.7 kPa; (ii) it was followed by a series of discreteincrements of increasing axial strain to a small end-of-test value of εa = 0.21 %. The variationof εv:∆u exhibited a curvilinear response in two sequences (see Fig. 4.9c): (i) it commencedwith an increasing volumetric strain to a small value εv = 0.11 % at ∆u = 3.7 kPa; (ii) it wasfollowed by a series of discrete increments of increasing volumetric strain to an end-of-testvalue of εv = 1.78 %.774.2.6.3 Test 6.0GB30-150The specimen was isotropically consolidated to p′c = 151 kPa, l = 102 mm and ec = 0.34 (seeTable 4.1). It was subject to 18 stages of seepage flow: the test commenced at TDH = 5 cm for∆t =50 min, yielding ∆u = 0.4 kPa and v = 0.004 cm/s; it was terminated at TDH = 140 cm,when the volume change measurement device approached the limit of its working range, aftert = 350 min, yielding ∆u = 7.4 kPa and v = 0.053 cm/s (see Table 4.2).The variation of v:∆u exhibited a curvilinear response in three sequences (see Fig. 4.9a): (i)it commenced with an increasing specific discharge at slightly decreasing rate to v = 0.034 cm/sat ∆u = 4.5 kPa; (ii) it was followed by a diminishing specific discharge to a local minimum v= 0.026 cm/s at ∆u = 5.3 kPa; (iii) it was then followed by an increasing specific discharge, atvarying rates, to the end of the test. A very small axial strain εa = 0.04 % (see Fig. 4.9b) wasrecorded at the end of the test. The variation of εv:∆u exhibited a curvilinear response in twosequences (see Fig. 4.9c): (i) it commenced with an increasing volumetric strain to a very smallvalue εv = 0.07 % at ∆u = 4.5 kPa; (ii) it was followed by an an approximately proportionalincrease to an end-of-test value εv = 1.87 %. Visual observations indicated that the volumetricdeformation occurred predominantly across the top half of the specimen (see Fig. D.4) . Anabundance of fine particles was observed at the top of the specimen, after dis-assembly of thecell (see Fig. D.5). The post-test particle size analysis yields a relatively high finer fractioncontent of S f = 0.44 compared to S f = 0.28 to 0.34 in the centre and bottom of the specimen(see Fig. D.31).4.2.7 Gradation 6.0GB35Four tests were conducted on gradation 6.0GB35. The specimens were reconstituted by slurrydeposition and isotropically consolidated to target cell pressures of 50 kPa, 100 kPa (in twosequential tests) and 150 kPa, respectively. The corresponding test codes are 6.0GB35-50,6.0GB35-100, 6.0GB35-100(R), and 6.0GB35-150.4.2.7.1 Test 6.0GB35-50The specimen was isotropically consolidated to p′c = 54 kPa, l = 98 mm and ec = 0.33 (see Table4.1). It was subject to 17 stages of seepage flow: the test commenced at TDH = 2 cm for ∆t =60 min, yielding ∆u = 0.2 kPa and v = 0.002 cm/s; it was terminated at TDH = 105 cm aftert = 370 min, when the range of the volume change measurement device approached its limit,yielding ∆u = 6.6 kPa and v = 0.039 cm/s (see Table 4.2).The variation of v:∆u exhibited a curvilinear response in two sequences (see Fig. 4.10a): (i)the relation was directly proportional to ∆u = 2.1 kPa and v = 0.018 cm/s; (ii) it was followed78by a decreasing specific discharge to a local minimum of v = 0.016 cm/s at ∆u = 2.7 kPa; (iii) itwas then followed by an increasing specific discharge at varying rates to the end of the test. Thevariation of εa:∆u exhibited a curvilinear response in two sequences (see Fig. 4.10b): (i) therewas no axial deformation to ∆u = 2.1 kPa; (ii) it was followed by an increasing axial strain to asmall end-of-test value εa = 0.21 %. The variation of εv:∆u exhibited a curvilinear response intwo sequences (see Fig. 4.10c): (i) it commenced with an increasing volumetric strain to a verysmall value of εv = 0.07 % at ∆u = 1.7 kpa; (ii) it was followed by an increasing volumetricstrain, at an approximately constant rate, to an end-of-test value of εv = 2.19 %. Visual obser-vations (see Fig. D.6) established that the volumetric deformation developed predominantly inthe top half of the specimen. An abundance of fine particles at the top of the specimen wasobserved after dis-assembly of the cell (see Fig. D.7).4.2.7.2 Test 6.0GB35-100The specimen was isotropically consolidated to p′c = 102 kPa, l = 100 mm and ec = 0.37 (seeTable 4.1). It was subject to twelve stages of seepage flow: the test commenced at TDH = 1cm for ∆t = 20 min, yielding ∆u = 0.1 kPa and v = 0.003 cm/s; it was terminated at TDH = 60cm after t = 150 min, when the volume change measurement device approached the limit of itsworking range, yielding ∆u = 4.7 kPa and v = 0.028 cm/s (see Table 4.2).The variation of v:∆u exhibited a curvilinear response in two sequences (see Fig. 4.10a): (i)the relation was directly proportional to ∆u = 1.2 kPa and v = 0.016 cm/s; (ii) it was followed byan increasing specific discharge, at a varying rate to the end of the test. The variation of εa:∆uexhibited a curvilinear response in two sequences (see Fig. 4.10b): (i) the relation was directlyproportional response to a very small value of εa = 0.02 % at ∆u = 1.2 kPa; (ii) it was followedby a series of discrete increments of increasing axial strain to an end-of-test value of εa = 0.62%. The variation of εv:∆u also exhibited a curvilinear response in two sequences (see Fig.4.10c): (i) it was negligible to ∆u = 1.2 kPa; (ii) it was followed by an increasing volumetricstrain to an end-of-test value of εv = 1.54 %. Visual observations (see Fig. D.8) establishedthat the volumetric deformation occurred mainly at one side of the specimen, at the top half. Alarge amount of fine particles was detected at the top of the specimen, after dis-assembly of thecell (see Fig. D.9).4.2.7.3 Test 6.0GB35-100(R)The specimen was isotropically consolidated to p′c = 102 kPa, l = 102 mm and ec = 0.37 (seeTable 4.1). It was subject to 23 stages of seepage flow: the test commenced with TDH = 1 cmfor ∆t = 30 min, yielding ∆u = 0.1 kPa and v = 0.001 cm/s; it was terminated at TDH = 80cm after t = 270 min, when the volume change measurement device approached the limit of itsworking range, yielding ∆u = 5.9 kPa and v = 0.035 cm/s (see Table 4.2).79The variation of v:∆u exhibited a curvilinear response comprising two sequences (see Fig.4.10a): (i) the relation was directly proportional to ∆u = 1.8 kPa and v = 0.025 cm/s in stage 4;(ii) it was followed by an initially decreasing specific discharge in stage five, and an increasingspecific discharge at varying rates to the end of the test. The variation of εa:∆u exhibited acurvilinear response comprising two sequences (see Fig. 4.10b): (i) the axial strain increasedto a very small value of εa = 0.04 % at ∆u = 1.8 kPa; (ii) it was followed by a series of discreteincrements of increasing axial strain to a end-of-test value εa = 0.70 %. The variation of εv:∆ualso exhibited a curvilinear response comprising two sequences (see Fig. 4.10c): (i) the volu-metric strain increased to a small value εv = 0.16 % at ∆u = 1.8 kPa; (ii) it was followed by anincreasing volumetric strain, at an approximately constant rate, to an end-of-test value of εv =2.19 %. Post-test particle size analysis indicated a slightly greater finer fraction content of S f =0.39 in the top layer of the specimen than the finer fraction content of S f = 0.33 to 0.36 in theother four layers of the specimen (see Fig. D.32).4.2.7.4 Test 6.0GB35-150The specimen was isotropically consolidated to p′c = 152 kPa, l = 102 mm and ec = 0.34 (seeTable 4.1). It was subject to 19 stages of seepage flow: the test commenced with TDH = 1 cmfor ∆t = 50 min, yielding ∆u = 0.1 kPa and v = 0.002 cm/s; it was terminated at TDH = 115cm after t = 320 min, when the volume change measurement device approached the limit of itsworking range, yielding ∆u = 6.9 kPa and v = 0.047 cm/s (see Table 4.2).The variation of v:∆u exhibited a curvilinear response in three sequences (see Fig. 4.10a):(i) the relation was directly proportional to v = 0.027 cm/s at ∆u = 2.7 kPa; (ii) it was followedby an initial small increase at a smaller rate to v = 0.028 cm/s at ∆u = 3.1 kPa and a subsequentdecrease to a local minimum of v = 0.023 cm/s at ∆u = 3.4 kPa; (ii) it was then followed byan increasing specific discharge at varying rates to the end of the test. A very small εa = 0.02% was measured at the end of the test (see Fig. 4.10b). The variation of εv:∆u exhibited acurvilinear response in two sequences (see Fig. 4.10c): (i) it commenced with an increasingvolumetric strain to a small value of εv = 0.14 % at ∆u = 2.7 kpa; (ii) it was followed by anincreasing volumetric strain, at a slightly increasing rate, to an end-of-test value of εv = 2.46 %.Post-test particle size analysis did not indicate a substantial variation of fine fraction content ofS f = 0.32 to 0.38, between five layers of the specimen (see Fig. D.33).4.2.8 Gradation 6.5GB35Two tests were conducted on gradation 6.5GB35. The specimens were reconstituted by slurrydeposition and isotropically consolidated to target cell pressures of 50 kPa and 100 kPa, respec-tively. The corresponding test codes are 6.5GB35-50 and 6.5GB35-100.804.2.8.1 Test 6.5GB35-50The specimen was isotropically consolidated to p′c = 54 kPa, l = 95 mm and ec = 0.30 (seeTable 4.1). It was subject to seven stages of seepage flow: the test commenced with TDH = 1cm for ∆t = 10 min, yielding ∆u = 0.1 kPa and v = 0.003 cm/s; it was terminated at TDH = 35cm after t = 90 min, when sudden continuing deformations occurred, yielding ∆u = 1.3 kPa andv = 0.037 cm/s (see Table 4.2).The variation of v:∆u exhibited a curvilinear response in two sequences (see Fig. 4.11a):(i) there was a directly proportional relation during the first four stages to ∆u = 1.2 kPa and v =0.018 cm/s; (ii) it was followed by a decrease of specific discharge from stage four to five andthen a marked marked increase in specific discharge during the final two stages. A small axialstrain εa = 0.14 % was recorded at the end of the test (see Fig. 4.11b). The variation of εv:∆uexhibited a curvilinear response in two sequences (see Fig. 4.11c): (i) there was negligiblevolumetric strain to ∆u = 1.2 kPa; (ii) it was followed by a marked increase in volumetric strainduring the final three stages to an end-of-test value of εv = 1.89 %, accompanied by a decreasefrom ∆u = 1.4 to 1.3 kPa. A post-test particle size analysis of the specimen showed that thefiner fraction content of S f = 0.48 in the top layer was greater than the finer fraction content ofS f = 0.30 to 0.33 in the other three layers (see Fig. D.34).4.2.8.2 Tests 6.5GB35-100The specimen was isotropically consolidated to p′c = 101 kPa, l = 96mm and ec = 0.31 (seeTable 4.1). It was subject to six stages of seepage flow: the test commenced with TDH = 1 cmfor ∆t = 10 min, yielding ∆u = 0.1 kPa and v = 0.003 cm/s; it was terminated at TDH = 30 cmafter t = 60 min, when sudden, continuing deformations occurred, yielding ∆u = 1.6 kPa and v= 0.033 cm/s (see Table 4.2).The variation of v:∆u exhibited a curvilinear response in two sequences (see Fig. 4.11a):(i) the relation was directly proportional to v = 0.023 cm/s at ∆u = 1.4 kPa; (ii) it was followedby a marked increase in specific discharge to the end of the test. A very small end-of-test valueεa = 0.04 % was measured (see Fig. 4.11b). The variation of εv:∆u exhibited a curvilinearresponse in two sequences (see Fig. 4.11c): (i) there was an initial increase to a very smallvalue εv = 0.04 % at ∆u = 0.8 kPa; (ii) it was followed by a marked increase at a much greaterrate during the final three stages to an end-of-test value of εv = 1.68 %. Visual observations (seeFig. D.10) indicated that the volumetric deformation was restricted to one side of specimen. Alarge amount of fine particles was observed at the top of the specimen, after dis-assembly ofthe cell (see Fig. D.11), which was in agreement with the relatively high finer fraction contentof S f = 0.42 in the top layer, compared to the finer fraction content of S f = 0.33 to 0.35 in theother three layers (see Fig. D.35).814.3 Tests on soilsThe BT test series comprises all tests conducted on soils of sub-angular particles. The initial andthe end-of-test conditions of all tests on soils are tabulated in Tables 4.3 and 4.4, respectively.A summary of typical responses in tests on soil gradations is provided in Section 4.5.4.3.1 Gradation 5.1BT20Two tests were conducted on gradation 5.1BT20, which has a very similar particle size distribu-tion as glass beads gradation 4.8GB20. The specimens were reconstituted by slurry depositionand isotropically consolidated at target cell pressures of 50 kPa and 150 kPa, respectively. Thecorresponding test codes are 5.1BT20-50 and 5.1BT20-150.4.3.1.1 Test 5.1BT20-50The specimen was isotropically consolidated to p′c = 53 kPa, l = 101 mm and ec = 0.63 (seeTable 4.3). It was subject to 26 stages of seepage flow: the test commenced at TDH = 2 cm for∆t = 30 min, yielding ∆u = 0.2 kPa and v = 0.003 cm/s; it was terminated at TDHmax after t =400 min, yielding ∆u = 8.7 kPa and v = 0.114 cm/s (see Table 4.4).The variation of v:∆u exhibited a curvilinear response in three sequences (see Fig. 4.12a):(i) the relation was directly proportional to ∆u = 0.3 kPa and v = 0.005 cm/s; (ii) it was followedby an increase at an initially increasing rate and subsequently diminishing rate to ∆u = 5.4 kPaand v = 0.081 cm/s; (iii) it was then followed by an increasing specific discharge at increasingrate to the end of the test. A negligible axial strain (see Fig. 4.12b) and a small end-of-testvalue εv = 0.10 % (see Fig. 4.12c) were recorded. A post-test particle size analysis showed thatthe fine fraction content varied from S f = 0.19 in the top layer to S f = 0.25 in the bottom layer(see Fig. D.36).4.3.1.2 Test 5.1BT20-150The specimen was isotropically consolidated to p′c = 152 kPa, l = 102 mm and ec = 0.65 (seeTable 4.3). It was subject to 25 stages of seepage flow: the test commenced at TDH = 1 cm for∆t = 20 min, yielding ∆u = 0.1 kPa and v = 0.003 cm/s; it was terminated at TDHmax after t =280 min, yielding ∆u = 6.0 kPa and v = 0.141 cm/s (see Table 4.4).The variation of v:∆u exhibited a curvilinear response in two sequences (see Fig. 4.12a): (i)the relation was directly proportional to ∆u = 0.3 kPa and v = 0.014 cm/s; (ii) it was followed byan increase, at an initially increasing rate and subsequently diminishing rate, to the end of thetest. There was no axial strain during the test (see Fig. 4.12b). The volumetric strain increasedto a small end-of-test value εv = 0.14 % (see Fig. 4.12c). A post-test particle size analysis82established that the fine fraction content S f = 0.13 in the top layer was smaller than the finerfraction content of S f = 0.19 to 0.22 in the four other layers (see Fig. D.37).4.3.2 Gradation 5.7BT20Three tests were conducted on gradation 5.7BT20. The specimens were reconstituted by slurrydeposition and isotropically consolidated to target cell pressures of 50 kPa, 100 kPa and 150kPa, respectively. The corresponding test codes are 5.7BT20-50, 5.7BT20-100 and 5.7BT20-150.4.3.2.1 Test 5.7BT20-50The specimen was isotropically consolidated to p′c = 54 kPa, l = 101 mm and ec = 0.59 (seeTable 4.3). It was subject to 21 stages of seepage flow: the test commenced at TDH = 1 cm for∆t = 50 min, yielding ∆u = 0.1 kPa and v = 0.003 cm/s; it was terminated at TDHmax after t =310 min, yielding ∆u = 5.4 kPa and v = 0.147 cm/s (see Table 4.4).The variation of v:∆u exhibited a curvilinear response in two sequences (see Fig. 4.13a): (i)the relation was directly proportional to ∆u = 0.3 kPa and v = 0.013 cm/s; (ii) it was followedby an increasing specific discharge, at an initially increasing rate and subsequently diminishingrate, to the end of the test. A negligible axial strain (see Fig. 4.13b) and small volumetric strainεv = 0.10 % (see Fig. 4.13c) were recorded at the end of the test. A post-test particle sizeanalysis established that the fine fraction content S f = 0.17 in the top layer was smaller than thefiner fraction content of S f = 0.20 to 0.23 in the other layers (see Fig. D.38).4.3.2.2 Test 5.7BT20-100The specimen was isotropically consolidated to p′c = 102 kPa, l = 101 mm and ec = 0.56 (seeTable 4.3). It was subject to 18 stages of seepage flow: the test commenced at TDH = 1 cm for∆t = 40 min, yielding ∆u = 0.1 kPa and v = 0.002 cm/s; it was terminated at TDHmax after t =280 min, yielding ∆u = 5.3 kPa and v = 0.146 cm/s (see Table 4.4).The variation of v:∆u exhibited a curvilinear response in two sequences (see Fig. 4.13a):(i) the relation was directly proportional to v = 0.011 cm/s at ∆u = 0.2 kPa; (ii) it was followedby an increasing specific discharge, at an initially increasing rate and subsequently diminishingrate, to the end of the test. There was no axial strain during the test (see Fig. 4.13b) and onlya very small volumetric strain, with an end-of-test value εv = 0.06 % (see Fig. 4.13c). A post-test particle size analysis indicated that the fine fraction content S f = 0.13 in the top layer wassmaller than the finer fraction content of S f = 0.21 to 0.26 in the other layers (see Fig. D.39).834.3.2.3 Test 5.7BT20-150The specimen was isotropically consolidated to p′c = 154 kPa, l = 102 mm and ec = 0.57 (seeTable 4.3). It was subject to 23 stages of seepage flow: the test commenced at TDH = 1 cm for∆t = 30 min, yielding ∆u = 0.1 kPa and v = 0.002 cm/s; it was terminated at TDHmax after t =300 min, yielding ∆u = 5.8 kPa and v = 0.141 cm/s (see Table 4.4).The variation of v:∆u exhibited a curvilinear response in two sequences (see Fig. 4.13a):(i) the relation was directly proportional to v = 0.008 cm/s at ∆u = 0.2 kPa; (ii) it was followedby an increasing specific discharge, at an initially increasing rate and subsequently diminishingrate, to the end of the test. There was no axial strain during the test (see Fig. 4.13b) and thevolumetric strain increased to a small end-of-test value εv = 0.17 % (see Fig. 4.13c). A post-testparticle size analysis indicated that the fine fraction content of S f = 0.18 to 0.19 in the top twolayers was somewhat lower than the finer fraction content of S f = 0.20 to 0.23 in the bottomthree layers (see Fig. D.40).4.3.3 Gradation 5.7BT35One test was conducted on gradation 5.7BT35, which has a nearly identical particle size distri-bution as glass beads gradation 6.0GB35. The specimen was reconstituted by slurry depositionand isotropically consolidated to a target cell pressure of 100 kPa. The test code is 5.7BT35-100.4.3.3.1 Test 5.7BT35-100The specimen was isotropically consolidated to p′c = 102 kPa, l = 102 mm and ec = 0.46 (seeTable 4.3). It was subject to 24 stages of seepage flow: the test commenced at TDH = 5 cm for∆t = 50 min, yielding ∆u = 0.3 kPa and v = 0.006 cm/s; it was terminated at TDHmax after t =360 min, yielding ∆u = 8.6 kPa and v = 0.114 cm/s (see Table 4.4).The variation of v:∆u exhibited a directly proportional response throughout the test (seeFig. 4.14a). A negligible axial strain (see Fig. 4.14b) and a small volumetric strain εv = 0.22 %(see Fig. 4.14c) were recorded at the end of the test. A post-test particle size analysis showedthat the finer fraction content varied from S f = 0.32 to 0.38 in the specimen (see Fig. D.41).4.3.4 Gradation 7.0BT20Two tests were conducted on gradation 7.0BT20. The specimens were reconstituted by slurrydeposition and isotropically consolidated to target cell pressures of 50 kPa and 150 kPa, respec-tively. The corresponding test codes are 7.0BT20-50 and 7.0BT20-150.844.3.4.1 Test 7.0BT20-50The specimen was isotropically consolidated to p′c = 56 kPa, l = 99 mm and ec = 0.55 (see Table4.3). It was subject to 23 stages of seepage flow: the test commenced at TDH = 1 cm for ∆t =30 min, yielding ∆u = 0.1 kPa and v = 0.003 cm/s; it was terminated at TDHmax after t = 320min, yielding ∆u = 3.3 kPa and v = 0.167 cm/s (see Table 4.4).The variation of v:∆u exhibited a curvilinear response in two sequences (see Fig. 4.15a): (i)the relation was directly proportional to ∆u = 0.2 kPa and v = 0.020 cm/s; (ii) it was followedby an increasing specific discharge at an initially slightly increasing rate and subsequently di-minishing rate to the end of the test. There was no axial strain during the test (see Fig. 4.15b).A very small volumetric strain εv = 0.07 % (see Fig. 4.15c) was measured at the end of the test.A post-test particle size analysis indicated that the fine fraction content of S f = 0.18 to 0.19 inthe top two layers was lower than the finer fraction content of S f = 0.19 to 0.22 in the bottomthree layers (see Fig. D.42).4.3.4.2 Test 7.0BT20-150The specimen was isotropically consolidated to p′c = 153 kPa, l = 100 mm and ec = 0.53 (seeTable 4.3). It was subject to 20 stages of seepage flow: the test commenced at TDH = 1 cm for∆t = 50 min, yielding ∆u = 0.1 kPa and v = 0.003 cm/s; it was terminated at TDHmax after t =320 min, yielding ∆u = 4.0 kPa and v = 0.158 cm/s (see Table 4.4).The variation of v:∆u exhibited a curvilinear response in two sequences (see Fig. 4.15a): (i)the relation was directly proportional to ∆u = 0.3 kPa and v = 0.014 cm/s; (ii) it was followedby an increasing specific discharge, at an initially increasing rate and subsequently diminishingrate, to the end of the test. There was no axial strain during the test (see Fig. 4.15b). Thevolumetric strain increased to a small end-of-test value of εv = 0.12 % (see Fig. 4.15c). Apost-test particle size analysis showed a substantially lower finer fraction content of S f = 0.14in the top layer than the finer fraction content of S f = 0.19 to 0.26 in the other layers (see Fig.D.43).4.3.5 Gradation 7.0BT35One test was conducted on gradation 7.0BT35. Tests on similar gradation of glass beads(6.5GB35, see Section 4.2.8) exhibited substantial contractive volume changes. The specimenof gradation 7.0BT35 was reconstituted by slurry deposition and isotropically consolidated toa target cell pressure of 50 kPa. The tests code is 7.0BT35-50.854.3.5.1 Test 7.0BT35-50The specimen was isotropically consolidated to p′c = 54 kPa, l = 102 mm and ec = 0.42 (seeTable 4.3). It was subject to twenty-eight stages of seepage flow: the test commenced at TDH= 1 cm for ∆t = 50 min, yielding ∆u = 0.1 kPa and v = 0.006 cm/s; it was terminated at TDHmaxafter t = 380 min, yielding ∆u = 9.5 kPa and v = 0.104 cm/s (see Table 4.4).The variation of v:∆u exhibited a response in two sequences (see Fig. 4.16a): (i) the rela-tion was directly proportional to ∆u = 7.3 kPa and v = 0.084 cm/s; (ii) it was followed by anincreasing specific discharge, at an initially decreasing and subsequently constant rate, to theend of the test. A very small axial strain εa = 0.02 % (see Fig. 4.16b) and a small volumetricstrain εv = 0.16 % (see Fig. 4.16c) were recorded at the end of the test.4.3.6 Gradation 8.6BT20One test was conducted on gradation 8.6BT20. The specimen was reconstituted by slurry de-position and isotropically consolidated to a target cell pressures of 50 kPa. The test code is8.6BT20-50.4.3.6.1 Test 8.6BT20-50The specimen was isotropically consolidated to p′c = 53 kPa, l = 98 mm and ec = 0.51 (see Table4.3). It was subject to twenty-six stages of seepage flow: the test commenced at TDH = 1 cmfor ∆t = 30 min, yielding ∆u = 0.1 kPa and v = 0.002 cm/s; it was terminated at TDHmax after t= 360 min, yielding ∆u = 7.5 kPa and v = 0.126 cm/s (see Table 4.4).The variation of v:∆u exhibited a curvilinear response in two sequences (see Fig. 4.17a):(i) the relation was directly proportional to v = 0.010 cm/s at ∆u = 0.4 kPa; (ii) it was followedby an increasing specific discharge, at an initially increasing rate and subsequently diminishingrate, to the end of the test. No axial strain (see Fig. 4.17b) and a very small volumetric strain εv= 0.06 % (see Fig. 4.17c) were measured at the end of the test. A post-test particle size analysisindicated that the fine fraction content of S f = 0.14 in the top layer was substantially lower thanthe finer fraction content of S f = 0.20 to 0.25 in the other layers (see Fig. D.44).4.3.7 Gradation 8.6BT35One test was conducted on gradation 8.6BT35. The specimen was reconstituted by slurry de-position and isotropically consolidated to a target cell pressure of 50 kPa. The test code is8.6BT35-50.864.3.7.1 Test 8.6BT35-50The specimen was isotropically consolidated to p′c = 55 kPa, l = 97 mm and ec = 0.46 (see Table4.3). It was subject to 25 stages of seepage flow: the test commenced at TDH = 1 cm for ∆t =40 min, yielding ∆u = 0.1 kPa and v = 0.001 cm/s; it was terminated at TDHmax after t = 370min, yielding ∆u = 9.5 kPa and v = 0.089 cm/s (see Table 4.4).The variation of v:∆u exhibited a response in two sequences (see Fig. 4.18a): (i) the re-lation was directly proportional to ∆u = 5.3 kPa and v = 0.042 cm/s; (ii) it was followed byan increasing specific discharge, at a slightly increasing rate, to the end of the test. The axialstrain was negligible throughout the test (see Fig. 4.18b) and the volumetric strain increased toa small end-of-test value of εv = 0.15 % (see Fig. 4.18c).4.3.8 Gradation 10.4BT25One test was conducted on gradation 10.4BT25. The specimen was reconstituted by slurrydeposition and isotropically consolidated to a target cell pressure of 50 kPa. The test code is10.4BT25-50.4.3.8.1 Test 10.4BT25-50The specimen was isotropically consolidated to p′c = 53 kPa, l = 98 mm and ec = 0.44 (see Table4.3). It was subject to 24 stages of seepage flow: the test commenced at TDH = 1 cm for ∆t =30 min, yielding ∆u = 0.1 kPa and v = 0.002 cm/s; it was terminated at TDHmax after t = 280min, yielding ∆u = 8.6 kPa and v = 0.117 cm/s (see Table 4.4).The variation of v:∆u exhibited a response in three sequences (see Fig. 4.19a): (i) therelation was directly proportional to v = 0.008 cm/s at ∆u = 0.3 kPa; (ii) it was followed by anincreasing specific discharge, at an initially increasing and subsequently diminishing rate, to alocal maximum of v = 0.081 cm/s at ∆u = 3.1 kPa; (iii) it was then followed by a decreasingspecific discharge to a local minimum of v = 0.069 cm/s at ∆u = 5.0 kPa and a subsequentlyincreasing specific discharge at an approximately constant rate to the end of the test. There wasno axial strain during the test (see Fig. 4.19b). The variation of εv:∆u exhibited a curvilinearresponse in two sequences: (i) it commenced with an increasing volumetric strain to a smallvalue εv = 0.17 % at ∆u = 2.5 kPa; (ii) it was followed by an isolated event of increasingvolumetric strain to εv = 0.72 %, after which it remained constant to the end of the test (see Fig.4.19c). Visual observations (see Figs. D.12) show more coarse particles protruding from thetop half of the specimen than from the bottom half. A relatively large portion of fine particleswas observed on top of the specimen after dis-assembly of the cell (see Fig. D.13). A post-testparticle size analysis showed that spatial variation in the specimen of finer fraction content of87S f = 0.25 to 0.28 was small (see Fig. D.45).4.3.9 Gradation 10.4BT30One test was conducted on gradation 10.4BT30. The specimen was reconstituted by slurrydeposition and isotropically consolidated to a target cell pressure of 50 kPa. The test code is10.4BT30-50.4.3.9.1 Test 10.4BT30-50The specimen was isotropically consolidated to p′c = 53 kPa, l = 98 mm and ec = 0.41 (see Table4.3). It was subject to 23 stages of seepage flow: the test commenced at TDH = 1 cm for ∆t =30 min, yielding ∆u = 0.1 kPa and v = 0.001 cm/s; it was terminated at TDHmax after t = 410min, yielding ∆u = 7.4 kPa and v = 0.130 cm/s (see Table 4.4).The variation of v:∆u exhibited a response in two sequences (see Fig. 4.20a): (i) the relationwas directly proportional to ∆u = 2.0 kPa and v = 0.003 cm/s; (ii) it was followed by an increas-ing specific discharge at a varying rate to the end of the test. A very small axial strain εv = 0.02% (see Fig. 4.20b) and a small volumetric strain εv = 0.20 % (see Fig. 4.20c) were recordedat the end of the test. Visual observations (see Fig. D.14) showed signs of local distress nearthe top of the specimen and established that fine particles had migrated out of the specimen(see Fig. D.15). A post-test particle size analysis indicated that the finer fraction content in thespecimen increased from S f = 0.27 in the top layer to S f = 0.32 in the bottom layer (see Fig.D.46).4.3.10 Gradation 10.4BT35Three tests were conducted on gradation 10.4BT35. The specimens were reconstituted by slurrydeposition and isotropically consolidated to target cell pressures of 50 kPa (in two sequentialtests) and 100 kPa. The respective test codes are 10.4BT35-50, 10.4BT35-50(R) and 10.4BT35-100.4.3.10.1 Test 10.4BT35-50The specimen was isotropically consolidated to p′c = 55 kPa, l = 103 mm and ec = 0.36 (seeTable 4.3). It was subject to 31 stages of seepage flow: the test commenced at TDH = 1 cm for∆t = 40 min, yielding ∆u = 0.1 kPa and v = 0.001 cm/s; it was terminated at TDHmax after t =430 min, yielding ∆u = 8.8 kPa and v = 0.113 cm/s (see Table 4.4).The variation of v:∆u exhibited a response in two sequences (see Fig. 4.21a): (i) the relationwas directly proportional to ∆u = 1.9 kPa and v = 0.044 cm/s; (ii) it was followed by a varying88specific discharge to the end of the test. The axial strain increased to a small end-of-test valueεa = 0.13 % (see Fig. 4.21b). The variation of εv:∆u exhibited a response in two sequences (seeFig. 4.21c): (i) it commenced with an increasing volumetric strain to a small value εv = 0.13 %and ∆u = 1.8 kPa; (ii) it was followed by a series of discrete increments of increasing volumetricstrain to an end-of-test value εv = 1.74 %. Visual observations (see Fig. D.16) showed signs oflocal distress near the top of the specimen and established that fine particles had migrated outof the specimen (see Fig. D.17).4.3.10.2 Test 10.4BT35-50(R)The specimen was isotropically consolidated to p′c = 57 kPa, l = 101 mm and ec = 0.37 (seeTable 4.3). It was subject to 25 stages of seepage flow: the test commenced at TDH = 1 cm for∆t = 30 min, yielding ∆u = 0.1 kPa and v = 0.001 cm/s; it was terminated at TDHmax after t =320 min, yielding ∆u = 6.3 kPa and v = 0.139 cm/s (see Table 4.4).The variation of v:∆u also exhibited a response comprising two sequences (see Fig. 4.21a):(i) the relation was directly proportional to ∆u = 0.5 kPa and v = 0.007 cm/s; (ii) it was followedby a varying specific discharge to the end of the test. The axial strain was negligible throughoutthe test (see Fig. 4.21b). The variation of εv:∆u exhibited a response comprising two sequences(see Fig. 4.21c): (i) it commenced with an increasing volumetric strain to a very small value εv= 0.06 % at ∆u = 0.8 kPa; (ii) it was followed by a series of discrete increments of increasingvolumetric strain to an end-of-test value εv = 1.08 %. Visual observations (see Figs. D.19, D.18)showed signs of local distress on one side of the specimen and established that fine particleshad migrated out of the specimen (see Fig. D.20). A post-test particle size analysis establishedthat the finer fraction content of S f = 0.35 in top of the specimen, was somewhat greater thanthe finer fraction content of S f = 0.30 to 0.32 in the rest of the specimen (see Fig. D.47).4.3.10.3 Test 10.4BT35-100The specimen was isotropically consolidated to p′c = 102 kPa, l = 100 mm and ec = 0.38 (seeTable 4.3). It was subject to 29 stages of seepage flow: the test commenced at TDH = 1 cm for∆t = 30 min, yielding ∆u = 0.1 kPa and v = 0.001 cm/s; it was terminated at TDHmax after t =370 min, yielding ∆u = 8.3 kPa and v = 0.118 cm/s (see Table 4.4).The variation of v:∆u exhibited a response in two sequences (see Fig. 4.21a): (i) the relationwas directly proportional to ∆u = 0.8 kPa and v = 0.013 cm/s; (ii) it was followed by a varyingspecific discharge to the end of the test. There was no axial strain to ∆u = 1.1 kPa. Theaxial subsequently decreased to a small end-of-test value εa = -0.18 % (see Fig. 4.21b). Thevariation of εv:∆u exhibited a response in two sequences (see Fig. 4.21c): (i) it commencedwith an increasing volumetric strain to a very small value εv = 0.06 % and ∆u = 1.1 kPa; (ii) it89was followed by a series of discrete increments of increasing volumetric strain to an end-of-testvalue εv = 2.04 %. Visual observations (see Figs. D.21 and D.22) showed signs of local distress,which were restricted to one side of the specimen. A post-test particle size analysis showed thatthe finer fraction content of S f = 0.44 in top layer of the specimen was substantially greaterthan the finer fraction content of S f = 0.29 to 0.34 in the other layers of the specimen (see Fig.D.48).4.4 Repeatability of the test resultsThree tests, on two glass beads gradations with finer fraction contents of S f = 0.20 and 0.35,respectively, and on one soil gradation with S f = 0.35, were performed to assess the repeatabil-ity of the test results. On the matter of specimen reconstitution and consolidation, the void ratioat the end of consolidation of ec = 0.48 of test specimen 4.8GB20-50 was somewhat greaterthan the void ratio of ec = 0.45 of the companion test specimen 4.8GB20-50(R). No substantialdifferences between void ratios at the end of consolidation were observed in the ensuing repro-ducibility tests of 6.0GB35-100 and 6.0GB35-100(R); and 10.4BT35-50 and 10.4BT35-50(R).On the matter of multi-stage seepage flow, the response in tests 4.8GB20-50 and 4.8GB20-50(R) was characterised by a similar variation of specific discharge with differential pore waterpressure across the specimen: a sequence of direct proportionality to ∆u = 0.2 kPa was followedby a sequence of increasing specific discharge, at an increasing and subsequently diminishingrate, in the absence of substantial axial or volumetric deformation. The response to seepage flowin test 6.0GB35-100 was characterised by two sequences: (i) a directly proportional relation ofv with ∆u to ∆u = 1.2 kPa, in absence of substantial axial and volumetric strains; (ii) it wasfollowed by an increasing specific discharge at a varying rate, accompanied by a series of dis-crete increments of increasing axial and volumetric strains. The response in the companion test6.0GB35-100(R) was characterised in a similar manner with the second sequence commencingat ∆u = 1.8 kPa. Likewise, the response to seepage flow in tests 10.4BT35-50 and 10.4BT35-50(R) was similar and characterised by two sequences: (i) a directly proportional relation ofv with ∆u, in the absence of substantial axial and volumetric strains; (ii) it was followed by avarying specific discharge, accompanied by a series of discrete increments of increasing volu-metric strain, in the absence of the development of substantial axial strain. Accordingly, thesecomparisons indicate that the test procedure, which comprises specimen preparation, consoli-dation and multi-stage seepage flow, yields repeatable results in test specimens of glass beadsand soils, with finer fraction contents of S f = 0.20 and S f = 0.35.4.5 SynthesisTwo commissioning tests on glass beads gradations, subject to upward seepage flow in the flex-ible wall permeameter, yielded consistent variations of the response variables. In particular, the90resolution and accuracy of the volume change measurement technique are believed sufficient toquantify the onset and progression of deformation, in specimens with volume of approximately780 cm3.The main test program consisted of 23 tests on eight glass beads gradations and 16 tests onten soil gradations. The responses can be broadly clustered into five groups:• The variation of v:∆u exhibited a response in two or three sequences: (i) the relation wasinitially directly proportional; (ii) it was followed by an increasing specific discharge atan initially increasing rate and subsequently decreasing rate; (iii) in some tests, it wasfollowed by a proportional relation to the end of the test. Negligible to small axial andvolumetric strains are measured. The forensic observation at the end of the test indicatesthat the top layer is depleted of fine particles. Glass beads test 6.0GB20-150 and soil test8.6BT20-50 are typical examples of this response.• The variation of v:∆u exhibited a response in two or three sequences, similar to the pre-vious response: (i) the relation was initially directly proportional; (ii) it was followed byan increasing specific discharge at an initially increasing rate and subsequently decreas-ing rate; (iii) in some tests, it was followed by a proportional relation to the end of thetest. The variation of εv:∆u exhibited a response in three sequences: (i) it commencedwith a very small strain; (ii) it was followed by a marked, isolated increase of volumetricstrain; (iii) it was then followed by a sequence of approximately constant strain throughthe remainder of the test. Glass beads test 4.8GB20-100 and soil test 10.4BT25-50 aretypical examples of this response.• The variation of v:∆u exhibited a response in two sequences: (i) the relation was initiallydirectly proportional; (ii) it was followed by a varying specific discharge to the end ofthe test. The variation of εv:∆u exhibited a response in two sequences: (i) it commencedwith a sequence of very small volumetric strain; (ii) it was followed by an increasing vol-umetric strain. The axial strain varied from negligible to substantial. Visual observationsestablished that the volumetric deformation did not occur uniformly across the specimen.Glass beads test 6.0GB35-100 and soil test 10.4BT35-100 are typical examples of thisresponse.• The variation of v:∆u exhibited a response in two sequences: (i) the relation was initiallydirectly proportional; (ii) it was followed by a marked increase in specific dischargeduring the final stages. There was no axial strain. The variation of εv:∆u exhibited aresponse in two sequences: (i) it commenced with a sequence of very small volumetricstrain; (ii) it was followed by a marked increase in volumetric strain after which the testwas terminated. Forensic observations established a relatively high finer fraction content91in the top layer at the end of the test, compared to the other layers of the specimen. Thisresponse only occurred in glass beads tests 6.5GB35-50 and 6.5GB35-100.• The variation of v:∆u exhibited a directly proportional relation in absence of any substan-tial axial or volumetric strain. Glass beads test 3.3GB20-50 and soil test 7.0BT35-50 aretypical examples of this response.Finally, the repeatability of the test results is examined. The comparisons of three sets ofcompanion tests indicate that the test procedure yields reproducible results in test specimens ofglass beads and soils, with finer fraction contents of S f = 0.20 and S f = 0.35.92Table 4.1: Initial test conditions: glass beads gradations.Test code l ec p′c B-value(mm) (-) (kPa) (-)GB-F-100 97 0.65 102 0.983.3GB20-50 102 0.53 54 0.954.8GB20-50 103 0.48 54 0.964.8GB20-50(R) 102 0.45 53 0.954.8GB20-100 102 0.47 104 0.984.8GB20-150 101 0.44 153 0.964.8GB35-50 103 0.36 57 0.984.8GB35-100 102 0.37 103 0.964.8GB35-150 103 0.37 152 0.966.0GB20-50 103 0.45 54 0.986.0GB20-100 102 0.41 102 0.976.0GB20-150 101 0.40 151 0.986.0GB25-50 100 0.38 53 0.986.0GB25-100 100 0.37 103 0.956.0GB25-150 101 0.36 151 0.956.0GB30-50 100 0.36 53 0.956.0GB30-100 98 0.34 102 0.966.0GB30-150 102 0.34 151 0.936.0GB35-50 98 0.33 54 0.976.0GB35-100 100 0.37 102 0.956.0GB35-100(R) 102 0.37 102 0.976.0GB35-150 102 0.34 152 0.916.5GB25-100 96 0.36 102 0.966.5GB35-50 95 0.30 54 0.966.5GB35-100 96 0.31 101 0.9493Table 4.2: End-of-test conditions: glass beads gradations.Test code ∆u v εa εv t Notes(kPa) (cm/s) (-) (-) (min)GB-F-100 8.1 0.117 0.04 0.18 2803.3GB20-50 3.6 0.127 0.00 0.05 280 See Fig. D.234.8GB20-50 5.3 0.148 0.00 0.12 3104.8GB20-50(R) 6.7 0.138 -0.01 0.03 3704.8GB20-100 6.0 0.142 0.02 0.50 3104.8GB20-150 6.0 0.142 0.04 0.48 410 See Fig. D.244.8GB35-50 8.7 0.114 1.15 1.60 360 See Fig. D.14.8GB35-100 7.3 0.128 0.79 1.20 360 See Fig. D.254.8GB35-150 8.6 0.114 0.05 0.78 390 See Fig. D.266.0GB20-50 3.9 0.098 0.00 0.09 260 See Fig. D.276.0GB20-100 6.9 0.133 0.30 0.82 370 See Fig. D.286.0GB20-150 3.9 0.067 0.01 0.09 240 See Fig. D.296.0GB25-50 6.2 0.138 0.68 1.61 380 See Fig. D.26.0GB25-100 2.5 0.057 0.00 0.01 1406.0GB25-150 8.0 0.120 0.05 1.58 320 See Fig. D.306.0GB30-50 5.9 0.073 0.44 1.90 360 See Fig. D.36.0GB30-100 7.5 0.065 0.21 1.78 3506.0GB30-150 7.4 0.053 0.04 1.87 350 See Figs. D.4, D.5 and D.31,6.0GB35-50 6.6 0.039 0.21 2.19 370 See Figs. D.6 and D.76.0GB35-100 4.7 0.028 0.62 1.54 150 See Figs. D.8 and D.96.0GB35-100(R) 5.9 0.035 0.70 2.19 270 See Fig. D.326.0GB35-150 6.9 0.047 0.02 2.46 320 See Fig. D.336.5GB25-100 1.6 0.077 0.08 1.65 2706.5GB35-50 1.3 0.037 0.14 1.89 90 See Fig. D.346.5GB35-100 1.6 0.033 0.04 1.68 60 See Figs. D.10, D.11 and D.3594Table 4.3: Initial test conditions: soil gradations.Test code l ec p′c B-value(mm) (-) (kPa) (-)5.1BT20-50 101 0.63 53 0.965.1BT20-150 102 0.65 152 0.955.7BT20-50 101 0.59 54 0.965.7BT20-100 101 0.56 102 0.985.7BT20-150 102 0.57 154 0.975.7BT35-100 102 0.46 102 0.987.0BT20-50 99 0.55 56 0.977.0BT20-150 100 0.53 153 0.967.0BT35-50 102 0.42 54 0.978.6BT20-50 98 0.51 53 0.988.6BT35-50 97 0.46 55 0.9410.4BT25-50 98 0.44 53 0.9810.4BT30-50 98 0.41 53 0.9310.4BT35-50 103 0.36 55 0.9710.4BT35-50(R) 101 0.37 57 0.9110.4BT35-100 100 0.38 102 0.9695Table 4.4: End-of-test conditions: soil gradations.Test code ∆u v εa εv t Notes(kPa) (cm/s) (-) (-) (min)5.1BT20-50 8.7 0.224 0.01 0.10 400 See Fig. D.365.1BT20-150 6.0 0.141 0.00 0.14 280 See Fig. D.375.7BT20-50 5.4 0.147 0.01 0.10 310 See Fig. D.385.7BT20-100 5.3 0.146 0.00 0.06 280 See Fig. D.395.7BT20-150 5.8 0.141 0.00 0.17 300 See Fig. D.405.7BT35-100 8.6 0.114 0.00 0.22 360 See Fig. D.417.0BT20-50 3.3 0.167 0.00 0.07 320 See Fig. D.427.0BT20-150 4.0 0.158 0.00 0.12 320 See Fig. D.437.0BT35-50 9.5 0.104 0.02 0.16 3808.6BT20-50 7.5 0.126 0.00 0.06 360 See Fig. D.448.6BT35-50 9.5 0.089 0.01 0.15 37010.4BT25-50 8.6 0.117 0.00 0.72 280 See Figs. D.12, D.13 and D.4510.4BT30-50 7.4 0.130 0.02 0.20 410 See Figs. D.14, D.15 and D.4610.4BT35-50 8.8 0.113 0.13 1.74 430 See Figs. D.16 and D.1710.4BT35-50(R) 6.3 0.139 0.00 1.08 320 See Figs. D.18, D.19, D.20 and D.4710.4BT35-100 8.3 0.118 -0.18 2.04 370 See Figs. D.21, D.22 and D.4896εvεaΔuvTDHt: total test timeΔt: duration of stage of seepage flowspecimenO-CHDI-CHDFigure 4.1: Response variables.97(a)(b)(c)Figure 4.2: GB-F-100 test results.98(a)(b)(c)Figure 4.3: 6.5GB25-100 test results.99(a)(b)(c)Figure 4.4: 3.3GB20 test results.100(a)(b)(c)Figure 4.5: 4.8GB20 test results.101(a)(b)(c)Figure 4.6: 4.8GB35 test results.102(a)(b)(c)Figure 4.7: 6.0GB20 test results.103(a)(b)(c)Figure 4.8: 6.0GB25 test results.104(a)(b)(c)Figure 4.9: 6.0GB30 test results.105(a)(b)(c)Figure 4.10: 6.0GB35 test results.106(a)(b)(c)Figure 4.11: 6.5GB35 test results.107(a)(b)(c)Figure 4.12: 5.1BT20 test results.108(a)(b)(c)Figure 4.13: 5.7BT20 test results.109(a)(b)(c)Figure 4.14: 5.7BT35 test results.110(a)(b)(c)Figure 4.15: 7.0BT20 test results.111(a)(b)(c)Figure 4.16: 7.0BT35 test results.112(a)(b)(c)Figure 4.17: 8.6BT20 test results.113(a)(b)(c)Figure 4.18: 8.6BT35 test results.114(a)(b)(c)Figure 4.19: 10.4BT25 test results.115(a)(b)(c)Figure 4.20: 10.4BT30 test results.116(a)(b)(c)Figure 4.21: 10.4BT35 test results.117Chapter 5Analysis of the test resultsIn this Chapter, the results of the commissioning tests and of the tests constituting the maintest program are analysed to establish the phenomenological responses of each test (researchobjective No. 4). The distinct seepage-induced instability responses of suffusion, suffosion andfluidisation were identified in the literature (see Section 2.1). The responses, which are charac-terised by mass loss, volume change and change in hydraulic conductivity, are used to informthe analysis of each test presented herein. Accordingly, the following evidence is considered toestablish the phenomenological response of the test specimens to seepage flow: 1) the potentialfor particle migration and particle rearrangement is determined by evaluating the characteristicsof the micro-structure of the gradation; 2) the variation of the hydraulic conductivity is quan-tified based on the variation of the specific discharge and the differential pore water across thespecimen; 3) the change of the element volume is quantified by the variation of the volumetricstrain; 4) any mass loss is inferred from the variation of hydraulic conductivity in conjunctionwith the variation of element volume; and, for some but not all tests, 5) forensic evidence ofmass loss or particle migration, presented in Appendix D.The calculation of the variation of the hydraulic conductivity, and the inference of any massloss from the variation of hydraulic conductivity volume change, are common to all gradationsand are therefore discussed first (see Section 5.1). The procedure for determining the micro-structure is presented in Section 5.2; the identification of the phenomenological response ofgradations of glass beads (GB test series), and gradations of soils (BT test series) to seepageflow, is presented in Sections 5.3 and 5.4, respectively. A synthesis of the analysis, includinga summary of quintessential responses identified in glass beads and soil tests, is presented inSection 5.5.1185.1 Analysis of the seepage regimeAnalysis of seepage flow in instability tests has been commonly undertaken to investigate cer-tain aspects of the phenomena (e.g. Chang and Zhang, 2013; Chapuis et al., 1996; Garner andSobkowicz, 2002; Ke and Takahashi, 2012; Lafleur et al., 1989; Li, 2008; Moffat et al., 2011;Skempton and Brogan, 1994; Wan and Fell, 2004). In this study, the seepage flow is analysedby calculating the hydraulic conductivity k using Eq. 2.11, with ∆ φ = ∆u / γw, and subse-quently evaluating the causes for changes in the seepage flow. Importantly, in the absence ofany change to the porous medium itself, the seepage flow is controlled by the flow regime. Thetransition from the Darcy flow regime to a laminar inertial flow regime (see Section 2.3) canoccur at relatively low specific discharge, at Reynolds numbers Re = 1 to 10 (Bear, 1972). It isassociated with a non-linear relation between v and ∆u. As an approximation to the upper limitof the Darcy flow regime, the specific discharge v at Re = 1, vRe=1, is determined for each test,using Eq. 2.12 with Ls = D′50.Additionally, the hydraulic conductivity is very sensitive to changes of the fabric of theporous medium. Inspection of the Kozeny-Carman Equation, Eq. 2.14, establishes that:• Particle migration from an element, in the absence of volume change, results in a greatervoid ratio, and hence, a greater hydraulic conductivity.• Contractive volume change of an element, in the absence of particle migration, results ina lower void ratio, and hence, a smaller hydraulic conductivity. Gradual, and generallysmall, contractive volume changes are associated with consolidation of a specimen sub-ject to an increasing mean effective stress1. Alternatively, contractive deformations areassociated with the rearrangement of particles, or even collapse of an open, low-densityfabric (Mitchell and Soga, 2005).• Expansive volume change of an element, in the absence of particle migration, results ina greater void ratio, and hence, a greater hydraulic conductivity. Small, elastic expansivedeformations occur as a result of the seepage-induced reduction of mean effective stress.Large, expansive volume changes associated with fluidisation (see Section 2.1.1) occurwhen the effective stress in the specimen reduces to zero (Terzaghi and Peck, 1948).An increase of hydraulic conductivity can thus only be attributed to particle migration orexpansive volume change. Two phenomena can yield a less than proportional increase of v with∆u: (i) a transition from the Darcy flow regime to a laminar inertial flow regime at Re = 1 to10; or, (ii) a contractive volume change as a result of particle rearrangement. A decrease ofhydraulic conductivity, observed as a decreasing v with increasing ∆u, can only be caused by1Shear-induced volume changes are not considered here, as the shear stress on the element remains constantduring multi-stage seepage flow.119contractive volume change as a result of particle rearrangement.For each test, the seepage regime is determined, based on the comparison of the variationof the specific discharge and vRe=1, and the hydraulic conductivity is calculated. Analysis ofthe variation of hydraulic conductivity and volume change, may or may not yield conclusiveevidence on the occurrence of mass loss during the test.5.2 Analysis of micro-structureKovacs (1981) noted that the stability of the packing of coarse particles in “loose clastic sed-iments” can be determined by comparing the inter-coarse void ratio to the maximum indexvoid ratio of the coarse fraction: a stable packing of coarse particles exists if es < es,max. Thisimplies that the fine particles do not contribute to the stability of the packing arrangement, incontrast to a micro-structure with es > es,max, where the fine grains must, of necessity, be loadbearing. Clast-supported micro-structures are commonly held susceptible to seepage-inducedinternal instability (e.g. Kenney and Lau, 1985; Kezdi, 1979; Kovacs, 1981; Skempton and Bro-gan, 1994; Wittmann, 1978) and the theoretical fine fraction content at which a clast-supportedmicro-structure exists can be determined assuming a credible range for the void ratios of thecoarse and fine fractions (see Section 2.2.1). Recently, Crawford-Flett (2014) introduced theconcepts of Thevanayagam et al. (2002) on the micro-structures of gap-graded mixtures, whichyielded further insights into the potential for suffusion. For glass beads, Crawford-Flett (2014)assumed theoretical values of emax = 0.91 and emin = 0.35 and distinguished between three typesof micro-structure. Considering the work of McGeary (1961) and Scott and Kilgour (1969) onpackings of spherical particles, an opportunity exists to relieve the micro-structure identifica-tion of gap-graded gradations of glass beads from its assumptions and found it on experimentalobservations instead. The concepts of Thevanayagam et al. (2002) will be respected and, wherenecessary, refined in order to attain a quantitative framework to distinguish between five typesof micro-structures, in contrast to the distinction of only three micro-structures by Crawford-Flett (2014).5.2.1 Micro-structures of glass beads gradations5.2.1.1 Construction of micro-structure identification diagramThe construction of the micro-structure identification diagram for the glass beads gradations isbased on the following assumptions:• The void ratio and micro-structure is uniform throughout the specimen.• The specimens are reconstituted using the modified slurry deposition method.120• The coarse and fine glass beads have the same shape and specific gravity and, hence,equal packing characteristics.• Glass beads have minimum and maximum index densities of emin = 0.57 and emax = 0.67,as determined experimentally and reported by McGeary (1961); Scott (1960); Scott andKilgour (1969) and the author (see Section 3.2).• The theoretical simple cubic fabric of glass beads, associated with a theoretical void ratioof ecub = 0.91 , is the loosest possible arrangement of spherical particles that can existwhere the particles are still in contact with each other. It can only exist if proper lateralsupport is provided (McGeary, 1961), for example in the case of a single column ofspheres with diameter d in a tube of equal diameter.The limits of stable arrangements of coarse spherical particles alone (S f = 0) are thus welldefined with es,min = 0.57 and es,max = 0.67, marked as reference points A (see Table 5.1) andB, respectively, in Fig. 5.1. In addition, consideration is given to the theoretical simple cubicparticle arrangement as it is the loosest possible arrangement where spherical particles are incontact with each other. The theoretical simple cubic particle arrangement, with ecub = 0.91,is marked as reference point C in Fig. 5.1. Similarly, the reference points for packings of finespherical particles alone (S f = 1) are: e f ,min = 0.57, e f ,max = 0.67, and ecub = 0.91, denoted asreference points D, E, and F, respectively, in Fig. 5.1.The lower bound of the void ratio of binary mixtures is established by determining the min-imum void ratio at each finer fraction content S f . McGeary (1961) showed that the theoreticallower bound can be attained experimentally if D′/d′ > 10. Considering the varying ratios of3.3≤ D′/d′ ≤ 6.5 of the glass beads gradations in this study, the theoretical lower bound is pre-ferred herein to simplify the conceptual construction of micro-structure identification diagram.The minimum void ratio is obtained when both the coarse component and the fine componentare in their respective densest states; it is calculated by combining Eqs. 2.2 and 2.3 with es =es,min = 0.57 and e f = e f ,min = 0.57 and solving for e, which yields e = 0.15 at S f = 0.27. Thisstate is denoted as reference point G in Fig. 5.1. The minimum void ratio at different finerfraction contents is then defined by the line connecting reference points A, G and D.Conceptually, any packing arrangement on the line connecting A and G, where es = es,min,can be characterised as a true clast-supported micro-structure: the coarse particle packing isinherently stable and does not require the support of the fine particles. The fine particles arethus presumed non-load bearing. The true clast-supported micro-structure is very similar tocase i of Thevanayagam et al. (2002). A true clast-supported micro-structure can also exist forpacking arrangements of coarse particles at a looser states than the maximum density packing,with es,min ≤ es ≤ es,max. This notion leads to the upper boundary B-H of true clast-supported121micro-structures where es = es,max. The fine particles are also presumed non-load bearing. Thevoid ratio and finer fraction content of reference point H, with es = es,max = 0.67 and e f = es,min= 0.57, can be calculated by combining Equations 2.2 and 2.3, yielding e = 0.17 at S f = 0.30.Reference point H is identical to the theoretical critical finer fraction content of Skempton andBrogan (1994) (see Eq. 2.1), with es = es,max = 0.67 and e f = es,min = 0.57.Now consider the case where the coarse particle arrangement of a true clast-supportedmicro-structure becomes looser, i.e. es,max < es < ecub. Such a packing arrangement in whichthe coarse particles are still in contact with each other can only exist with lateral support, whichis postulated to be provided by the fine particles filling the inter-coarse voids. A portion of thefine particles is thus presumed to be load-bearing. The upper limit of a transitional (referringto the support of the fine particles) clast-supported micro-structure is the simple cubic pack-ing arrangement of coarse particles, which is defined by the line connecting reference pointsC and I in Fig. 5.1, where I corresponds to the case where es = ecub and e f = e f ,min, yield-ing e = 0.21 and S f = 0.37. The transitional clast-supported micro-structure is very similar tocase ii of Thevanayagam et al. (2002), but the boundaries of this case are now explicitly defined.Introducing a small percentage of coarse particles to a ‘mixture’ of only fine particles atreference point D, would yield fully dispersed coarse particles in a matrix of fine particles. Thedomain of a micro-structure of fine particles with fully dispersed coarse particles characterisesa true matrix-supported micro-structure. This micro-structure, with its stable fine particle pack-ing, can only exist with e f ,min ≤ e f ≤ e f ,max. The true matrix-supported micro-structure is verysimilar to case iv-1 of Thevanayagam et al. (2002). According to Thevanayagam et al. (2002),the coarse particles contribute to the stability of the micro-structure if spaced less than ten par-ticle diameters apart. The measurements of McGeary (1961), who found that the influence oftwo boundaries negates when they are spaced more than ten particle diameters apart, justify theuse of this value. The limiting finer fraction content, Sf,L, above which the distance betweenthe coarse particles in a matrix-supported micro-structure is at least ten times the diameter, canbe then calculated using Eq. 2.5, with a = 10. For the glass beads gradations, typical values areD′ = 1.0 mm and d′ = 0.15 mm, which yields Rd = 6.7 as a first approximation. The limitingfiner fraction contents and corresponding void ratio at es = es,min are then Sf,L = 0.94 and e =0.54, which is denoted as reference point J in Fig. 5.1. Similarly, for es = es,max, Sf,L = 0.95 ande = 0.63, which is denoted as reference point K in Fig. 5.1. The domain between the matrix-supported micro-structures and the transitional clast-supported micro-structures is consideredto represent a transitional matrix-supported micro-structure. The transitional matrix-supportedmicro-structure is very similar to case iv-2 of Thevanayagam et al. (2002). The limit for a tran-sitional matrix-supported micro-structure is determined by considering a packing arrangementwhere coarse particles are just separated from each other by infinitesimally small fine particles;122i.e. the coarse particle packing is just slightly looser than the theoretical simple cubic packingarrangement. The distance between the coarse particles is thus occupied by a singular fine par-ticle, which in the limit also yields a simple cubic packing arrangement of fine particles. Thiscase, with es = e f = ecub, yields e = 0.29 at S f = 0.32, which is denoted by L on Fig. 5.1.Finally, the domain of micro-structures at states looser than the limits of the transitionalclast-supported and transitional matrix-supported micro-structures, is referred to as transitional.The fine particles are presumed to separate the coarse grains in these micro-structures, resultingin relatively open particle arrangements, typically with es > ecub and e f > ecub. The transitionalmicro-structure is very similar to case iii of Thevanayagam et al. (2002).5.2.1.2 Micro-structure typesAccordingly, the distinct micro-structures, and their potential for the presence of non-load bear-ing fine particles, and potential for rearrangement of coarse particles are:1. Clast-supported micro-structure, Type C (see Fig. 5.2a): only the coarse particles areload-bearing. The inter-coarse void ratio is smaller than the maximum void ratio of thecoarse particles: es,min < es < es,max, which indicates no potential for the rearrangementof coarse particles and a large portion of non-load bearing fine particles.2. Transitional clast-supported micro-structure, Type C-T (see Fig. 5.2b): the micro-structureis dominated by the inter-coarse particle contacts, with es,max < es < ecub, but at least aportion of the fine particle fraction provides support to the coarse particle packing ar-rangement. Therefore, a portion of the fine particles in this micro-structure is non-loadbearing, while the relatively high inter-coarse void ratio indicates a potential for coarse-particle rearrangement.3. Transitional micro-structure, Type T (see Fig. 5.2e): this relatively open micro-structurecomprises coarse and fine particles, which exist at states that are typically looser thanthe equivalent cubic packing arrangement; e f > ecub and es > ecub. The relatively highinter-coarse and inter-fine void ratios further suggest a potential for contractive particlerearrangement or even collapse of the micro-structure, towards states of es,max and e f ,max,respectively. Non load-bearing fine and coarse particles may or may not exist in a veryopen micro-structure.4. Transitional matrix-supported micro-structure, Type M-T (see Fig. 5.2c): the micro-structure is dominated by the inter-fine particle contacts with e f ,min < e f < ecub andsupported by partially dispersed coarse particles at es > ecub and S f < Sf,L. Therefore, avery large portion of the fine particles is expected to be load-bearing, while there is poten-tial for the rearrangement of particles to es,max and e f ,max, which would yield contractive123volumetric deformation.5. Matrix-supported micro-structure, Type M (see Fig. 5.2d): the stability of the micro-structure is governed by the fine particles alone, e f ,min < e f < e f ,max, as the coarseparticles are fully embedded in the fine particle packing arrangement at a finer fractioncontent S f > Sf,L. Therefore, all fine particles are deemed load-bearing and there appearsno potential for rearrangement of particles.The micro-structure identification diagram constructed for glass beads gradations in thisstudy, distinguishes between five types of micro-structure, as originally proposed by The-vanayagam et al. (2002). Considering that the influence of the specimen reconstitution tech-nique on the structure, defined by particle arrangement, composition and inter-particle forces,is not characterised by the type micro-structure (see Section 2.2.1), the following comments arelimited to specimens reconstituted by slurry deposition. In general, the re-introduction (in com-parison to the micro-structure identification diagram for glass beads gradations of Crawford-Flett, 2014), of transitional clast-supported type C-T and transitional matrix-supported typeM-T micro-structures, illustrates that both fine and coarse particles are load-bearing in a largedomain of admissible packing states of binary mixtures. It is noteworthy that the boundariesbetween type C and type C-T micro-structures and between type M and type M-T micro-structures, respectively, are based on measurements of the minimum and maximum indexvoid ratios of the fine and coarse fractions. The true clast-supported type C and matrix-supported type M micro-structures are thus restricted to well defined domains at relativelylow and high fine fraction contents, respectively. Accordingly, caution is warranted in inter-preting the response of type C-T and type M-T micro-structures, based on type C and type Mmicro-structures, respectively, as it may lead to misconceptions regarding the contribution of thecoarse and fine particles to the stability of the particle fabric. The ability to distinguish betweentype C and type C-T micro-structures, and between type M and type M-T micro-structures,respectively, therefore yields a key improvement of the micro-structure identification diagram.The boundary between the type C-T and type T micro-structures is based on the theoreticalreference point where the coarse particles, on average, are just in contact with each other. Theboundary between the type M-T and type T micro-structures is not as well defined, as the fineand coarse particles are load bearing in both domains. The distinction between the domains isintended to illustrate the relatively open nature of the type T micro-structure, in comparison tothe type M-T micro-structure.For each test, the initial micro-structure of the specimen is determined by projecting thefiner fraction content S f (see Table 5.2) and the void ratio at the end of consolidation ec on tothe micro-structure classification diagram for glass beads (see Fig. 5.3). The micro-structure atthe end of the test cannot be determined, because the seepage-induced mass loss is not measured124(see Section 3.1.2), which yields an unknown end-of-test void ratio.5.2.2 Micro-structures of soil gradationsA micro-structure identification diagram for sub-angular particles is constructed in a similarmanner, based on the measurement of emin = 0.63 and emax = 0.80 of the coarse fraction, re-spectively. The equivalent void ratio of a cubic packing of sub-angular particles ecub = 1.09, isapproximated by assuming that the relative increase of the the void space Vv in an sub-angularparticle packing is identical to the relative increase of 19 % in Vv from emax = 0.67 to ecub =0.91 in a packing of spherical particles. The micro-structure identification diagram for the sub-angular particles is presented in Fig. 5.4, with the reference points A to L reported in Table5.1. The different types of micro-structure of soil gradations are bounded by similar referencepoints as the micro-structure of glass beads gradations. With reference to Fig. 5.4, Type Cis bounded by reference points A-B-G-H; Type C-T is bounded by reference points B-C-H-I;Type M-T is bounded by reference points I-J-K-L; Type M is bounded by reference pointsD-E-J-K; and Type T is bounded by reference points C-E-L. For each test, the micro-structureof the specimen is determined by projecting the finer fraction content S f (see Table 5.5) andthe void ratio at the end of consolidation ec on to the micro-structure classification diagram forsoils of sub-angular particles (see Fig. 5.5).5.3 Analysis of tests on glass beadsThe 25 tests on ten gradations of glass beads are analysed with respect to the assigned micro-structure type, the seepage flow and any deformation of the test specimen. Glass beads teststhat exhibited a quintessential response to seepage flow are accentuated herein, and summarisedin Section 5.5.5.3.1 Gradation GB-FThe test on gradation GB-F serves as a benchmark for comparative analysis of the hydraulicconductivity of the fine component of the gap-graded glass beads gradations. The void ratioof the specimen ec = 0.65 (see Table 5.2) is very close to the loosest state emax = 0.67 of apacking of equal sized spheres. The seepage flow yields a constant hydraulic conductivityk = 0.014 cm/s and no substantial axial or volumetric deformation. Very similar hydraulicconductivities of k = 0.015 cm/s (Moffat and Fannin, 2006) and k = 0.019 cm/s (Crawford-Flett, 2014) were obtained for similarly loose specimens, tested in a rigid wall permeameter.The specific discharge at the last stage of seepage flow v = 0.117 cm/s (see Table 5.3) is smallerthan the calculated upper limit of specific discharge in the Darcy flow regime, vRe=1 = 0.257cm/s, which suggests the Darcy flow regime existed throughout the test. The constant hydraulicconductivity and the absence of any substantial volume change or change in length, (see Fig.1254.2), suggest no particle rearrangement or mass loss within the specimen . Accordingly, the testis deemed to have been terminated at a “pre-critical” condition (see Table 5.4).5.3.2 Gradation 3.3GB20The micro-structure of this specimen is identified as type C-T (see Table 5.2 and Fig. 5.3),based on the finer fraction content S f = 0.20 and the void ratio at the end of consolidation ec= 0.53. The specific discharge at the last stage of seepage flow of v = 0.127 cm/s is greaterthan the calculated specific discharge vRe=1 = 0.034 cm/s at which the transition from the Darcyflow regime to an inertial regime is anticipated. The sequence (see Fig. 4.4) of a constant initialhydraulic conductivity ki = kmax = 0.037 cm/s (see Table 5.3), and the absence of any substantialvolume change or change in length, suggest no particle rearrangement or mass loss within thespecimen. The subsequent sequence of a slight decrease to the hydraulic conductivity at thelast stage of seepage flow kl = kmin = 0.034 cm/s, in the absence of volume change, is attributedto the transition to the inertial flow regime. It also suggests no particle rearrangement or massloss occurred within the specimen. Accordingly, this quintessential test is deemed to have beenterminated at a “pre-critical” condition (see Table 5.4).5.3.3 Gradation 4.8GB20The micro-structure of the specimens of tests 4.8GB20-50, 4.8GB20-50(R), 4.8GB20-100 and4.8GB20-150 is type C-T, which exhibits non-load bearing fine particles and a potential for therearrangement of coarse particles, (see Table 5.2 and Fig. 5.3), based on S f = 0.20 and ec =0.44 to 0.48.The seepage flow in test 4.8GB20-50 exhibits an initial sequence of constant hydraulicconductivity ki = kmin = 0.022 cm/s (see Table 5.3), in the absence of any axial or volumetricdeformation (see Fig. 4.5), which suggests the fabric of the specimen did not change. It isfollowed by a sequence of increasing hydraulic conductivity to kmax = 0.032 cm/s, also in ab-sence of any substantial axial and volumetric deformation (see Table 5.4), which is attributedto migration of fine particles out of a stable, coarse-particle dominated micro-structure. Thesubsequent sequence of decreasing hydraulic conductivity to kl = 0.028 cm/s (see Table 5.3),again in the absence of any substantial deformation, commences approximately at the calcu-lated value of vRe=1 = 0.031 cm/s, at which the transition from the Darcy flow regime to theinertial flow regime is anticipated. Accordingly, the response of the specimen to seepage flow ischaracterised by an initial sequence where the fabric remains unchanged and a subsequent se-quence of migration of fine particles out of a stable, coarse-particle dominated micro-structure:this response is termed suffusion. The differential pore water pressure at the onset of suffusion∆usu = 0.2 kPa, is by definition the greatest differential pore water pressure at which a constanthydraulic conductivity is maintained, prior to an increase in hydraulic conductivity. The corre-126sponding mean effective stress at the onset of suffusion is p′su = 54 kPa.The seepage flow in test 4.8GB20-50(R) yields an initially constant hydraulic conductivityki = 0.025 cm/s (see Table 5.3), followed by an increase to kmax = 0.028 cm/s and subsequentdecrease to kl = kmin = 0.020 cm/s, which is attributed to the transition to the inertial flowregime. The variation of hydraulic conductivity, in the absence of volume change (see Fig. 4.5and Table 5.4), is similarly termed suffusion with ∆usu = 0.2 kPa and p′su = 53 kPa.The seepage flow in test 4.8GB20-100 yields an initially constant hydraulic conductivityki = kmin = 0.017 cm/s (see Table 5.3), followed by an increase to a hydraulic conductivity atthe last stage of seepage flow kl = kmax = 0.025 cm/s. The increase in hydraulic conductivity isaccompanied by an isolated event of contractive volume change, in the absence of axial defor-mation (see Fig. 4.5). Although the occurrence of some contractive deformation is indicative oflimited particle rearrangement, this isolated event does not appear to progress with increasinghydraulic load. The predominant response of increasing hydraulic conductivity, in the absenceof progressive development of volumetric deformation, is attributed to the migration of fineparticles out of a largely stable, coarse-particle dominated micro-structure. Considering thepredominant phenomenon, the quintessential response to seepage flow is termed suffusion (seeTable 5.4) with ∆usu = 0.2 kPa and p′su = 104 kPa. Comparison (see Table 5.3) of the specificdischarge at the last stage of seepage flow v = 0.142 cm/s and vRe=1 = 0.031 cm/s, indicates atransition to the inertial flow regime occurred. It is postulated that the anticipated decrease ofhydraulic conductivity associated with the transition to the inertial flow regime, which was notmeasured, was offset by the effects of suffusion.The seepage flow in test 4.8GB20-150 yields an initially constant hydraulic conductivity ki= kmin = 0.018 cm/s (see Table 5.3), and a subsequent increase to kl = kmax = 0.024 cm/s. Theincrease in hydraulic conductivity is accompanied by an isolated event of contractive volumechange, in the absence of axial deformation (see Fig. 4.5). The predominant phenomenonof increasing hydraulic conductivity, in the absence of progressive development of volumetricdeformation, is similarly attributed to the migration of fine particles out of a largely stable,coarse-particle dominated micro-structure. The migration of fine particles is confirmed by thereduced percentage finer fraction in the top layer of the specimen at the end of the test. Theresponse to seepage flow is also termed suffusion (see Table 5.4), with ∆usu = 0.2 kPa and p′su= 153 kPa. It is postulated that the anticipated decrease of hydraulic conductivity, associatedwith the transition to the inertial flow regime, was offset by the effects of suffusion.1275.3.4 Gradation 4.8GB35The micro-structure of the specimens of tests 4.8GB35-50, 4.8GB35-100 and 4.8GB35-150 istype T, which exhibits a potential for particle rearrangement, (see Table 5.2 and Fig. 5.3), basedon S f = 0.35 and ec = 0.36 to 0.37.The response of 4.8GB35-50 exhibits an initial sequence of constant hydraulic conductivityki = kmax = 0.015 cm/s (see Table 5.3), in the absence of any axial and volumetric strain (see Fig.4.6), which suggests the fabric of the specimen did not change. The subsequent sequence of areduced and slightly varying hydraulic conductivity to kl = kmin = 0.013 cm/s, is accompaniedby the incremental development of non-uniform contractive volumetric and axial deformations,which is attributed to particle rearrangement. Accordingly, the response of the specimen toseepage flow is characterised by an initial sequence where the fabric remains unchanged and asubsequent sequence of particle rearrangement: this response is termed suffosion. Comparison(see Table 5.3) of the specific discharge at the last stage of seepage flow v = 0.114 cm/s andvRe=1 = 0.027 cm/s, indicates a transition to the inertial flow regime occurred. However, it isbelieved that the effects of suffosion have masked the anticipated decrease in hydraulic con-ductivity associated with the transition to the inertial flow regime. The differential pore waterpressure at the onset of suffosion ∆uso = 1.5 kPa, is defined as the greatest differential pore wa-ter pressure at which the original specimen volume was maintained. The corresponding meaneffective stress at the onset of suffosion is p′so = 56 kPa (see Table 5.4).The seepage flow in test 4.8GB35-100 yields an initially constant hydraulic conductivity ki= kmax = 0.018 cm/s, followed by an approximately constant hydraulic conductivity kl = kmin =0.017 cm/s (see Table 5.3). Although the end-of-test specimen appears not depleted of a largeportion of fines (see Fig. D.25), the seepage-induced incremental development of non-uniformvolumetric deformations, accompanied by the incremental development of axial and volumetricdeformations (see Fig. 4.6 and Table 5.4) is indicative of suffosion, with ∆uso = 1.6 kPa and p′so= 102 kPa. It is postulated that the effects of suffosion have obscured the anticipated decreasein hydraulic conductivity, associated with the transition to the inertial flow regime (see Table5.3).The seepage flow in test 4.8GB35-150 yields an initially constant hydraulic conductivityki = kmax = 0.017 cm/s (see Table 5.3), followed by a decreasing hydraulic conductivity to kl= kmin = 0.013 cm/s. It is postulated that the decrease in hydraulic conductivity was causedby the combined effects of contractive volume change (see Fig. 4.6) and a transition to theinertial flow regime (see Table 5.3). The small variation in finer fraction content in the post-testspecimen, indicates that the extend of migration of fine particles was limited. The seepage-induced incremental development of non-uniform contractive volumetric deformation, although128in absence of substantial axial deformation (see Table 5.4), is termed suffosion, with ∆uso = 2.8kPa and p′so = 150 kPa.5.3.5 Gradation 6.0GB20The micro-structure of the specimens of tests 6.0GB20-50, 6.0GB20-100 and 6.0GB20-150 istype C-T, which exhibits non-load bearing fine particles and a potential for the rearrangementof coarse particles, (see Table 5.2 and Fig. 5.3), based on S f = 0.20 and ec = 0.40 to 0.45.The seepage flow in test 6.0GB20-50 yields an initially constant hydraulic conductivity ki= kmin = 0.024 cm/s (see Fig. 4.7 see Table 5.3), followed by an increase to kmax = 0.027 cm/sand subsequent decrease to kl = 0.025 cm/s, which is attributed to the transition to the inertialflow regime at vRe=1 = 0.024 cm/s. The increase in hydraulic conductivity and the absence ofany substantial axial or volumetric deformation (see Table 5.4), are attributed to the migrationof fine particles out of a stable, coarse-particle dominated micro-structure. The migration offine particles is confirmed by the relatively low finer fraction content in the top layer of thespecimen at the end of the test. The response to seepage flow is termed suffusion, with ∆usu =0.2 kPa and p′su = 54 kPa.The seepage flow in test 6.0GB20-100 yields an initially constant hydraulic conductivity ki= kmax = 0.037 cm/s (see Fig. 4.7 and Table 5.3), and a subsequently decreasing hydraulic con-ductivity to kl = kmin= 0.019 cm/s, which is attributed to the transition to the inertial flow regimeat vRe=1 = 0.024 cm/s. The reduced finer fraction content in the top layer (see Fig. D.28), andthe absence of progressive volumetric deformations (see Table 5.4), suggests a seepage-inducedmigration of fine particles out of a largely stable, coarse-particle dominated micro-structure.The effect of the isolated event of contractive volume change is speculated to have negated theeffect of the migration of particles on the hydraulic conductivity. The response to seepage flowis termed suffusion. The conditions at the onset of particle migration are associated with theonset of contractive volume change at ∆usu = 0.3 kPa and p′su = 102 kPa.The seepage flow in test 6.0GB20-150 yields an initially slightly decreasing hydraulic con-ductivity from ki = 0.010 cm/s to kmin = 0.009 cm/s (see Fig. 4.7 and Table 5.3), followedby a subsequently increasing hydraulic conductivity to kmax = kl = 0.017 cm/s. Comparisonof the specific discharge at the last stage of seepage flow v = 0.067 cm/s and vRe=1 = 0.024cm/s, indicates a transition to the inertial flow regime occurred. The increase in hydraulic con-ductivity and the absence of any substantial axial or volumetric deformation (see Table 5.4),suggest migration of fine particles out of a stable, coarse-particle dominated micro-structure.The migration of fine particles is confirmed by the reduced finer fraction content S f = 0.05 inthe top layer of the specimen at the end of the test (see Fig. D.29). The quintessential response129to seepage flow is suffusion. It is postulated that the anticipated decrease in hydraulic conduc-tivity, associated with the transition to the inertial flow regime, was compensated by the effectof suffusion. The changes in hydraulic conductivity during the first stage of seepage flow, forwhich ∆u = 0.4 kPa, suggest that particle migration occurred during this stage. Accordingly,this stage is considered an upper limit for the onset of suffusion, with ∆usu = 0.4 kPa and p′su =151 kPa.5.3.6 Gradation 6.0GB25The micro-structure of the specimens of tests 6.0GB25-50, 6.0GB25-100 and 6.0GB25-150 istype C-T, which exhibits non-load bearing fine particles and a potential for the rearrangementof coarse particles, (see Table 5.2 and Fig. 5.3), based on S f = 0.25 and ec = 0.36 to 0.38.The seepage flow in test 6.0GB25-50 yields an initially constant hydraulic conductivity ki= 0.021 cm/s (see Table 5.3), followed by a varying hydraulic conductivity between kmin =0.017 cm/s and kl = kmax = 0.022 cm/s. Comparison of the specific discharge at the last stageof seepage flow v = 0.138 cm/s and vRe=1 = 0.022 cm/s, indicates a transition to the inertialflow regime occurred. The incremental development of non-uniform contractive volumetricdeformation (see Fig. 4.8 and Table 5.4), accompanied by a contractive axial deformation, isattributed to the rearrangement of the particle packing of fine and coarse particles. The volu-metric deformation is believed to have masked the transition to the inertial flow regime. Theresponse to seepage flow is suffosion and the conditions at the onset of suffosion are ∆uso = 0.7kPa and p′so = 53 kPa.The seepage flow in test 6.0GB25-100 yields an initial hydraulic conductivity ki = kmax =0.026 cm/s (see Fig. 4.8 and Table 5.3), which decreases to kmin = kl = 0.022 cm/s. The smallreduction in constant hydraulic conductivity is attributed to a transition from the Darcy flowregime to the inertial flow regime. The absence of any substantial volumetric or axial defor-mation (see Table 5.4), suggest no particle migration or rearrangement of the coarse-particledominated micro-structure. The test was terminated at “pre-critical” condition.The seepage flow in test 6.0GB25-150 yields an initially constant hydraulic conductivity ki= kmax = 0.032 cm/s (see Table 5.3) and a subsequently smaller and varying hydraulic conduc-tivity with kmin = 0.014 cm/s and kl = 0.015 cm/s. During the sequence of varying hydraulicconductivity, the specific discharge exceeded vRe=1 = 0.022 cm/s, which indicates a transitionfrom the Darcy flow regime to the inertial flow regime. The incremental development of non-uniform contractive volumetric deformation, in absence of any axial deformation (see Fig. 4.8and Table 5.4), is attributed to the rearrangement of the particle packing of fine and coarse par-ticles, which is believed to have masked the transition to the inertial flow regime. Accordingly,130the response to seepage flow is suffosion, with ∆uso = 1.4 kPa, and p′so = 150 kPa. Forensicevidence indicates that the coarse-particle rearrangement was accompanied by a migration offine particles from the top half of the specimen (see Fig. D.30).5.3.7 Gradation 6.0GB30The micro-structure of the specimens of tests 6.0GB30-50, 6.0GB30-100 and 6.0GB30-150 istype T, which exhibits a potential for particle rearrangement, (see Table 5.2 and Fig. 5.3), basedon S f = 0.30 and ec = 0.34 to 0.36.The seepage flow in test 6.0GB30-50 yields an initially constant hydraulic conductivity ki =kmin = 0.008 cm/s (see Table 5.3), and a subsequently varying hydraulic conductivity betweenkmax = 0.013 cm/s and kl = 0.012 cm/s. During the sequence of varying hydraulic conductivity,the specific discharge exceeded vRe=1 = 0.021 cm/s, which indicates a transition from the Darcyflow regime to the inertial flow regime. The development of non-uniform contractive volumet-ric and axial deformations (see Fig. 4.9 and Table 5.4) is attributed to the seepage-inducedrearrangement of the particle packing. The substantial contractive deformations are believed tohave masked the transition to the inertial flow regime. Accordingly, the response to seepageflow is suffosion with ∆uso = 3.4 kPa and p′so = 51 kPa.The seepage flow in test 6.0GB30-100 yields an initial hydraulic conductivity ki = kmax =0.017 cm/s (see Table 5.3), which decreases sharply to kmin = 0.006 cm/s, and subsequentlyvaries to kl = 0.008 cm/s. The development of contractive volumetric and axial deformation(see Fig. 4.9 and Table 5.4) is attributed to the rearrangement of the particle packing of fineand coarse particles, and is held to have obscured the transition to the inertial flow regime (seeTable 5.3). The response to seepage flow is suffosion, with ∆uso = 3.7 kPa and p′so = 100 kPa.The response to seepage flow in test 6.0GB30-150 is very similar with ki = kmax = 0.009cm/s, kmin = 0.005 cm/s and kl = 0.008 cm/s (see Table 5.3). The presence of an abundanceof fine particles on top of the specimen (see Fig. D.5) suggests the non-uniform contractivevolumetric deformation (see Fig. 4.9 and Table 5.4), which is believed to have masked thetransition to the internal flow regime (see Table 5.3), was accompanied by an upward migrationof fine particles. The response to seepage flow is identified as suffosion with ∆uso = 4.5 kPaand p′so = 149 kPa.5.3.8 Gradation 6.0GB35The micro-structure of the specimens of tests 6.0GB35-50, 6.0GB35-100, 6.0GB35-100(R) and6.0GB35-150 is type T, which exhibits a potential for particle rearrangement, (see Table 5.2 andFig. 5.3), based on S f = 0.35 and ec = 0.33 to 0.37.131The seepage flow in test 6.0GB35-50 yields an initially constant hydraulic conductivity ki= kmax = 0.009 cm/s (see Table 5.3), which subsequently decreases to kmin = 0.005 cm/s, andincreases to kl = 0.006 cm/s. During the sequence of varying hydraulic conductivity, the spe-cific discharge exceeded vRe=1 = 0.021 cm/s, which indicates a transition from the Darcy flowregime to the inertial flow regime. The development of substantial non-uniform contractive vol-umetric deformation, accompanied by much smaller axial deformation (see Fig. 4.10 and Table5.4), is attributed to a seepage-induced rearrangement of the particle packing, and is believedto have obscured the transition to the internal flow regime. The presence of an abundance offine particles in the top of the specimen suggests that the rearrangement of the particle packingwas accompanied by an upward migration of fine particles. The response to seepage flow issuffosion, with ∆uso = 1.7 kPa and p′so = 53 kPa.The seepage flow in test 6.0GB35-100 yields an initially constant hydraulic conductivity ki= kmax = 0.014 cm/s (see Table 5.3), which subsequently decreases to kmin = kl = 0.006 cm/s.The presence of an abundance of fines in the top of the specimen suggests that the developmentof non-uniform contractive volumetric and axial deformations (see Fig. 4.10 and Table 5.4),was accompanied by an upward migration of fine particles, and is held to have masked the an-ticipated the transition from the Darcy to the inertial flow regime (see Table 5.3). Accordingly,the quintessential response to seepage flow is suffosion, with ∆uso = 1.2 kPa and p′so = 101 kPa.The suffosive response in test 6.0GB35-100(2) is very similar to the response in test 6.0GB35-100 with ki = kmax = 0.016 cm/s and kl = kmin = 0.006 cm/s (see Table 5.3), accompanied bycontractive volumetric and axial deformations (see Fig. 4.10 and Table 5.4), which are believedto have obscured the transition to the inertial flow regime. The conditions at the onset of suf-fosion are ∆uso = 1.8 kPa and p′so = 101 kPa. Forensic evidence indicates that the extent of themigration of fine particles was limited (see Fig. D.32).The seepage flow in test 6.0GB35-150 yields an initially constant hydraulic conductivity ki= kmax = 0.010 cm/s (see Table 5.3), which decreases to kmin = 0.005 cm/s, and subsequentlyvaries to kl = 0.007 cm/s. Comparison of the specific discharge at the last stage of seepageflow v = 0.047 cm/s and vRe=1 = 0.021 cm/s, indicates a transition to the inertial flow regimeoccurred. The development of non-uniform contractive volumetric deformation, in absence ofaxial deformations (see Fig. 4.10 and Table 5.4), which is believed to have masked the antici-pated transition to the inertial flow regime (see Table 5.3), is attributed to the rearrangement ofthe particle packing. Accordingly, the response to seepage flow is identified as suffosion, with∆uso = 2.7 kPa and p′so = 150 kPa. Forensic evidence indicates that the extent of the migrationof fine particles was limited (see Fig. D.33).1325.3.9 Gradation 6.5GB25The micro-structure of the specimens of test 6.5GB25-100 is type C-T, which exhibits non-loadbearing fine particles and a potential for the rearrangement of coarse particles, (see Table 5.2and Fig. 5.3), based on S f = 0.25 and ec = 0.36.The seepage flow yields an initially slightly decreasing hydraulic conductivity from ki =0.032 cm/s to kmin = 0.028 cm/s (see Table 5.3). It is followed by a marked increase to kmax= 0.045 cm/s after which it decreases slightly to kl = 0.044 cm/s. During the test, the spe-cific discharge exceeded vRe=1 = 0.019 cm/s (see Table 5.3), which indicates a transition fromthe Darcy flow regime to the inertial flow regime occurred. The development of non-uniformcontractive volumetric deformation, in the absence of substantial axial deformations (see Fig.4.3 and Table 5.4), is attributed to the rearrangement of the particle packing of fine and coarseparticles, and is believed to have masked the transition to the inertial flow regime. Accordingly,the response to seepage flow is suffosion with ∆uso = 1.1 kPa and p′so = 101 kPa.5.3.10 Gradation 6.5GB35The micro-structure of the specimens of tests 6.5GB35-50 and 6.5GB25-100 is type M-T, whichexhibits a potential for particle rearrangement, (see Table 5.2 and Fig. 5.3), based on S f = 0.35,ec = 0.30 to 0.31 and Sf,L = 0.89 to 0.90.The seepage flow in test 6.5GB35-50 yields an initially constant hydraulic conductivity ki =0.015 cm/s (see Table 5.3), followed by a decrease to kmin = 0.010 cm/s and a marked increaseto kmax = kl = 0.027 cm/s at the end of the test. Comparison of the specific discharge at thelast stage of seepage flow v = 0.037 cm/s and vRe=1 = 0.017 cm/s, indicates a transition fromthe Darcy flow regime to the inertial flow regime occurred. The development of contractivevolumetric deformation (see Fig. 4.11 and Table 5.4), accompanied by small axial deforma-tions, is attributed to the rearrangement of the particle packing, and is believed to have maskedthe transition to the inertial flow regime. The presence of an abundance of fine particles in thetop of the specimen (see Fig. D.34), suggests the rearrangement of the particle packing wasaccompanied by an upward migration of fine particles. Accordingly, the response to seepageflow is suffosion with the conditions at the onset of suffosion ∆uso = 1.2 kPa and p′so = 54 kPa.The continuing deformations in the last stage are interpreted as an overall instability of the soilstructure, associated with failure. More specifically, failure is defined as continuing deforma-tion at a constant differential pore water pressure ∆u f . The differential pore water pressure andthe mean effective stress at the onset of failure are ∆u f = 1.4 kPa and p′f = 53 kPa, respectively.133The seepage flow in test 6.5GB35-100 yields an initial hydraulic conductivity ki = 0.016cm/s (see Table 5.3) which decreased slightly to kmin = 0.014 cm/s. It is followed by a markedincrease to kmax = kl = 0.020 cm/s. The development of non-uniform contractive volumetricdeformation (see Fig. 4.11 and Table 5.4), in absence of substantial axial deformations, isattributed to the rearrangement of the particle packing, which is believed to have masked theanticipated transition (see Table 5.3) to the inertial flow regime. The presence of an abundanceof fine particles in the top of the specimen (see Fig. D.35), suggests the rearrangement of theparticle packing was accompanied by an upward migration of fine particles. Accordingly, theresponse to seepage flow is suffosion with ∆uso = 0.8 kPa and p′so = 100 kPa. Failure wasreached in the last stage multi-stage seepage flow, at ∆u f = 1.6 kPa and p′f = 97 kPa.5.4 Analysis of tests on soilsThe 16 tests on ten gradations of soils are analysed with respect to the assigned micro-structuretype, the seepage flow, and any deformation of the test specimen. Soil tests that exhibited aquintessential response to seepage flow are accentuated herein, and summarised in Section 5.5.5.4.1 Gradation 5.1BT20The micro-structure of the specimens in tests 5.1BT20-50 and 5.1BT20-150 is type C-T, whichexhibits non-load bearing fine particles and a potential for the rearrangement of coarse particles,(see Table 5.5 and Fig. 5.5), based on S f = 0.20 and ec = 0.63 to 0.65.The seepage flow in test 5.1BT20-50 yields an initially constant hydraulic conductivity ki= 0.016 cm/s (see Table 5.6), followed by an increase in hydraulic conductivity to kmax = 0.025cm/s. The subsequent decrease to kl = kmin = 0.013 cm/s is attributed to the transition to theinertial flow regime at vRe=1 = 0.038 cm/s. The increase in hydraulic conductivity and absenceof any substantial axial or volumetric deformation (see Fig. 4.12 and Table 5.7), is attributedto migration of fine particles out of a stable, coarse-particle dominated micro-structure. Thevariation of S f = 0.19 to 0.25 across the end-of-test specimen (see Fig. D.36) indicates that theextend of particle migration was relatively small. The response to seepage flow is indicative of asuffusive response. The conditions at the onset of suffusion are ∆usu = 0.3 kPa and p′su = 53 kPa.The response to seepage flow in test 5.1BT20-150 was very similar with ki = 0.036 cm/s(see Table 5.6), kmax = 0.040 cm/s and kl = kmin = 0.023 cm/s, in the absence of any substantialaxial or volumetric deformations (see Fig. 4.12 and Table 5.7). Forensic observations yielded atop layer of the end-of-test specimen that was depleted of fine particles (see Fig. D.37), whichconfirms the migration of fine particles. The response is also termed suffusion, with ∆usu = 0.3kPa and p′su = 152 kPa.1345.4.2 Gradation 5.7BT20The micro-structure of the specimens in tests 5.7BT20-50, 5.7BT20-100 and 5.1BT20-150 istype C-T, which exhibits non-load bearing fine particles and a potential for the rearrangementof coarse particles, (see Table 5.5 and Fig. 5.5), based on S f = 0.20 and ec = 0.56 to 0.59.The seepage flow in test 5.7BT20-50 yields an initially constant hydraulic conductivity ki= 0.030 cm/s (see Table 5.6), followed by an increase to kmax = 0.042 cm/s. The subsequentdecrease to kl = kmin = 0.027 cm/s is attributed to the transition to the inertial flow regime atvRe=1 = 0.030 cm/s. The increase in hydraulic conductivity, and the absence of any substantialaxial or volumetric deformation (see Fig. 4.13 and Table 5.7), is attributed to the migration offine particles out of a stable, coarse-particle dominated micro-structure. Forensic observations(see Fig. D.38) confirm the migration of fine particles from the top layer of the specimen. Ac-cordingly, the response to seepage flow is suffusion, with ∆usu = 0.3 kPa and p′su = 54 kPa.The responses to seepage flow in tests 5.7BT20-100 and 5.7BT20-150 are very similar withki = 0.027 cm/s, kmax = 0.035 cm/s, kl = kmin = 0.027 cm/s in tests 5.7BT20-100 (see Fig. 4.12and Table 5.6) and ki = 0.025 cm/s, kmax = 0.030 cm/s, kl kmin = 0.024 cm/s in test 5.7BT20-150. In both tests, the migration of fine particles is inferred from the sequence of an increasinghydraulic conductivity, in the absence of any substantial axial or volumetric deformation. Aforensic observation (see Fig. D.39) confirmed the migration of a substantial portion of fineparticles from the top of the specimen in test 5.7BT20-100, The variation of finer fractioncontent S f = 0.18 to 0.23 in the end-of-test specimen 5.7BT20-150, indicates that the extend ofthe migration of fine particles was limited in this test (see Fig. D.39). In both tests, the responseis suffusion. The conditions at the onset of suffusion were ∆usu = 0.2 kPa and p′su = 102 kPa intests 5.7BT20-100, and ∆usu = 0.2 kPa and p′su = 154 kPa in test tests 5.7BT20-150 (see Table5.7).5.4.3 Gradation 5.7BT35The micro-structure of the specimen in test 5.7BT35-100 is type T, which exhibits a potentialfor particle rearrangement, (see Table 5.5 and Fig. 5.5), based on S f = 0.35 and ec = 0.46.The seepage flow yields a constant hydraulic conductivity ki = kl = 0.013 cm/s (see Fig.4.14 and Table 5.6), in the absence of substantial axial or volumetric deformation (see Table5.7). Comparison of the specific discharge at the last stage of seepage flow v = 0.114 cm/s andvRe=1 = 0.026 cm/s, suggests a transition from the Darcy flow regime to the inertial flow regimeoccurred. However, in contrast to several other tests, the anticipated decreasing hydraulic con-ductivity associated with a transition to the inertial flow regime (see Table 5.6), is not detectedin the response to seepage flow. A speculative explanation is that a limited suffusive response135occurred, but that the effect on the hydraulic conductivity was compensated by the transitionto the inertial flow regime. Considering that the variation at the end of the test of S f = 0.32 to0.38 in the specimen (see Fig. D.41) falls with the range of S f = 0.35 +/- 0.03 with which aspecimen can be reconstituted (see Section 3.3.1), no conclusive evidence is available to firmlyestablish the migration of fine particles. The constant hydraulic conductivity and absence ofany progressive substantial volume change or change in length, suggest no significant particlerearrangement or, as previously noted, no substantial mass loss within the specimen. The testis deemed to have been terminated at a “pre-critical” condition.5.4.4 Gradation 7.0BT20The micro-structure of the specimens in tests 7.0BT20-50 and 7.0BT20-150 is type C-T, whichexhibits non-load bearing fine particles and a potential for the rearrangement of coarse particles,(see Table 5.5 and Fig. 5.5), based on S f = 0.20 and ec = 0.53 to 0.55.The seepage flow in test 7.0BT20-50 yields an initially constant hydraulic conductivity ki =0.065 cm/s (see Fig. 4.15 and Table 5.6), followed by an increase in hydraulic conductivity tokmax = 0.073 cm/s. The subsequent decrease to kl = kmin = 0.049 cm/s is attributed to the tran-sition to the inertial flow regime at vRe=1 = 0.024 cm/s. The increase in hydraulic conductivity,and the absence of any substantial axial or volumetric deformation (see Table 5.7), is attributedto migration of fine particles out of a stable, coarse-particle dominated micro-structure. Thevariation of finer fraction content S f = 0.18 to 0.22 in the end-of-test specimen, indicates thatthe extend of the migration of fine particles was limited (see Fig. D.42). The response to seep-age flow is suffusion, with ∆usu = 0.2 kPa and p′su = 56 kPa.The response to seepage flow in test 7.0BT20-150 was qualitatively very similar (see Fig.4.15) with ki = kmin = 0.037 cm/s, kmax = 0.044 cm/s, kl = 0.038 cm/s (see Table 5.6), in theabsence of any substantial volume change. A forensic observation (see Fig. D.43) confirmed themigration of a substantial portion of fine particles from the top of the specimen. The responseis suffusion, with ∆usu = 0.3 kPa and p′su = 153 kPa (see Table 5.7).5.4.5 Gradation 7.0BT35The micro-structure of the specimen in test 7.0BT35-50 is type T, which exhibits a potential forparticle rearrangement, (see Table 5.5 and Fig. 5.5), based on S f = 0.35 and ec = 0.42.The initially constant hydraulic conductivity ki = kmax = 0.012 cm/s (see Fig. 4.16 andTable 5.6), and subsequent slight decrease to kmin = kl = 0.011 cm/s, which is attributed to thetransition from a Darcy to the inertial flow regime at vRe=1 = 0.021 cm/s, and the absence of anysubstantial axial or volumetric deformation (see Table 5.7), suggest no particle rearrangement136or mass loss within the specimen. This quintessential test is deemed to have been terminated ata “pre-critical” condition.5.4.6 Gradation 8.6BT20The micro-structure of the specimen in test 8.6BT20-50 is type C-T, which exhibits non-loadbearing fine particles and a potential for the rearrangement of coarse particles, (see Table 5.5and Fig. 5.5), based on S f = 0.20 and ec = 0.51.The seepage flow yields an initially constant hydraulic conductivity ki = 0.017 cm/s (see Fig.4.17 and Table 5.6), followed by an increase in hydraulic conductivity to kmax = 0.024 cm/s.The subsequent decrease to kl = kmin = 0.016 cm/s is attributed to the transition to the inertialflow regime at vRe=1 = 0.019 cm/s. The increase in hydraulic conductivity, and the absence ofany substantial axial or volumetric deformation (see Table 5.7), is attributed to migration of fineparticles out of a stable, coarse-particle dominated micro-structure. A forensic observation (seeFig. D.44) confirmed the migration of a substantial portion of fine particles from the top of thespecimen. The quintessential response to seepage flow is suffusion, with ∆usu = 0.4 kPa andp′su = 53 kPa.5.4.7 Gradation 8.6BT35The micro-structure of the specimen in test 8.6BT35-50 is type T, which exhibits a potential forparticle rearrangement, (see Table 5.5 and Fig. 5.5), based on S f = 0.35 and ec = 0.46.The seepage flow in test 8.6BT35-50 yields an initially constant hydraulic conductivityki = kmin = 0.008 cm/s (see Fig. 4.18 and Table 5.6), followed by an increase in hydraulicconductivity to kl = kmax = 0.009 cm/s. The increase in hydraulic conductivity and the absenceof any substantial axial or volumetric deformation (see Table 5.7), is attributed to migrationof fine particles out of a stable, coarse-particle dominated micro-structure. Accordingly, theresponse to seepage flow is suffusion, with ∆usu = 5.3 kPa and p′su = 52 kPa. It is postulated thatthe anticipated decreasing hydraulic conductivity, associated with the transition to the inertialflow regime (see Table 5.6), was masked by the effects of suffusion.5.4.8 Gradation 10.4BT25The micro-structure of the specimen in test 10.4BT25-50 is type C-T, which exhibits non-loadbearing fine particles and a potential for the rearrangement of coarse particles, (see Table 5.5and Fig. 5.5), based on S f = 0.25 and ec = 0.44.The seepage flow in test 10.4BT25-50 yields an initially constant hydraulic conductivity ki137= 0.019 cm/s (see Table 5.6), followed by an increase in hydraulic conductivity to kmax = 0.034cm/s and a marked decrease, which was then followed by a varying hydraulic conductivity to kl= kmin = 0.013. The initial increase in hydraulic conductivity at a small differential pore waterpressure and the absence of any substantial axial or volumetric deformation (see Fig. 4.19) is at-tributed to migration of fine particles out of a stable, coarse-particle dominated micro-structure.The subsequent decrease of hydraulic conductivity is accompanied by an isolated event of con-tractive volumetric deformation (Table 5.7), which is attributed to a small rearrangement of alargely stable coarse particle dominated micro-structure, similar to the isolated event of vol-umetric strain development in tests 4.8GB20-100 and 4.8GB20-150. It is postulated that theanticipated decreasing hydraulic conductivity, associated with the transition to the inertial flowregime (see Table 5.6), was masked by the effects of the isolated contractive volume change.Forensic observations indicated that a limited portion of fine particles had migrated from the tophalf of the specimen (see Figs. D.12, D.13, and D.45). The quintessential response to seepageflow is suffusion, with ∆usu = 0.4 kPa and p′su = 53 kPa.5.4.9 Gradation 10.4BT30The micro-structure of the specimen in test 10.4BT30-50 is type C-T, which exhibits non-loadbearing fine particles and a potential for the rearrangement of coarse particles, (see Table 5.5and Fig. 5.5), based on S f = 0.30 and ec = 0.41.The seepage flow in test 10.4BT30-50 yields an initially constant hydraulic conductivityki = kmin = 0.012 cm/s (see Fig. 4.20 and Table 5.6), followed by an increase in hydraulicconductivity to kmax = 0.021 cm/s, and a subsequent decrease to kl = 0.017 cm/s, which isattributed to a transition to the inertial flow regime. The increase in hydraulic conductivity, andthe absence of any progressive axial or volumetric deformation (see Table 5.7), is attributedto migration of fine particles out of a stable, coarse-particle dominated micro-structure. Thevariation of finer fraction content S f = 0.27 to 0.32 in the end-of-test specimen 5.7BT20-150indicates that the extend of the migration of fine particles was limited (see Fig. D.46). Theresponse to seepage flow is suffusion, with ∆usu = 2.0 kPa and p′su = 52 kPa.5.4.10 Gradation 10.4BT35The micro-structure of the specimens in tests 10.4BT35-50, 10.4BT35-50(R) and 10.4BT35-100 is type M-T, which exhibits a potential for particle rearrangement, (see Table 5.5 and Fig.5.5), based on S f = 0.35 and ec = 0.36 to 0.38 and Sf,L = 0.82.The seepage flow in test 10.4BT35-50 yields an initially constant hydraulic conductivityki = 0.011 cm/s (see Table 5.6), followed by a varying hydraulic conductivity, between kmin =0.009 cm/s and kmax = 0.023 cm/s, to kl = 0.013 cm/s. The specific discharge exceeds vRe=1138= 0.013 cm/s during the test, which indicates a transition from the Darcy flow regime to theinertial flow regime occurred. The seepage-induced incremental development of non-uniformcontractive volumetric deformation (see Fig. 4.21 and Table 5.4), in the absence of substantialaxial deformation is attributed to the rearrangement of the particle packing and is believed tohave masked the transition to the inertial flow regime. Visual observations established that therearrangement of the particle packing was accompanied by migration of fine particles out of thespecimen (see Fig. D.17). Accordingly, the response to seepage flow is suffosion with ∆uso =1.8 kPa and p′so = 54 kPa.The response to seepage flow (see Table 5.6) in tests 10.4BT35-50(R) and 10.4BT35-100 isvery similar with ki = 0.013 cm/s, kmax = 0.051 cm/s, kmin = 0.013 cm/s, kl = 0.022 cm/s in test10.4BT35-50(R), and ki = 0.013 cm/s, kmax = 0.035 cm/s, kmin = 0.009 cm/s , kl = 0.014 cm/sin test 10.4BT35-100. The seepage-induced incremental development of non-uniform contrac-tive volumetric deformation (see Fig. 4.21 and Table 5.4), in the absence of substantial axialdeformation is attributed to the rearrangement of the particle packing. Forensic observations inboth tests established that the particle rearrangement was accompanied by a migration of fineparticles from the bottom to the top of the specimen (see Figs. D.47 and D.48). The responseis suffosion. The conditions at the onset of suffosion are ∆uso = 0.8 kPa and p′so = 57 kPain test 10.4BT35-50(R), and ∆uso = 1.1 kPa and p′so = 102 kPa, in the quintessential soil test10.4BT35-100, respectively.5.5 SynthesisThe results of 25 tests on ten glass beads gradations, including the commissioning tests, and16 tests on ten soil gradations have been analysed. The analysis of the test results consideredevidence on: 1) the potential for particle migration and particle rearrangement, based on a welldefined micro-structure identification diagram; 2) the variation of the hydraulic conductivity;3) axial and volumetric strains; 4) any mass loss inferred from the variation of hydraulic con-ductivity, in conjunction with the variation of element volume; and, for some but not all tests,5) forensic observations and measurements of mass loss.A micro-structure identification diagram (Thevanayagam et al., 2002), with explicit defini-tions of all boundaries, has been constructed for gap-graded gradations of glass beads and soils,based on considerations of the experimental limits of stable particle packing arrangements ofthe fine fraction and coarse fraction. Gap-graded materials can exhibit one of five differenttypes of micro-structure: clast-supported type C; transitional clast-supported type C-T; transi-tional type T; transitional matrix-supported type M-T, and matrix supported type M. The abilityto distinguish between type C and type C-T micro-structures, and between type M and typeM-T micro-structures yields a key improvement of the micro-structure identification diagram,139because of the distinction between micro-structures that exhibit potential for particle rearrange-ment and micro-structures that do no exhibit this potential. The potential for seepage-inducedinternal instability is examined for each micro-structure type: it is inferred that type C and C-Tmicro-structures exhibit a substantial portion of non-load bearing fine particles, whereas a po-tential for particle rearrangement was established in type C-T, M-T and T micro-structures, butnot type C or M micro-structures.Analysis of the test result leads to one of two evidence-based phenomenological responses.A suffusive response was characterised by an initially constant hydraulic conductivity, in ab-sence of axial and volumetric deformation, and a subsequent increases in hydraulic conductivityat differential pore water pressures greater than a threshold value ∆usu, which is either accom-panied by negligible or very small axial and volumetric strains, or else it is accompanied bya non-progressive development of small axial or volumetric strains. A suffusive response as-sociated with negligible or very small strains, which is in agreement with the strict definitionestablished from a review of the literature (see Section 2.1.3), was established in four tests ontwo glass beads gradations and in ten tests on six soil gradations: glass beads test 6.0GB20-150and soil test 8.6BT20-50 are typical examples of this response. The somewhat broader defini-tion of suffusion associated with the non-progressive development of small axial or volumetricstrains, was invoked to characterise the response in three tests on two glass beads gradationsand in one test on a soil gradation: glass beads test 4.8GB20-100 and soil test 10.4BT25-50are typical examples of this response. The need to extend the definition of suffusion will bediscussed in Section 6.2.1. A type C-T micro-structure was identified in 17 of 18 soil and glassbeads specimens that exhibited a suffusive response; a type T micro-structure was identified inone soil specimen.A suffosive response was characterised by a sequence of an initially constant hydraulic con-ductivity, in absence of axial and volumetric deformation, and a subsequent varying hydraulicconductivity at differential pore water pressures greater than a threshold value ∆uso, accom-panied by the progressive development of non-uniform contractive volumetric deformation.Forensic evidence in a four glass beads tests and three soil tests, suggests that the rearrange-ment of the coarse particle packing was accompanied by a downstream migration of fine parti-cles. A suffosive response was established in 15 tests on six gradations of glass beads and threetests on one gradation of soils, which yielded type C-T, type T or type M-T micro-structures.Glass beads test 6.0GB35-100 and soil test 10.4BT35-100 are typical examples of a suffosiveresponse. In two tests that exhibited a suffosive response, on gradation 6.5GB35 with a typeM-T micro-structure, failure, defined as continuing deformations at a constant differential porewater pressure ∆u f , was reached.140In addition, a “pre-critical” condition, characterised by a constant hydraulic conductivity inthe absence of axial and volumetric deformation, was established in three tests on three grada-tions of glass beads and in two tests on two gradations of soils. Glass beads test 3.3GB20-50 andsoil test 7.0BT35-50 are typical examples of a test terminated at a “pre-critical” condition. It ispostulated that a “pre-critical” condition is a precursor to suffusion, suffosion or fluidisation.141Table 5.1: Reference points in the micro-structure identification diagram for glass beadsand soil gradations.Glass beads SoilsReference point e S f e S fA 0.57 0.00 0.63 0.00B 0.67 0.00 0.80 0.00C 0.91 0.00 1.09 0.00D 0.57 1.00 0.63 1.00E 0.67 1.00 0.80 1.00F 0.91 1.00 1.09 1.00G 0.15 0.27 0.18 0.28H 0.17 0.30 0.21 0.33I 0.21 0.37 0.25 0.40J1 0.54 0.94 0.60 0.95K1 0.63 0.95 0.75 0.94L 0.29 0.32 0.37 0.34Note:1 For gradations with D′/d′ = 6.7, for a grada-tion specific value of Sf,L, use Eq. 2.5.142Table 5.2: Micro-structure: glass beads test specimens.Test code S f 1 ec2 es 3 e f 4 Sf,L 5 Micro-(-) (-) (-) (-) (-) structure6GB-F-100 0 0.65 0.65 - - -3.3GB20-50 0.20 0.53 0.91 2.64 - C-T4.8GB20-50 0.20 0.48 0.86 2.42 - C-T4.8GB20-50(R) 0.20 0.45 0.81 2.23 - C-T4.8GB20-100 0.20 0.47 0.84 2.34 - C-T4.8GB20-150 0.20 0.44 0.80 2.19 - C-T4.8GB35-50 0.35 0.36 1.10 1.03 - T4.8GB35-100 0.35 0.37 1.11 1.07 - T4.8GB35-150 0.35 0.37 1.11 1.06 - T6.0GB20-50 0.20 0.45 0.81 2.26 - C-T6.0GB20-100 0.20 0.41 0.76 2.06 - C-T6.0GB20-150 0.20 0.40 0.75 1.99 - C-T6.0GB25-50 0.25 0.38 0.84 1.53 - C-T6.0GB25-100 0.25 0.37 0.83 1.49 - C-T6.0GB25-150 0.25 0.36 0.81 1.44 - C-T6.0GB30-50 0.30 0.36 0.94 1.19 - T6.0GB30-100 0.30 0.34 0.91 1.13 - T6.0GB30-150 0.30 0.35 0.92 1.15 - T6.0GB35-50 0.35 0.33 1.04 0.94 - T6.0GB35-100 0.35 0.37 1.10 1.05 - T6.0GB35-100(R) 0.35 0.37 1.10 1.05 - T6.0GB35-150 0.35 0.34 1.06 0.97 - T6.5GB25-100 0.25 0.36 0.81 1.43 - C-T6.5GB35-50 0.35 0.30 1.00 0.87 0.90 M-T6.5GB35-100 0.35 0.31 1.01 0.87 0.89 M-TNotes:1 From Table 3.2.2 From Table 4.1.3 Using Eq. 2.2.4 Using Eq. 2.3.5 Determined only to distinguish between type M and type M-T,using Eq. 2.5.6 See also Fig. 5.3.143Table 5.3: Seepage regime: tests on glass beads gradations.Test code ki1 kmin kmax kl2 v3 vRe=14 Seepage(cm/s) (cm/s) (cm/s) (cm/s) (cm/s) (cm/s) regime5GB-F-100 0.014 0.014 0.014 0.014 0.117 0.257 I3.3GB20-50 0.037 0.034 0.037 0.034 0.127 0.034 I-II4.8GB20-50 0.022 0.022 0.032 0.028 0.148 0.031 I-II4.8GB20-50(R) 0.025 0.020 0.028 0.020 0.138 0.031 I-II4.8GB20-100 0.017 0.017 0.025 0.025 0.142 0.031 I-II4.8GB20-150 0.018 0.018 0.024 0.024 0.142 0.031 I-II4.8GB35-50 0.015 0.013 0.015 0.013 0.114 0.027 I-II4.8GB35-100 0.018 0.017 0.018 0.017 0.128 0.027 I-II4.8GB35-150 0.017 0.013 0.017 0.013 0.114 0.027 I-II6.0GB20-50 0.024 0.024 0.027 0.025 0.098 0.024 I-II6.0GB20-100 0.037 0.019 0.037 0.019 0.133 0.024 I-II6.0GB20-150 0.010 0.009 0.017 0.017 0.067 0.024 I-II6.0GB25-50 0.021 0.017 0.022 0.022 0.138 0.022 I-II6.0GB25-100 0.026 0.022 0.026 0.022 0.057 0.022 I-II6.0GB25-150 0.032 0.014 0.032 0.015 0.120 0.022 I-II6.0GB30-50 0.008 0.008 0.013 0.012 0.073 0.021 I-II6.0GB30-100 0.017 0.006 0.017 0.008 0.065 0.021 I-II6.0GB30-150 0.009 0.005 0.009 0.008 0.053 0.021 I-II6.0GB35-50 0.009 0.005 0.009 0.006 0.039 0.021 I-II6.0GB35-100 0.014 0.006 0.014 0.006 0.028 0.021 I-II6.0GB35-100(R) 0.016 0.006 0.016 0.006 0.035 0.021 I-II6.0GB35-150 0.010 0.005 0.010 0.007 0.047 0.021 I-II6.5GB25-100 0.032 0.028 0.045 0.044 0.077 0.019 I-II6.5GB35-100 0.015 0.010 0.027 0.027 0.037 0.017 I-II6.5GB35-50 0.016 0.014 0.020 0.020 0.033 0.017 I-IINotes:1 Initial hydraulic conductivity ki.2 Hydraulic conductivity at the last stage of seepage flow kl .3 Specific discharge at the last stage of seepage flow, from Table 4.2.4 Using Eq. 2.12 with Ls = D′50.5 I = Darcy flow regime; II = Inertial flow regime.144Table 5.4: Responses: tests on glass beads gradations.Test code - kmin / ki1 kmax / ki1 εa2 εv3 ∆min f 4 ∆mobs 5 Pheno- ∆usu ∆uso ∆u f p′su p′so p′f(-) (-) (%) (%) menon6 (kPa) (kPa) (kPa) (kPa) (kPa) (kPa)GB-F-100 1.0 1.0 0.04 0.18 n - PC - - - - - -3.3GB20-50 0.9 1.0 0.00 0.05 n n PC - - - - - -4.8GB20-50 1.0 1.5 0.00 0.12 y - SU 0.2 - - 54 - -4.8GB20-50(R) 0.8 1.1 -0.01 0.03 y - SU 0.2 - - 53 - -4.8GB20-100 1.0 1.5 0.02 0.50 y - SU 0.2 - - 104 - -4.8GB20-150 1.0 1.3 0.04 0.48 y y SU 0.2 - - 153 - -4.8GB35-50 0.9 1.0 1.15 1.60 n - SO - 1.5 - - 56 -4.8GB35-100 0.9 1.0 0.79 1.20 n n SO - 1.6 - - 102 -4.8GB35-150 0.8 1.0 0.05 0.78 n n SO - 2.8 - - 150 -6.0GB20-50 1.0 1.1 0.00 0.09 y y SU 0.2 - - 54 - -6.0GB20-100 0.5 1.0 0.30 0.82 o y SU 0.3 - - 102 - -6.0GB20-150 0.9 1.7 0.01 0.09 y y SU 0.4 - - 151 - -6.0GB25-50 0.8 1.0 0.68 1.61 o - SO - 0.7 - - 53 -6.0GB25-100 0.8 1.0 0.00 0.01 n - PC - - - - - -6.0GB25-150 0.4 1.0 0.05 1.58 o y SO - 1.4 - - 150 -6.0GB30-50 1.0 1.6 0.44 1.90 o - SO - 3.4 - - 51 -6.0GB30-100 0.4 1.6 0.21 1.78 o - SO - 3.7 - - 100 -6.0GB30-150 0.6 1.0 0.04 1.87 o y SO - 4.5 - - 149 -6.0GB35-50 0.6 1.0 0.21 2.19 o y SO - 1.7 - - 53 -6.0GB35-100 0.4 1.0 0.62 1.54 o y SO - 1.2 - - 101 -6.0GB35-100(R) 0.4 1.0 0.70 2.19 o y SO - 1.8 - - 101 -6.0GB35-150 0.5 1.0 0.02 2.46 o n SO - 2.7 - - 150 -6.5GB25-100 0.9 1.4 0.08 1.65 o - SO - 1.1 - - 101 -6.5GB35-50 0.7 1.8 0.14 1.89 o y SO-F - 1.2 1.4 - 54 536.5GB35-100 0.9 1.3 0.04 1.68 o y SO-F - 0.8 1.6 - 100 97Notes:1 Calculated using Table 5.3; 2 End-of-test axial strain, from Table 4.2; 3 End-of-test volumetric strain, from Table 4.2; 4 Mass loss inferredfrom analysis of seepage regime and deformations y = yes, n = no, o = inconclusive; 5 Mass loss established through forensic observations;6 PC = Pre-critical condition; SU = Suffusion; SO = Suffosion; SO-F = Suffusion and Failure.145Table 5.5: Micro-structure: soil test specimens.Test code S f 1 ec2 es3 e f 4 Sf,L5 Micro-structure(-) (-) (-) (-) (-) Type65.1BT20-50 0.20 0.63 1.04 3.15 - C-T5.1BT20-150 0.20 0.65 1.06 3.25 - C-T5.7BT20-50 0.20 0.59 0.99 2.95 - C-T5.7BT20-100 0.20 0.56 0.95 2.85 - C-T5.7BT20-150 0.20 0.57 0.96 2.85 - C-T5.7BT35-100 0.35 0.46 1.25 1.31 - T7.0BT20-50 0.20 0.55 0.94 2.75 - C-T7.0BT20-150 0.20 0.53 0.91 2.65 - C-T7.0BT35-50 0.35 0.42 1.18 1.20 - T8.6BT20-50 0.20 0.51 0.89 2.55 - C-T8.6BT35-50 0.35 0.46 1.25 1.31 - T10.4BT25-50 0.25 0.44 0.92 1.76 - C-T10.4BT30-50 0.30 0.41 1.01 1.37 - C-T10.4BT35-50 0.35 0.36 1.09 1.03 0.82 M-T10.4BT35-50(R) 0.35 0.37 1.11 1.06 0.82 M-T10.4BT35-100 0.35 0.38 1.12 1.09 0.82 M-TNotes:1 From Table 3.22 From Table 4.33 Using Eq. 2.24 Using Eq. 2.35 Determined only to distinguish between type M and type M-T, usingEq. 2.56 See also Fig. 5.5146Table 5.6: Seepage regime: tests on soil gradations.Test code ki1 kmin kmax kl2 v3 vRe=14 Seepage(cm/s) (cm/s) (cm/s) (cm/s) (cm/s) regime55.1BT20-50 0.016 0.013 0.025 0.013 0.224 0.038 I-II5.1BT20-150 0.036 0.023 0.040 0.023 0.141 0.038 I-II5.7BT20-50 0.030 0.027 0.042 0.027 0.147 0.030 I-II5.7BT20-100 0.027 0.027 0.035 0.027 0.146 0.030 I-II5.7BT20-150 0.025 0.024 0.030 0.024 0.141 0.030 I-II5.7BT35-100 0.013 0.013 0.013 0.013 0.114 0.026 I-II7.0BT20-50 0.065 0.049 0.073 0.049 0.167 0.024 I-II7.0BT20-150 0.037 0.037 0.044 0.038 0.158 0.024 I-II7.0BT35-50 0.012 0.011 0.012 0.011 0.104 0.021 I-II8.6BT20-50 0.017 0.016 0.024 0.016 0.126 0.019 I-II8.6BT35-50 0.008 0.008 0.009 0.009 0.089 0.018 I-II10.4BT25-50 0.019 0.013 0.034 0.013 0.117 0.015 I-II10.4BT30-50 0.012 0.012 0.021 0.017 0.130 0.014 I-II10.4BT35-50 0.011 0.009 0.023 0.013 0.113 0.013 I-II10.4BT35-50(R) 0.013 0.013 0.051 0.022 0.139 0.013 I-II10.4BT35-100 0.013 0.009 0.035 0.014 0.118 0.013 I-IINotes:1 Initial hydraulic conductivity ki.2 Hydraulic conductivity at the last stage of seepage flow kl .3 Specific discharge at the last stage of seepage flow, from Table 4.4.4 Using Eq. 2.12 with Ls = D′50.5 I = Darcy flow regime; II = Inertial flow regime.147Table 5.7: Responses: tests on soil gradations.Test code kmin / ki1 kmax / ki1 εa2 εv3 ∆min f 4 ∆mobs 5 Pheno- ∆usu ∆uso p′su p′so(-) (-) (%) (%) menon6 (kPa) (kPa) (kPa) (kPa)5.1BT20-50 0.8 1.6 0.01 0.10 y o SU 0.3 - 53 -5.1BT20-150 0.6 1.1 0.00 0.14 y y SU 0.3 - 152 -5.7BT20-50 0.9 1.4 0.01 0.10 y o SU 0.3 - 54 -5.7BT20-100 1.0 1.3 0.00 0.06 y y SU 0.2 - 102 -5.7BT20-150 1.0 1.2 0.00 0.17 y o SU 0.2 - 154 -5.7BT35-100 1.0 1.0 0.00 0.22 o n PC - - - -7.0BT20-50 0.8 1.1 0.00 0.07 y o SU 0.2 - 56 -7.0BT20-150 1.0 1.2 0.00 0.12 y y SU 0.3 - 153 -7.0BT35-50 0.9 1.0 0.02 0.16 n - PC - - - -8.6BT20-50 0.9 1.4 0.00 0.06 y y SU 0.4 - 53 -8.6BT35-50 1.0 1.1 0.01 0.15 y - SU 5.3 - 52 -10.4BT25-50 0.7 1.8 0.00 0.72 y y SU 0.3 - 53 -10.4BT30-50 1.0 1.8 0.02 0.20 y y SU 2.0 - 52 -10.4BT35-50 0.8 2.1 0.13 1.74 n y SO - 1.8 - 5410.4BT35-50(R) 1.1 3.9 0.00 1.10 n y SO - 0.8 - 5610.4BT35-100 0.7 2.7 -0.18 2.00 n y SO - 1.1 - 102Notes:1 Calculated using Table 5.6; 2 End-of-test axial strain, from Table 4.4; 3 End-of-test volumetric strain, from Table4.4; 4 Mass loss inferred from analysis of seepage regime and deformations y = yes, n = no, o = inconclusive; 5 Massloss established through forensic observations; 6 PC = Pre-critical condition; SU = Suffusion; SO = Suffosion; SO-F= Suffusion and Failure.148Figure 5.1: Micro-structure identification diagram for glass beads.(b) Type C-T(c) Type M-T(a) Type C(d) Type M(e) Type TFigure 5.2: Micro-structure types.149(a)(b)Figure 5.3: Identification of initial micro-structure of glass beads test specimens: a)overview; b) detail.Figure 5.4: Micro-structure identification diagram for soils.150(a)(b)Figure 5.5: Identification of initial micro-structure of soil test specimens: a) overview; b)detail.151Chapter 6Factors governing suffusion andsuffosionIn the previous Chapter suffusion was identified in seven tests on two glass beads gradationsand in eleven tests on seven soil gradations, whereas suffosion was identified in 15 tests on sixglass beads gradations and in three tests on one soil gradation. In addition, a “pre-critical” end-of-test condition, which is postulated to be a precursor to suffusion, suffosion or fluidisation,was reached in three tests on three glass beads gradations and in two tests on two soil gradations.In this Chapter, the factors governing suffusion and suffosion are examined by means ofcomparison of select sets of tests from this study, which exhibit variation in only one of theindependent variables of finer fraction content, gap ratio, particle shape, mean effective stress,and differential pore water pressure across the specimen. A reader guide to this Chapter ispresented in Fig. 6.1. The findings are compared to a limited body of evidence found in theliterature, for which an experimental database is compiled in Section 6.1. The discussion of thefactors governing suffusion and suffosion is guided by the four research hypotheses introducedin Section 1.1. Hypotheses Nos. 1 to 3 are first examined based on the tests on glass beadsof this study that exhibited suffusion or suffosion, which includes one commissioning test,supplemented by select tests found in the literature. Hypothesis No. 1, which seeks to establishthe suitability of the volume change to distinguish between suffusion and suffosion, is thentested in Section 6.2. Section 6.3 addresses the dependence of the onset of suffosion on effectivestress (hypothesis No. 2). Hypothesis No. 3, which was proposed to investigate the relationbetween the soil micro-structure and seepage-induced internal instability, is tested in Section6.4. The influence of the particle shape on seepage-induced internal instability (hypothesis No.4) is examined in Section 6.5, by the explicit comparison of select tests on glass beads and soils,for which the particle shape is the only independent variable. The factors governing suffusionand suffosion, identified in the evaluation of hypotheses Nos. 1 to 3 based on tests on glass152beads, are then further examined in Section 6.6, based on the tests on soils of this study andsupplemented by select tests on soils found in the literature. Finally, a unified approach for thecharacterisation of suffosion, and a summary of the factors governing suffusion and suffosionare presented in Sections 6.7, and 6.8, respectively.6.1 Compilation of experimental database on suffusion andsuffosionIn addition to the glass beads and soil tests of this study, a database of gap-graded materials,reported in the literature, that exhibited a suffusive or suffosive response to seepage flow iscompiled (see Table 6.1), based on the following criteria:• The gap-graded media are of sand and gravel size (either glass beads or soil).• The test results that are reported include values of effective stress, hydraulic conductivity,volume change and, preferably, include information on mass loss.Herein, important features of the test method, the materials and the response to seepage floware presented for each study. Details of the tests are presented in the discussion, if necessary.Select tests of the studies of Moffat (2005), Li (2008), Sail et al. (2011), and Crawford-Flett(2014), who all prepared specimens using the modified slurry deposition technique, and selecttests of studies of Skempton and Brogan (1994) and Chang and Zhang (2013), who used a moisttamping technique, are included in the database (see Fig. 6.2).1Crawford-Flett (2014) identified the phenomenological response to upward seepage flow offour tests on a glass beads gradation 6.6GB22, tested in a rigid wall permeameter, as suffusion.Using the same rigid wall permeameter, Li (2008) reported the response of five tests on a glassbeads gradation FR7, subject to upward or downward flow, and four tests on a glass beads gra-dation FR8, subject to downward flow. All tests exhibited suffosion. Sail et al. (2011) subjecteda glass beads gradation G4-C to downward seepage flow in a rigid wall permeameter. The re-sponse was identified as suffosion. Skempton and Brogan (1994) reported a suffusive responseon their soil gradation A, tested in a rigid wall permeameter. Chang and Zhang (2013), using aflexible wall permeameter, reported a suffusive response for four downward flow seepage testson a soil gradation GS, isotropically consolidated to varying values of mean effective stress; theresults of 18 seepage tests on the same soil gradation GS, anisotropically consolidated to vary-ing values of mean effective stress and deviatoric stress, exhibited suffosion. Moffat (2005),using a rigid wall permeameter, conducted three tests on a soil gradation T-0 subject to upwardseepage flow, and five tests on a soil gradation T-5 subject to upward or downward seepageflow; all tests on gradations T-0 and T-5 exhibited a suffosive response.1The studies of the author, Moffat (2005), Li (2008) and Crawford-Flett (2014) have all been conducted at theUniversity of British Columbia.1536.2 Test of hypothesis No. 1The first hypothesis was proposed as follows: Volume change is a characteristic variable ofseepage-induced internal instability and serves to distinguish between suffusion and suffosion.The hypothesis is first examined based on the glass beads tests of this study in Section 6.2.1.The findings are subsequently compared to select tests reported in the literature in Section 6.2.2.6.2.1 Glass beads tests of this studyIn this study, the response to seepage flow in seven tests on two gradations of glass beads wasidentified as suffusion (see Table 5.4). In Section 5.3, a suffusive response was characterised byan initial sequence of no volumetric strain and a constant hydraulic conductivity, and a subse-quent sequence of an increasing hydraulic conductivity accompanied by either: (i) a negligibleor very small volume change; or, (ii) an isolated event of a small volumetric deformation toend-of-test values of εv = 0.48 to 0.82 % (see Table 5.4), which did not progress in subsequentstages of seepage flow. The former is in agreement with the definition of suffusion establishedfrom a review of the literature (see Section 2.1.3). Considering the latter, it is speculated that,given the variation of finer fraction content of S f = +/- 0.03 and +/- 0.08 across five layers of thetrial specimens 4.8GB20 (see Fig. 3.25) and 6.0GB20 (see Figs. 3.28 and 3.29), respectively, ina few zones, locally unstable particle packings are formed during reconstitution of specimenswith S f = 0.20. These small zones of unstable particle packings may, with the slightest distur-bance, of for example, a small hydraulic load, rearrange to a stable packing. The absence ofprogressive volume change is then explained by the predominant presence of a stable particlepacking throughout the specimen and the limited spatial extend of zones of locally unstablepackings. The predominant response in such specimens is one of local migration of fine par-ticles, largely in the absence of volume change. In light of this discussion on the practicallimitations of the reconstitution of homogeneous specimens of glass beads, it seems prudent tobroaden the strict definition of suffusion, as established from the literature review, to “suffusionis characterised as seepage-induced mass loss, without change in volume, or with a small non-progressive change in volume, accompanied by an increase of hydraulic conductivity.”In addition to suffusion, a suffosive response was identifed in 15 tests on six different gra-dations of glass beads. A suffosive response was characterised by an initial sequence of novolumetric strain, and a subsequent sequence of a progressive development of substantial vol-umetric strain (see for example test 6.0GB35-100 in Fig. 4.10c), to end-of-test values rang-ing from εv = 0.78 to 2.46 % (see Table 5.4). This characterisation is in agreement with thedefinition of suffosion established from a review of the literature (see Section 2.1.3). Visualobservations established that the progressive, seepage-induced contractive volume change as-sociated with suffosion did not occur equally throughout the specimen (see Appendix D.1),but was rather of a local nature. It is speculated that the entire specimen yields a potential for154particle rearrangement, but that small spatial variations in the particle packing cause the spatialvariation in the volumetric deformation with the most susceptible zones to rearrange first. Theuse of end-of-test values of εv is limited for the purpose of comparison between tests, as thevalues depend, amongst other factors, on the differential pore water pressure at the end of thetest, which varies between tests. Notwithstanding this caveat, the end-of-test values of εv in asuffusive specimen are typically small, whereas the end-of-test values of εv in a suffosive spec-imen are substantial, and in all but one test, equal to or greater than 1.20 %. Inspection of thevariation of the volumetric strain with the differential pore water pressure across the specimenin tests on glass beads that exhibited a suffosive response (Figs. 4.8c to 4.11c), suggests thatprogression of volumetric deformation occurs at an approximately constant rate in each test.The rate of deformation appears to be very similar for a given gradation, and is independent ofeffective stress. To quantify the progression of volumetric deformation, the average unit rate ofvolumetric deformation Ev is defined as the increase of volume strain, ∆εv, per unit increase indifferential pore water pressure:Ev =∆εv∆ul−∆uso (6.1)where ∆ul is the differential pore water pressure at the last stage of seepage flow.The increase in contractive strain is calculated from the onset of suffosion (see Fig. 6.3)to the end of the test, if continuing deformations associated with failure did not occur, else ifcontinuing deformations did occur during the last stage of seepage flow (in tests on gradation6.5GB35, see Fig. 4.11c), then to the last stage prior to failure. Determination of the aver-age unit rate of deformation for every test that exhibited a suffosive response (see Table 6.2),yields a relatively narrow range of values for each gradation: 4.8GB35 yields Ev = 0.12 to 0.21%/kPa; gradation 6.0GB25 yields Ev = 0.23 to 0.28 %/kPa; gradation 6.0GB30 yields Ev = 0.44to 0.71 %/kPa; gradation 6.0GB35 yields Ev = 0.43 to 0.55 %/kPa; gradation 6.5GB35 yieldsEv = 0.75 to 0.89 %/kPa. A relatively large unit rate of volumetric deformation of Ev = 3.18%/kPa is obtained for the test on gradation 6.5GB25. The unit rate of deformation increaseswith increasing gap ratio for gradations 6.0GB25 and 6.5GB25, and for gradations 4.8GB35,6.0GB35 and 6.5GB35, respectively (see Fig. 6.4). The variation of Ev with D′15/d′85 demon-strates a strong dependence between the variables. The physical explanation is sought in theincreasing particle size of the coarse component with increasing D′15/d′85: rearrangement oflarger coarse particles apparently yields larger volumetric deformations than the rearrangementof smaller coarse particles. The average unit rate of volumetric deformation can thus be con-sidered a gradation specific variable, which, in contrast to the end-of-test volumetric strain, isnot dependent on the conditions at which the test was terminated. The relatively narrow rangeof the average unit rate of volumetric deformation, for any gradation, suggests it is a usefulvariable to quantify the volumetric deformations associated with suffosion.155Considering that, in a flexible wall permeameter, the measurement of axial deformationis simpler and less expensive than the measurement of the total volumetric deformation, thenecessity of the measurement of volumetric deformation to quantify seepage-induced internalinstability will be discussed herein. In all tests on gradations of glass beads that exhibited a suf-fosive response, comparison of the end-of-test values of axial and volumetric strains, yieldsgreater volumetric strains than axial strains (see Table 6.2). More specifically, nine of 15tests (4.8GB35-150, 6.0GB25-150, 6.0GB30-100, 6.0GB30-150, 6.0GB35-60 6.0GB35-150,6.5GB25-100, 6.5GB35-50 and 6.5GB35-100) exhibited substantial volumetric deformationsin the absence of substantial axial deformations (see Figs. 6.5 to 6.10). A tentative expla-nation for the absence of substantial axial deformation is that a partial, local collapse of themicro-structure does not significantly affect the overall stability of the specimen. These find-ings demonstrate that volumetric deformation is an essential variable to characterise seepage-induced deformations. Furthermore, it should be noted that the response to seepage flow infour tests on gradation 4.8GB20 (see Table 5.4), and in two tests on gradation 6.5GB35, isqualitatively very similar in the domain of axial deformations and hydraulic conductivity. Thesuffosive response of the tests on gradation 6.5GB35 can only be appreciated and distinguishedfrom the suffusive response of the tests on gradation 4.8GB20 in terms of the progressive in-crease of volumetric deformation during multi-stage seepage flow (compare Figs. 4.11c and4.5c, respectively). Considering the localised nature of the contractive volumetric deformationassociated with suffosion, which can occur in the absence of axial deformation, measurement oftotal volume change proves necessary to avoid any mis-interpretation of the phenomenologicalresponse to seepage flow.6.2.2 Glass beads tests of other studiesIn the experimental database compiled from the literature, the results of seepage tests on fourother gap-graded gradations of glass beads are reported. Crawford-Flett (2014) reports a suf-fusive response to seepage flow in tests on gradation 6.6GB22 (see Table 6.1), which is fairlysimilar to gradation 6.0GB20 of this study. Isolated events of contractive axial deformation,which did not progress with subsequent stages of seepage flow, were observed in all tests ongradation 6.2GB20, which yielded end-of-test values ranging from εa = 0.1 to 0.6 % (see Ta-ble 6.2). It thus appears that Crawford-Flett (2014) also, at least implicitly, adopted a notionof a small, non-progressive volume change to characterise suffusion. These findings thus sup-port the broader definition of suffusion adopted in the previous Section. Using a rigid wallpermeameter, Li (2008) reported a suffosive response in a total of nine tests on glass beadsgradations FR7 and FR8 (see Table 6.1), yielding end-of-test values ranging from εa = 0.1 to4.7 %. Sail et al. (2011) also report substantial end-of-test axial strain of εa = 4.9% in thesuffosive response of one test on glass beads gradation G4-C, using a rigid wall permeameter.156The range of axial strains measured in the tests of Li (2008) and Sail et al. (2011), on glassbeads gradations using a rigid wall permeameter, imply that equivalent substantial volumetricstrains occurred in these tests. Moreover, Li (2008) appears to have relied on visual observa-tions to identify the phenomenological response. Visual observations of particle rearrangementwere necessary to identify suffosion in test FR7-50-D, which only exhibited a very small valueof εa = 0.1 %. In two other tests, FR8-100-D and FR7-100-D, visual observations of particlerearrangement are reported as a “disturbance to the specimen” during multi-stage seepage flow,which are not captured by the axial strain measurement. It is speculated that such a disturbance,visible through the side wall of the permeameter, would have been captured if the total volumechange could have been measured. Considering the need to use visual observations, in additionto measurements of axial strain, to identify the phenomenological response, the findings of Li(2008) and Sail et al. (2011) lend further evidence to the limitation of the axial strain alone toquantify suffosion. The findings from the literature thus confirm the belief that volume changeis a necessary variable to characterise, and to distinguish between, suffusion and suffosion.6.3 Test of hypothesis No. 2Hypothesis No. 2 seeks to establish the conditions at which the onset of suffosion initiates,especially in relation to the role of effective stress: The onset of suffosion is dependent oneffective stress. The hypothesis is first tested based on the glass beads tests of this study inSection 6.3.1. The findings are subsequently compared to select tests reported in the literaturein Section 6.3.2.6.3.1 Glass beads tests of this studyThe parametric study, with the mean effective stress as one of the test variables, permits a directcomparison between specimen response, where the effective stress is the only variable thatchanges: any difference in the response can be attributed to the influence of the effective stress.The influence of the effective stress on the onset of suffosion can thus be assessed by comparingthe suffosive response in 15 tests on six gradations (4.8GB35, 6.0GB25, 6.0GB30, 6.0GB35,6.5GB25 and 6.5GB35, see Table 6.2), which have been isotropically consolidated to a range ofmean effective stress p′c = 53 to 152 kPa (see Table 4.1). For each test that exhibited a suffosiveresponse, a plot is generated of the hydro-mechanical path, which is defined by the progressionof the mean effective stress p′ of the element and the differential pore water pressure across theelement ∆u. Consider for example gradation 4.8GB35, which is the gradation with the smallestgap ratio that exhibited suffosion. The hydro-mechanical path of tests 4.8GB35-50, 4.8GB35-100 and 4.8GB35-150 is plotted in Fig. 6.11: the onset of suffosion is marked. Following aninitial response of no volume change in the tests (see also Fig. 4.6c), the onset of suffosion isassociated with the beginning of contractive volume change. The volume change progressed157in a localised manner with increasing differential pore water pressure across the specimen, butoverall instability of the soil structure did not occur. The onset of suffosion in test 4.8GB35-50at ∆uso = 1.5 kPa and p′so = 53 kPa (see Table 5.4) initiates at a slightly smaller differential porewater pressure and substantially smaller mean effective stress than the onset of suffosion in test4.8GB35-100 at ∆uso = 1.6 kPa and p′so = 103 kPa. The onset of suffosion in test 4.8GB35-150initiates at ∆uso = 2.8 kPa and p′so = 150 kPa. These data suggest that the differential pore waterpressure at the onset of suffosion increases with increasing mean effective stress. The relation isapproximately proportional with a gentle slope that may or may not pass through the origin. Thehydro-mechanical paths of the other glass beads tests that exhibited suffosion are plotted in Figs.6.12 to 6.16. In particular, the tests on gradations 6.0GB25 (see Fig. 6.12), 6.0GB30 (see Fig.6.13), 6.0GB35 (see Fig. 6.14), and 6.5GB35 (see Fig. 6.16), exhibit a trend of increasing ∆usowith increasing p′so at a gentle slope. Theoretically, in the limit of zero mean effective stress,any differential pore water pressure across an element yields fluidisation of the element. It istherefore speculated that the relation between mean effective stress and differential pore waterpressure at the onset of suffosion, must pass through the origin. Presenting the conditions at theonset of volume change of all glass beads tests in one plot, (see Fig. 6.17) indicates a moderatepositive correlation between the mean effective stress and the differential pore water pressure atthe onset of volume change, which demonstrates that the differential pore water pressure at theonset of suffosion is dependent on the mean effective stress. However, comparison of the testson gradation 6.0GB30, with S f = 0.30, gradations 6.0GB25 and 6.5GB25, with S f = 0.25, andtests on gradations 4.8GB35, 6.0GB35 and 6.5GB35, with S f = 0.35 (see Fig. 6.17), revealsa greater variation in values for ∆uso with S f than with p′so. The influence of the finer fractioncontent is addressed in Section 6.4.6.3.2 Glass beads tests of other studiesFrom the literature, only Li (2008) has reported a suffosive response in tests on specimensof glass beads subject to varying values of effective stress. Li (2008), however, was primarilyinterested in the seepage-induced failure of internally unstable gradations, and did not explicitlyreport the conditions at the onset of suffosion. The response to seepage flow of five tests ongradation FR7 and four test on gradation FR8 is re-interpreted by examining the variation ofaxial deformation, and reproduced in terms of ∆u and p′, by assuming K0 = 0.4 (after Moffat,2005). Recall that volumetric deformation has been identified to be a more correct measurefor suffosion than axial deformation, and it is apparent that the onset of suffosion based onmeasurements of axial deformation should be considered an upper limit for the conditions atthe onset of suffosion. Re-interpretation of the tests on FR7 and FR8 (see Fig. 6.18) indicatesthat the onset of suffosion occurred at differential pore water pressures ranging from ∆uso =0.0 to 3.0 kPa in all but test FR7-100-D, which exhibited ∆uso = 6.8 kPa. These values fallin a similar range as the values of ∆uso = 0.7 kPa to 4.5 kPa reported for tests on glass beads158in this study (compare Figs. 6.17 and 6.18), but the re-interpreted data do not yield a cleardependence of ∆uso on p′so. Further, in four of five tests on FR7 and in one of four tests onFR8 (see Fig. 6.18), the onset of suffosion appears to have initiated at lower differential porewater pressures than the onset of failure, yielding a similar overall response of suffosion priorto failure as in the two tests on gradation 6.5GB35 of this study (see Fig. 6.16). Hence, theapparent range of values of ∆uso, and the observed progression from suffosion to failure in testson the glass bead gradations FR7 and FR8 reported by Li (2008), are generally similar to theresponse to seepage flow on the gradations of glass beads tested in this study. The findings fromthe literature are however inconclusive regarding the influence of the mean effective stress onthe onset of suffosion.6.4 Test of hypothesis No. 3The third hypothesis was defined with the purpose of establishing a causative relation betweenthe micro-structure of a particle arrangement and its potential for seepage-induced internalinstability: The micro-structure of internally unstable materials controls the phenomenologicalresponse to seepage flow. The hypothesis is first tested based on the glass beads tests of thisstudy in Section 6.4.1. The findings are subsequently compared to select glass beads testsreported in the literature in Section 6.4.2.6.4.1 Glass beads tests of this studyInspection of the responses on gradations of glass beads tested in this study with varying S f andvarying D′15/d′85 (see Fig. 6.19) demonstrates that glass beads gradations with S f = 0.20 andD′15/d′85 > 4 exhibited suffusion, whereas a suffosive response was only identified in gradationswith 0.25≤ S f ≤ 0.35 and D′15/d′85 > 4. In the following detailed investigation of the influenceof the micro-structure, a distinction is made between micro-structures that exhibited a suffusiveresponse (see Section 6.4.1.1), and micro-structures that exhibited a suffosive response (seeSection 6.4.1.2).6.4.1.1 Micro-structures that exhibited a suffusive responseThe micro-structure of specimens of two glass beads gradations that exhibited a suffusive re-sponse are both type C-T, with, as noted previously, S f = 0.20 (see Fig. 6.20). From the analysisof the micro-structures of gap-graded materials (see Section 5.2), it was postulated that a por-tion of the fine particles in the C-T type micro-structure is load bearing, based on the relativelyloose state of the coarse fraction es > es,max. This postulate can be examined by assessingthe micro-structures of the gap-graded gradations of perfectly spherical particles simulated byShire and O‘Sullivan (2013) using DEM. It is unknown to what extent the method to configurethe particle arrangement, and hence its micro-structure, adapted by Shire and O‘Sullivan (2013)159approximates the physical specimen preparation by slurry deposition. The micro-structure ofthe gradations simulated by Shire and O‘Sullivan (2013), which are included herein for thepurpose of comparison, can be assessed using the micro-structure identification diagram thathas been developed for the glass beads (see Fig. 6.20 and Table 6.3): gradations G1-10 andG1-20 exhibit a type C micro-structure, gradations G1-30 and G2-20 exhibit a type C-T micro-structure, and gradation G1-40 exhibits a type M-T micro-structure. For these gradations, Shireand O‘Sullivan (2013) reported the portion of load bearing fine particles b (see Table 6.3). Inthe type C micro-structure of gradations G1-10 and G1-20, the portion of load bearing fineparticles is b = 0.01 and 0.02, respectively, whereas in type C-T micro-structure of gradationsG1-30 and G2-20, it is 0.24 and 0.12, respectively. In the type M-T micro-structure of gradationG1-40, 0.61 of the fine particles are non-load bearing. The very small values of b = 0.01 to 0.02in gradations with a type C micro-structure, confirm the postulate that the fine particles do notcontribute significantly to the stability of this type of micro-structure. In contrast, the findingsof DEM simulations indicate that a substantial portion of the fine particles of b≥ 0.12 is indeedload bearing in type C-T micro-structures. The findings further suggest that nearly half of thefine particles may be non-load bearing in type M-T.Given this insight from DEM simulations, a novel procedure is proposed to determine theportion of load-bearing fine particles in micro-structures of gap-graded gradations: it is basedon the concept of an equivalent inter-coarse void ratio ese (see Eq. 2.7). Thevanayagam et al.(2002) hypothesised that the shear response of a specimen with a type C micro-structure witha value es, was similar to the shear response of an equivalent specimen with a type C-T micro-structure with an equal value of ese. It is postulated that a type C-T micro-structure can only bestable if the volume of all load-bearing coarse and fine spherical particles is equal to the volumeof all coarse particles in a particle packing without fine particles, in the loosest stable state; thatis ese = es,max. The portion of load bearing fine particles b then follows from Eq. 2.7, if ese =es,max:b = 1− 1S fes,max− ees,max +1(6.2)with 0 ≤ b ≤ 1.Values for b of the four gradations of spherical particles with type C or type C-T micro-structures simulated by Shire and O‘Sullivan (2013), are calculated using Eq. 6.2 with es,max =0.67 (see Table 6.3). The calculated value of the portion of load bearing fine particles of b = 0in type C micro-structures of gradations C1-10 and C1-20, respectively, are in very good agree-ment with the findings of the DEM simulations of (1-b) = 0.01 and (1-b) = 0.02, respectively.Of the two gradations that classify as micro-structure type C-T, namely G1-30 and G2-20, the160calculated portion of load bearing fine particles using Eq. 6.2 yield nearly identical values, b= 0.21 and 0.09, respectively, as the values of b = 0.21 and 0.12 determined using DEM simu-lations. Although Eq. 6.2 was developed based on the concept of a type C-T micro-structure,the calculated value of b = 0.48 for the G1-40 specimen with micro-structure type M-T is com-parable to that of b = 0.61 determined by DEM (see Table 6.3). Accordingly, Eq. 6.2 appearsto provide a very good estimate of the portion of load bearing fine particles in type C and typeC-T micro-structures, and a reasonable estimate of the portion of load bearing fine particles intype M-T micro-structures.In absence of DEM simulations of the gradations tested in this study, the portion of non-load bearing fine particles (1-b) in the specimens of glass beads is calculated using Eq. 6.2 (seeTable 6.2): the portion of non-load bearing fine particles in specimens that exhibited suffusion,which were all of a type C-T micro-structure, varies from (1-b) = 0.56 - 0.81. The potentialfor non-load bearing fine particles is thus fulfilled in C-T micro-structures that exhibited suffu-sion in this study. Inspection of the variation of the relative increase in hydraulic conductivity,kmax/ki, with the portion of non-load bearing fine particles (1-b) (see Fig. 6.21), in four testson gradation 4.8GB20 and three tests on gradation 6.0GB20, suggests that an increasing por-tion of non-load bearing fine particles is associated with a greater relative increase of hydraulicconductivity. The portion of non-lead bearing fine particles then appears a useful parameter toquantify the potential for particle migration in gap-graded gradations of glass beads.The potential for suffusion is not only determined by the presence of non-load bearing fineparticles, but also by the constrictions of the micro-structure, which need to be sufficiently largeto permit passage of the fine particles. Crawford-Flett (2014) demonstrated that the controllingconstriction size in transitional clast-supported micro-structures is controlled by the smallestcoarse particles. Accordingly, D′15 appears a suitable index for the controlling constrictionsize, which renders the gap ratio D′15/d′85 a suitable index for the transportation potential offine particles from a type C or type C-T supported micro-structure. A gap-ratio of D′15/d′85 = 4(Kezdi, 1979), or D′15/d′85 = 4 to 5 (Sherard, 1979) has traditionally been used as a transporta-tion criterion of fine particles through the pores of a coarse particle arrangement. The suffusiveresponse of tests on gradations 4.8GB20 and 6.0GB20 with D′15/d′85 = 4.8 and D′15/d′85 = 6.0,respectively, broadly confirm the validity of D′15/d′85 = 4 to 5 as a transportation criterion offine particles through the pores of a coarse particle arrangement.6.4.1.2 Micro-structures that exhibited a suffosive responseGradations that exhibited a suffosive response in this study yield type C-T (gradations 6.0GB25and 6.5GB25), type T (6.0GB30, 4.8GB35 and 6.0GB35) or type M-T (6.5GB35) micro-structures (see Table 6.2). The potential for particle rearrangement of a micro-structure was161qualitatively assessed as es > es,max, or e f > e f ,max in Section 5.2. In order to investigate the in-fluence of the micro-structure on the onset of suffosion, two quantitative indices are proposed tocharacterise the potential for particle rearrangement. The modified inter-coarse state parameterΨs, as a variation to the inter-coarse state parameter proposed by Thevanayagam and Mohan(2000), is defined as the difference between the inter-coarse void ratio es and the maximumindex void ratio corresponding to a stable packing of the coarse particle component es,max:Ψs = es− es,max (6.3)Similarly, the inter-fine state parameter Ψ f , is defined as the difference between the inter-fine void ratio e f and the maximum index void ratio corresponding to a stable packing of thefine particle component e f ,max:Ψ f = e f − e f ,max (6.4)The modified inter-coarse state parameter Ψs is selected as a measure of the potential forparticle rearrangement in the type C-T micro-structure (gradations 6.0GB25 and 6.5GB25) asΨs relates to the dominant particle type in the micro-structure. Similarly, for the fine particledominated micro-structure type M-T (gradation 6.5GB35) the inter-fine state parameter Ψ f isselected as a measure for the potential for particle rearrangement. Now consider the matterof assigning a state parameter to the transitional type T micro-structure. The inter-coarse voidratio of the type T micro-structure is greater than the theoretical cubic packing arrangement, es> ecub, as is the case in type M-T micro-structures, which indicates that the coarse particles arenot in contact with each other. A comparison of the modified state parameters for all gradationsof glass beads that exhibited a suffosive response (see Fig. 6.22), indicates that the values ofΨs and Ψ f for gradations with type T micro-structure (6.0GB30, 4.8GB35 and 6.0GB35) aresimilar to the values of Ψs and Ψ f of the gradations with an M-T micro-structure (6.5GB35).The inter-fine state parameter is therefore selected as a measure for the potential for particlerearrangement in transitional type T micro-structures.Following is an examination of the influence of the proposed state parameters on suffosion.On the matter of the spatial variation of the volumetric deformation in tests that exhibited asuffosive response, positive values of Ψs or Ψ f , indicate an average potential for particle re-arrangement of coarse or fine particles, respectively, throughout the specimen. It is speculatedthat small spatial variations in the micro-structure cause the spatial variation in the volumetricdeformation, with the most susceptible zones rearranging first. On the matter of the influenceof the proposed state parameters on the conditions at the onset of suffosion, the three tests ongradations 6.0GB25 and 6.5GB25 with a type C-T micro-structure yield Ψs = 0.14 to 0.17(see Table 6.2) and differential pore water pressures at the onset of suffosion ∆uso = 0.7 to1621.4 kPa (see Table 5.4). The twelve tests on gradations with type M-T and T micro-structuresgave Ψ f = 0.19 to 0.53 and ∆uso = 0.8 to 4.5 kPa. The variation of the differential pore waterpressure at the onset of suffosion with the Ψ f in type T and type M-T micro-structures (seeFig. 6.23) yields a strong linear strong correlation, with correlation coefficient rc = 0.75. In-terestingly, the variation of Ψ f in type T and type M-T micro-structures, together with Ψs intype C-T micro-structures, yields a stronger linear correlation (rc = 0.78) with the differentialpore water pressure at the onset of suffosion. Considering the relative plotting positions of theconditions at the onset of suffosion in glass beads gradations (see Fig. 6.17), it appears thatthe gradation 6.0GB30, which yields a type T micro-structure and exhibited the largest valuesof ∆uso, corresponds to the largest value of the characteristic state parameter. Furthermore,gradations 6.0GB25 and 6.5GB25, which yield a type C-T micro-structure and exhibited thelowest values of ∆uso, correspond to the smallest value of the characteristic state parameter. Itis speculated that, in a type C-T micro-structure, local rearrangement of coarse particles canonly occur if a small portion of load-bearing fine particles becomes detached, which occurs ata relatively low differential pore water pressure proportional to Ψs. In contrast, in type T, andalso in type M-T micro-structure, an increasing Ψ f is speculated to yield stronger force chainsthrough fewer fine particles, which requires relatively large differential pore water pressures toinitiate suffosion. These findings suggests that the proposed state parameters are predictors ofthe relative susceptibility to suffosion: Ψ f is a characteristic state parameter of type T and typeM-T micro-structures, and Ψs is a characteristic state parameter of a type C-T micro-structure.6.4.2 Glass beads tests of other studiesFrom the literature, the suffusive response in four tests on glass beads gradation 6.6GB22 re-ported by Crawford-Flett (2014) (see Table 6.1), for which D′15/d′85 = 6.6 and S f = 0.22, sug-gests a refinement of the upper boundary of finer fraction content for which suffusion is thepredominant phenomenological response to S f < 0.25 (see Fig. 6.19). Additionally, one teston glass beads gradation G4-C of Sail et al. (2011), with D′15/d′85 = 7.4 and S f = 0.40, exhibitedsuffosion, as did five tests on glass beads gradation FR-7, with D′15/d′85 = 7.1 and S f = 0.30,and four tests on glass beads gradation FR-8, with D′15/d′85 = 7.9 and S f = 0.30, reported by Li(2008), respectively. The findings of Li (2008) and of Sail et al. (2011) thus appear in broadagreement with the findings of this study; hence it can be reasonably claimed that a suffosiveresponse occurs for glass beads gradations with 0.25 ≤ S f ≤ 0.40 and D′15/d′85 > 4.On the matter of factors controlling the potential for suffusion, Crawford-Flett (2014) foundthat a glass beads gradation 5.2GB22 with S f = 0.22 and D′15/d′85 = 5.2, which is very similarto gradation 4.8GB20 of this study, exhibited fluidisation when subject to upward seepage flow.Considering the previous discussion on the potential for suffusion, it would appear that a valueof D′15/d′85 > 5, instead of D′15/d′85 > 4, is an appropriate index for the transportation criterion163of fine particles being able to pass through the pores of a coarse particle arrangement, especiallyconsidering the uncertainty associated with values for d′85 and D′15. Secondly, the response ofthe four tests on gradation 6.6GB22 (Crawford-Flett, 2014), which exhibited a type C-T micro-structure, can be compared to the suffusive response on gradations of glass beads tested in thisstudy. The portion of non-load bearing fine particles is calculated for each specimen (see Table6.2), yielding a range of (1-b) = 0.65 to 0.68. Comparison of the portion of non-load bearingfine particles, and the relative increase in hydraulic conductivity of kmax/ki = 1.7 to 1.9 (see Fig.6.21), appears to confirm the previous finding that the portion of non-lead bearing fine particlesis a useful parameter to quantify the potential for particle migration in gap-graded gradationsof glass beads.On the matter of factors controlling the potential for suffosion, it appears that the tests ongradations FR7 and FR8, of Li (2008), yielded a micro-structure on the boundary between typeC-T and type T (see Fig. 6.20). Similarly, gradation G4-C of Sail et al. (2011) is identified toexhibit a type T micro-structure. The response in tests on gradations FR7, FR8, and G4-C isthus in agreement with the previous finding that a suffosive can occur in type C-T, type T, or typeM-T micro-structures. Further, the values of the inter-coarse and inter-fine state parameters ofgradations FR7 and FR8 are very similar to gradation 6.0B30 (see Fig. 6.22), which exhibits thesame finer fraction content S f = 0.30. The values of the proposed state parameters of gradationG4-C follow the trend of increasing values of Ψs and decreasing values of Ψ f with increasingfiner fraction content. The range of values of ∆uso = 0.0 to 6.8 kPa inferred for five tests ongradation FR7 and four tests on gradation FR8, shows considerable overlap with the range ofvalues of ∆uso = 3.4 to 4.5 kPa of the tests on gradation 6.0GB30. Although the values of thestate parameters, and the conditions at the onset of suffusion, of the glass beads tests reportedin the literature are similar, the limited data from the literature is deemed to yield inconclusiveevidence on the belief that the state parameters are predictors of the relative susceptibility tosuffosion.6.5 Test of hypothesis No. 4The fourth hypothesis seeks to establish a causative relation between particle shape and thestability of the soil structure subject to seepage flow: Gap-graded gradations of rounded par-ticles are more susceptible to seepage-induced internal instability than identical gap-gradedgradations of sub-angular particles.6.5.1 Glass beads and soil testsA parametric study, with particle shape as one of the control variables, permits a direct com-parison between tests where the particle shape is the only variable that changes. Any difference164in the response to seepage flow can then be attributed to the particle shape. Tests with particleshape as the single control variable were made on gradations at the lower end of the percentagefiner fraction, with S f = 0.20, and on gradations at the upper end, with S f = 0.35, respectively.No two seepage tests were found in the literature on internally unstable gradations that onlydiffer in particle shape. Accordingly, hypothesis No. 4 is examined solely based on tests con-ducted in this study.6.5.1.1 Influence of particle shape on gradations with S f = 0.20Consideration is first given to the two tests on the soil gradation 5.1BT20 and the two testson the glass beads gradation 4.8GB20, which had been isotropically consolidated to 55 kPaand 155 kPa, and 54 kPa and 153 kPa, respectively (see Table 6.4). Tests 5.1BT20-50 and5.1BT20-150 only differ from tests 4.8GB20-50 and 4.8GB20-150, respectively, in the particleshape of the sub-angular grains of the BT soil and the spherical grains of the GB glass beads.All gradations were reconstituted using the modified slurry deposition technique without anyattempt to densify the specimen, which yielded a type C-T micro-structure (see Table 6.4). Theinitial condition of the soils yielded a much larger void ratio of ec = 0.63 and 0.65 for tests5.1BT20-50 and 5.1BT20-150, respectively, compared to that of ec = 0.48 and 0.44, for tests4.8GB20-50 and 4.8GB20-150, respectively. The difference in initial void ratio is attributedsolely to the particle shape, in agreement with the findings of Youd (1973) (see Section 2.2.2.2).The phenomenological response in the 4.8GB20 glass beads tests, and the 5.1BT20 soil tests, isidentified as suffusion. Suffusion initiated at ∆usu = 0.2 to 0.3 kPa in all four tests, independentof particle shape (see Table 6.4). The increase in hydraulic conductivity of kmax/ki = 1.1 to 1.6(see Table 5.6) in the soils tests is similar to the range of kmax/ki = 1.3 to 1.5 in the glass beadstests (see Table 5.3). The finding indicates that the particle shape did not significantly affect theresponse of specimens with a type C-T soil structure at S f = 0.20. Now consider the three testson the soil gradation 5.7BT20 and three tests on the glass beads gradation 6.0GB20 (see Table6.4). Again, the void ratios ec = 0.56 to 0.59 of the soil gradation were substantially greaterthan the void ratios of ec = 0.40 to 0.45 of the glass beads gradation. The phenomenologicalresponse in the 6.0GB20 glass beads tests and the 5.7BT20 soil tests is identified as suffusion,with ∆usu = 0.1 to 0.4 kPa (see Table 6.4) and kmax/ki= 1.0 to 1.7 (compare Tables 5.3 and 5.6),independent of particle shape. This comparison further supports the finding that the particleshape does not significantly affect the response of the materials with a type C-T micro-structureat S f = 0.20.An explanation for the nearly identical response to seepage flow of soils and glass beadswith S f = 0.20 is sought by further examining the factors controlling suffusion (Crawford-Flett,2014): (i) the presence of non-load bearing fine particles; (ii) the hydraulic load; and (iii) thecontrolling constriction size of the material. The postulate of a presence of non-load bearing165fine particles in a type C-T micro-structure was confirmed for the glass beads specimens inSection 6.4.1.1 and it can be reasonably extended to soil specimens. On the matter of thehydraulic load, Crawford-Flett (2014) convincingly argued that, under conditions of upwardseepage flow, movement of a non-load bearing particle initiates when the seepage-induced dragforce exceeds the buoyant weight of the particle. The fine particles of glass beads with d′85= 0.19 mm and soils with d′85 = 0.17 mm (see Table 3.1) are nearly identical in size which,in addition to the very similar values of specific gravity of the solids of Gs = 2.49 and 2.70,respectively, implies that the buoyant weight of individual glass beads and sub-angular particlesis approximately the same. Although the drag force depends on particle shape, the difference indrag coefficient of the glass beads (with sphericity SQP = 0.94) and the sub-angular soil particles(with SQP = 0.88) is negligible for flow regimes of this study with Re < 10, taking into accountthe observations of Haider and Levenspiel (1989). The hydraulic threshold for migration ofthe finer fraction of glass beads and the finer fraction of sub-angular soil particles can thusbe reasonably anticipated to be of similar value, which is in agreement with the experimentalresults of this study. Given the comparable response to seepage flow of the glass bead and soiltest specimens at S f = 0.20, it is postulated that the constrictions of the glass beads and soilsare of a similar size in these specimens, which exhibit a type C-T micro-structure.6.5.1.2 Influence of particle shape on gradations with S f = 0.35The effect of the particle shape at S f = 0.35 can be appreciated by comparing the responseto seepage flow in tests on glass beads gradations 6.0GB35 and 6.5GB35, with tests on soilgradations 5.7BT35 and 7.0BT35, respectively. All gradations were reconstituted using themodified slurry deposition technique, without any attempt to densify the specimen. The voidratio ec = 0.46 of gradation 5.7BT35 (see Table 6.4) was substantially greater than the voidratio of ec = 0.37 of gradation 6.0GB35. The response to seepage flow is identified as suffosionin the glass beads test 6.0GB35-100 (see Table 6.4), whereas soil test 5.7BT35-100 reacheda “pre-critical” condition, despite a much greater differential pore water pressure across thespecimen of ∆u = 8.6 kPa at the end of test 5.7BT35-100, compared to ∆u = 4.7 kPa at the endof test 6.0GB35-100. Both specimens exhibited a type T micro-structure (see Table 6.4). Thecomparison of tests 6.5GB35-50, which exhibited a type M-T micro-structure, and 7.0BT35-50,which exhibited a type T micro-structure, appears to indicate a similar influence of the particleshape. Again, the void ratio ec = 0.42 of gradation 7.0BT35 (see Table 6.4) was substantiallygreater than the void ratio of ec = 0.30 of gradation 6.5GB35. The glass beads test exhibited asuffosive response, which eventually resulted in failure at ∆u f = 1.4 kPa, whereas the soil testreached a “pre-critical” conditions even at an end-of-test value of ∆u = 9.5 kPa. Inspection ofthe response to seepage flow of other soil specimens with S f = 0.35 (see Table 6.4) indicatesthat the gap ratio needs to be increased to D′15/d′85 = 8.6 for a suffusive response to occur(test 8.6BT35-50), and to D′15/d′85 = 10.4 for a suffosive response to occur (tests 10.4BT35-16650, 10.4BT35-50(R) and 10.4BT35-100). It can thus reasonably be concluded that specimensof sub-angular particles with S f = 0.35 yield stronger type T or M-T micro-structures thanspecimens with a nearly identical gradation of spherical particles. In these micro-structures,both the fine and coarse particles exist at relatively loose states and hence, large portions ofthe fine and coarse particles are presumed load-bearing and contributing to the stability of themicro-structure. Collapse of a type T or M-T micro-structure implies rearrangement of particlesand hence the greater resistance against particle rotation, and especially contact slipping, of sub-angular particles (Holtz et al., 2011; see also Section 2.2.2.2) is postulated to yield a strongertransitional micro-structure of sub-angular particles than of spherical particles.6.6 Factors governing suffusion and suffosionThe influence of several factors governing suffusion and suffosion has been established in thepreceding using glass beads tests of this study, and supplemented by limited evidence found inthe literature of glass beads tests. What follows is a general discussion on the factors governingsuffusion and suffosion. The findings from the the glass beads tests are compared to the soiltests from this study and the literature, and where necessary expanded upon.6.6.1 Volume changeIn this study, the response to seepage flow in seven tests on two glass beads gradations (see Ta-ble 5.4), and in eleven tests on seven soil gradations was identified as suffusion (see Table 5.7).The volumetric deformation was negligible to small in ten soil tests, ranging from end-of-testvalues of εv = 0.06 to 0.12 % (see Table 5.4). In one test, 10.4BT25-50, volumetric deformationoccurred as an isolated, non-progressive event, yielding an end-of-test value εv = 0.72 %, butthe predominant response was one of local migration of fine particles, largely in the absence ofvolume change. Accordingly, the broader definition of suffusion as seepage-induced mass losswithout change in volume, or else a small non-progressive change in volume, accompanied byan increase of hydraulic conductivity, appears a useful, and necessary extension of the defini-tion of suffusion established from the review of the literature (see Section 2.1.3).In addition to suffusion, a suffosive response was identifed in 15 tests on six different glassbeads gradations (see Table 5.4), and in three tests on one soil gradation (see Table 5.7). Suffo-sion was characterised by an initial sequence of no volumetric strain and a subsequent sequenceof a progressive development of substantial volumetric strain, in both the glass beads and soiltests. The end-of-test values εv = 0.78 to 2.46 % in the glass beads tests (see Table 5.4) andεv = 1.10 to 2.00 % in the soil tests (see Table 5.7) yield comparable values. The average unitrate of deformation in the suffosive response in tests on gradation 10.4BT35 varied within arelatively narrow range of Ev = 0.19 to 0.27 %/kPa (see Table 6.4), which broadly confirms the167prior finding that the average unit rate of deformation is a useful variable to quantify volumetricdeformation associated with suffosion, independent of particle shape. The three tests on grada-tion 10.4BT35 exhibited substantial volumetric deformations in the absence of substantial axialdeformations (see Figs. 6.24), which enhances the finding that measurement of total volumechange is necessary to avoid any mis-interpretation of the phenomenological response to seep-age flow.From the literature, Chang and Zhang (2013), found that axial and radial strains enablea useful distinction between an “initiation phase”, which is believed equivalent to suffusionin this study, and a “development phase”, which is believed equivalent to suffosion in thisstudy. Secondly, Moffat et al. (2011), testing soil T-0 in a rigid wall permeameter, observed that“[s]mall movements occurred [..] prior to the development of a relatively large displacementduring the final stage [..]”. Hence, it appears that the rearrangement of the particle packingof the soil was not recorded in the measurement of axial strain, an issue which Li (2008), asnoted in Section 6.2.2, also alluded to in the tests on glass beads. These observations from theliterature further confirm the need to measure volume change in order to correctly characterisesuffosion.6.6.2 Effective stressThe independence of the effective stress on the onset of suffusion in gap-graded glass beadswas conclusively established by Crawford-Flett (2014). The prior comparison between tests onglass beads gradations and soil gradations with S f = 0.20 (see Section 6.5.1.1), established thatthe particle shape does not have a significant effect on the response of these gradations, whichexhibit a type C-T micro-structure. Inspection of the variation (see Fig. 6.25) of the mean effec-tive stress p′su and the differential pore water across the specimen at the onset of suffusion ∆usu,in tests on glass beads and soils, reveals that in all but two soil tests, a small value ∆usu = 0.1to 0.4 kPa is independent of p′su. The gradations which exhibit suffusion at these small valuesof ∆usu, namely 4.8GB20, 6.0GB20, 5.1BT20, 5.7BT20, 7.0BT20, 8.6BT20 and 10.4BT25,exhibit a type C-T micro-structure (see Tables 5.2 and 5.5). This observations thus confirms thefinding of Crawford-Flett (2014) that the onset suffusion is not dependent on effective stress.However, one test on gradation 8.6BT35 (see Table 6.4), with a type T micro-structure, and onetest on gradation 10.4BT30, with a type C-T micro-structure, exhibited a suffusive responsethat initiated at relatively large values of ∆usu = 5.3 and 2.0 kPa, respectively. From the liter-ature, Chang and Zhang (2013) also appear to have reported a suffusive response on a similargradation GS, with S f = 0.35, which exhibited relatively large values of ∆usu = 0.8 to 2.5 kPa.These limited data indicate that suffusion can be dependent on effective stress in gradationswith S f > 0.25, which is tentatively attributed to the characteristics of the micro-structure forthis range of finer fraction contents. Notwithstanding this caveat, the findings of the suffusive168response in tests on glass beads and soils, support the conclusion of Crawford-Flett (2014) thatthe onset of suffusion is not dependent on effective stress, and extend it to the domain of thesoils. However, it is noteworthy that this holds only true in specimens that exhibit a type C-Tmicro-structures with S f ≤ 0.25.The examination of the conditions at the onset of suffosion in 15 tests on six glass beadsgradations indicated a trend of increasing ∆uso with increasing p′so at a gentle slope. In compar-ison, the onset of suffosion in soil tests 10.4BT35-50, 10.4BT35-50(R) and 10.4BT35-100 (seeFig. 6.26) initiates at values of ∆uso = 0.8 to 1.8 kPa and p′so = 54 to 102 kPa. The data of thesesoils test are deemed insufficient to further examine the influence of the effective stress on theonset of suffosion.From the literature, Moffat (2005) identified a suffosive response in three tests on soil gra-dation T-0 and in five tests on soil gradation T-5. Re-interpretation of the response, based on thevariation of the axial strain, of the tests on gradation T-0 (see Fig. 6.27) indicates that suffosioninitiated prior to failure in all three tests, where failure is defined as continuing deformations ata constant hydraulic gradient. In the tests, the hydraulic gradients at the onset of axial deforma-tion are relatively small, and yield an apparent increase in hydraulic gradient with increasingeffective stress with a gentle slope. Similarly, in two of five tests on gradation T-5 (see Fig.6.28), suffosion was identified to initiate at relatively small hydraulic gradients, prior to failureof the specimen. Chang and Zhang (2013) reported a suffosive response in 18 tests on a soilgradation GS, which was anisotropically consolidated to varying states of effective stress. Thereported values of ∆uso = 0.9 to 3.8 kPa in these tests (see Fig. 6.29) plot in a similar range asthe values of ∆uso in the glass beads and soil tests (see Figs. 6.17 and 6.26). More importantly,the data exhibit a trend of increasing ∆uso with increasing p′so with a gentle non-linear slope:the trend line can reasonably be extrapolated through the origin. The tests of Moffat (2005), andin particular Chang and Zhang (2013), thus appear consistent with the finding of this study thatthe onset of suffosion is governed by effective stress: the variation of differential pore waterpressure at the onset of suffosion with the mean effective stress, exhibits a gentle positive slope.6.6.3 Micro-structureA suffusive response has been identified in tests on glass beads gradations with D′15/d′85 > 5and S f < 0.25, whereas a suffosive response was identified in tests on glass beads gradationswith similar gap-ratios D′15/d′85 > 5, but greater finer fraction contents 0.25 ≤ S f ≤ 0.40 (seeFig. 6.19). For the soils tested in this study, a suffusive response was identified in tests onseven gradations with 0.20 ≤ S f ≤ 0.35 and D′15/d′85 > 5, whereas a suffosive response wasidentified in tests on one gradation with S f = 10.4 and S f = 0.35. Inclusion of the suffusiveresponse on a gradation A of Skempton and Brogan (1994), the suffusive or suffosive response169on a gradation GS of Chang and Zhang (2013), and the suffosive response on gradations T-0and T-5 of Moffat (2005), allows for the limits for seepage-induced internal instability in soilsto be refined (see Fig. 6.30) as follows: a suffusive response in soil gradations with 0.15 ≤S f ≤ 0.35 and D′15/d′85 > 5; and, a suffosive response in soil gradations with 0.30 ≤ S f ≤0.40 and D′15/d′85 > 8. Accordingly, gradations with D′15/d′85 > 5 and with a relatively smallfiner fraction content S f ≤ 0.25, are susceptible to suffusion, independent of particle shape. Incontrast, the particle shape has a profound influence on the susceptibility of gradations with0.30 ≤ S f ≤ 0.40: glass beads gradations are susceptible to suffosion if D′15/d′85 > 5, whereassoil gradations are susceptible to suffosion if D′15/d′85 > 8.6.6.3.1 Influence of micro-structure on suffusionFrom the glass beads tests, the portion of non-load bearing fine particles was found a usefulparameter to quantify the potential for migration of fine particles in type C-T micro-structures.Given that the response for gradations with S f = 0.20, which yielded a type C-T micro-structure,is not affected by the particle shape, it is reasonable to claim that the proposed procedure todetermine the portion of non-load bearing fine particles (i.e. Eq. 6.2) is also valid for soil gra-dations with a type C-T micro-structure. Six of seven soil gradations that exhibited a suffosiveresponse yield a type C-T micro-structure (gradations 5.1BT20, 5.7BT20, 7.0BT20, 8.6BT20,10.4BT25 and 10.4BT30, see Table 6.4). The range of the calculated values of (1-b) = 0.42to 0.81 (see Table 6.4) for the soil gradations, together with the range of increasing hydraulicconductivity kmax/ki = 1.0 to 1.8 (see Table 5.6), are similar to the glass beads specimens (seeFig. 6.31). This observation confirms the previous finding that the portion of non-lead bearingfine particles is a useful parameter to quantify the potential for migration of fine particles fortype C-T micro-structures, independent of particle shape.On the matter of suffusion in gradations 8.6BT35 and 10.4BT30, which exhibited type Tand C-T micro-structures, respectively, it seems reasonable to claim that a substantial portionof the fine particles is non-load bearing in these micro-structures in soils, considering it wasestablished for glass beads gradations in Section 6.4.1.1. Considering the substantial potentialfor particle rearrangement in a type T micro-structure, it is further speculated that gradation8.6BT35 may exhibit suffosion when subject to more adverse hydraulic loads. The findings ofChang and Zhang (2013) established that certain soils can exhibit either suffusion or suffosion,depending on the stress condition. In contrast, considering the suffusive response in soils withS f ≤ 0.25 and D′15/d′85 > 5, which exhibit a C-T micro-structure, and the limited potential forparticle rearrangement in this micro-structure (see Section 6.4.1.1), it is postulated that thesegradations are only susceptible to suffusion.1706.6.3.2 Influence of micro-structure on suffosionThe proposed state parameters appear to be predictors of the relative susceptibility to suffosion:Ψ f is a characteristic state parameter of type T and type M-T micro-structures and Ψs is acharacteristic state parameter of a type C-T micro-structure. Calculation of the modified inter-fine state parameter of the soil specimens of gradation 10.4BT35, using Eq. 6.4 with es,max= 0.80, yields very similar values of Ψ f = 0.23 to 0.29 (see Table 6.4), compared to Ψ f =0.20 to 0.40 (see Table 6.2) of the glass beads specimens with S f = 0.35 (see also Fig. 6.32).The tests on gradation 10.4BT35 yielded a suffosive response with ∆uso = 0.8 to 1.8 kPa (seeTable 5.7 and Fig. 6.26), which is of very similar range to the values of ∆uso = 0.8 to 2.8 kPaobtained from the tests on glass beads gradations 4.8GB35, 6.0GB35 and 6.5GB35 (see Table5.7), despite the difference in particle shape and gap ratio. The similar values of the differentialpore water pressures required to initiate suffosion, implies that the glass beads specimens, andsoil specimens, exhibit similar micro-structures with similar force chains. The combination ofvalues of Ψ f and ∆uso for the soil gradation plot in the same range of glass beads gradations(see Fig. 6.33), which tentatively confirms that the proposed state parameters are predictors ofthe relative susceptibility to suffosion, independent of particle shape.6.7 A unified approach for suffosionA suffosive response was characterised in this study (see Section 5.5) by the sequence of aninitially constant hydraulic conductivity in the absence of volumetric deformation, and a sub-sequent varying hydraulic conductivity at differential pore water pressures across the specimengreater than a threshold value, accompanied by the progressive development of non-uniformcontractive volumetric deformations. In some tests, the onset of suffosion was followed bycontinuing deformations in the last stage associated with failure. The prior discussion has es-tablished that volume change is a necessary variable to correctly quantify suffosion and that theonset of suffosion occurs at a critical combination of mean effective stress p′ and differentialpore water pressure across the specimen ∆u. Considering the variables necessary to quantifya suffosive response have been identified, an opportunity exists to develop a unified approachto characterise suffosion (see Fig. 6.34). An element is consolidated to p′c and subsequentlysubject to an increasing differential pore water pressure across the specimen ∆u, yielding up-ward seepage flow. In Fig. 6.34 a typical hydro-mechanical path is depicted for a specimensubject to upward seepage flow in a flexible wall permeameter, with a decreasing value p′ forincreasing values of ∆u. No volume change occurs if the hydro-mechanical path remains belowthe suffosion boundary. Suffosion initiates if the hydro-mechanical path reaches the suffosionboundary at point (p′so, ∆uso). Suffosion progresses with increasing ∆u, at an approximatelyconstant unit rate of deformation Ev. Failure, defined as continuing volumetric deformations ata constant differential pore water pressure across the specimen, occurs if the hydro-mechanical171path reaches the failure boundary at point (p′f , ∆u f ).The proposed state parameters appear good predictors of the conditions at the onset ofsuffosion (see Fig. 6.33): the modified inter-coarse state parameter Ψs is characteristic for typeC-T micro-structure, whereas the inter-fine state parameter Ψ f is characteristic for type T andtype M-T micro-structures. The average unit rate of deformation Ev appears mainly dependenton the particle shape and D′15/d′85: Ev was found to increase with increasing values of D′15/d′85for glass beads gradations (see Fig. 6.4) .6.8 SummaryFactors controlling suffusion and suffosion have been discussed in this Chapter, guided by theexamination of four research hypotheses using the tests of this study and select tests reportedin the literature (see Fig. 6.1). Based on seven tests on two glass beads gradations and eleventests on seven soil gradation of this study, and supplemented by experimental observations inthe literature, it is proposed to broaden the definition of suffusion to “a seepage-induced massloss without change in volume, or a with small, non-progressive contractive volume change,accompanied by an increase in hydraulic conductivity”. The suitability of volume change asa variable to characterise seepage-induced internal instability was tested in the discussion ofhypothesis No. 1. The average unit rate of deformation Ev was introduced as a characteristicvariable of the progression of suffosion and it was deemed useful to quantify seepage-inducedvolumetric deformation in 15 tests on six glass beads gradations and in three tests on one soilgradation. Considering the localised nature of the progressive contractive volumetric deforma-tion associated with suffosion, which can occur in absence of axial deformation, measurementof total volume change proved necessary to avoid any mis-interpretation of the phenomenolog-ical response to seepage flow.The influence of effective stress on suffosion was tested in the discussion of hypothesis No.2. The differential pore water pressure at the onset of suffosion was found to increase with in-creasing mean effective stress in 15 tests on six glass beads gradations that exhibited suffosionin this study, and in 18 tests on one soil gradation reported in the literature. It was confirmed forglass beads and soil gradations, that the onset of suffusion is not dependent on effective stressin specimens with S f ≤ 0.25, which exhibit a transitional clast-supported micro-structure.In the discussion of hypothesis No. 3, the dependence of the micro-structure on suffusionand suffosion was investigated. Based on the results of tests reported in this study and testsreported in the literature, a suffusive response was identified in tests on glass beads gradationswith D′15/d′85 > 5 and S f < 0.25, whereas it was identified in tests on soil gradations with 0.15≤ S f ≤ 0.35 and D′15/d′85 > 5. The micro-structures of seven glass beads specimens and of ten172soil specimens of this study that exhibited a suffusive response are transitional clast-supported;one soil specimen yielded a transitional micro-structure. Similarly, a suffosive response wasidentified in tests on glass beads gradations with D′15/d′85 > 5 and 0.25 ≤ S f ≤ 0.40. Theparticle shape has a profound influence on the susceptibility to suffosion: suffosion was identi-fied in soil gradations with 0.30 ≤ S f ≤ 0.40 and D′15/d′85 > 8. Micro-structures of gradationsthat exhibited a suffosive response are transitional clast-supported, transitional, or transitionalmatrix supported. It appears that specimens with a transitional micro-structure may exhibitsuffusion, prior to suffosion. A novel procedure to determine the portion of load-bearing fineparticles in clast-supported and transitional clast-supported micro-structures was validated withDEM simulations. The findings inferred from seven tests on two glass beads gradations andfrom ten tests on six soil gradations that exhibited a suffusive response and a transitional clast-supported micro-structure in this study, supported by evidence found in the literature, indicatethat the portion of non-load bearing fine particles is a useful parameter to quantify the poten-tial for particle migration, independent of particle shape. The novel concepts of the modifiedinter-coarse state parameterΨs and the inter-fine state parameterΨ f have been introduced. Thefindings inferred from 15 tests on six glass beads gradations and from three tests on soil grada-tions suggests that the proposed state parameters are predictors of the relative susceptibility tosuffosion, independent of particle shape: Ψ f is a characteristic state parameter of transitionaland transitional matrix-supported micro-structures and Ψs is a characteristic state parameter ofa transitional clast-supported micro-structure.Discussion of hypothesis No. 4 examined the influence of the particle shape on seepage-induced internal instability. The comparison between five glass beads tests and five soils testsestablished that the particle shape does not significantly affect the response of gap-graded gra-dations with S f = 0.20, It was postulated that the constrictions of the glass beads and soils areof a similar size in these specimens. The comparison between two glass beads tests and twosoils tests demonstrated that sub-angular particles result in transitional micro-structure at S f= 0.35 that is more resistant to collapse than a similar micro-structure of spherical particles,which is attributed to the greater resistance against particle rotation and contact slipping, of thesub-angular particles.Finally, a unified approach is presented to characterise suffosion in the domain of mean ef-fective stress, differential pore water pressure across the specimen, and volumetric deformation.The approach unifies the notion of a hydro-mechanical boundary at the onset of suffosion, anda hydro-mechanical boundary where seepage-induced failure of internally unstable gradationsinitiates. Failure is is defined as continuing deformations at a constant differential pore waterpressure across the specimen.173Table 6.1: Experimental database on suffusion and suffosion compiled from the literature.Study Apparatus1 Flow2 No. of tests Material3 Gradation S f D′15/d′85 R4 Pheno-direction (-) type (-) (-) (-) menon5Crawford- RWP U 4 GB 6.6GB22 0.22 6.0 R SU(2014)Li RWP U/D 5 GB FR7 0.30 7.1 R SO(2008) D 4 GB FR8 0.30 7.9 R SOSail et. al (2011) RWP D 1 GB G4-C 0.40 7.4 R SOSkempton and RWP U 1 Soil A 0.15 11 SA SUBrogan (1994)Chang and FWP D 22 Soil GS 0.35 7.9 SA6 SU/SOZhang (2013)Moffat RWP D 3 Soil T-0 0.40 13.7 SA SO(2005) U/D 5 Soil T-5 0.40 14.3 SA SONotes:1 RWP = Rigid Wall Permeameter, FWP = Flexible Wall Permeameter2 U = Upward seepage flow, D = Downward seepage flow3 Soil or Glass Beads (GB)4 roundness, R; R = Rounded, SA = sub-angular5 SU = Suffusion; SO = Suffosion6 The particle shape of the coarse component was identified by the author, based on Fig. 3-4.(a) of Chang (2012)174Table 6.2: Seepage-induced internal instability in tests on glass beads.Study Test code εa1 εv1 Pheno- Ev3 Micro- (1-b)5 Ψs, Ψ f 6(%) (%) menon2 (%/kPa) structure4 (-) (-)This 4.8GB20-50 0.00 0.12 SU - C-T 0.56 0.19study 4.8GB20-50(2) -0.01 0.03 SU - C-T 0.67 0.144.8GB20-100 0.02 0.50 SU - C-T 0.60 0.174.8GB20-150 0.04 0.48 SU - C-T 0.69 0.136.0GB20-50 0.00 0.09 SU - C-T 0.66 0.146.0GB20-100 0.30 0.82 SU - C-T 0.78 0.096.0GB20-150 0.01 0.09 SU - C-T 0.81 0.06.0GB25-50 0.68 1.61 SO 0.28 C-T 0.69 0.176.0GB25-150 0.05 1.58 SO 0.23 C-T 0.74 0.146.5GB25-100 0.08 1.65 SO 3.18 C-T 0.75 0.144.8GB35-50 1.15 1.60 SO 0.21 T - 0.364.8GB35-100 0.79 1.20 SO 0.20 T - 0.404.8GB35-150 0.05 0.78 SO 0.12 T - 0.396.0GB30-50 0.44 1.90 SO 0.71 T - 0.526.0GB30-100 0.21 1.78 SO 0.44 T - 0.466.0GB30-150 0.04 1.87 SO 0.62 T - 0.486.0GB35-50 0.21 2.19 SO 0.43 T - 0.276.0GB35-100 0.62 1.54 SO 0.43 T - 0.386.0GB35-100(2) 0.70 2.19 SO 0.50 T - 0.386.0GB35-150 0.02 2.46 SO 0.55 T - 0.306.5GB35-50 0.14 1.89 SO 0.75 M-T - 0.206.5GB35-100 0.04 1.68 SO 0.89 M-T - 0.20Crawford- 6.6GB22-0 0.3 - SU - C-T 0.65 0.16Flett 6.6GB22-25 0.1 - SU - C-T 0.68 0.15(2014) 6.6GB22-50 0.2 - SU - C-T 0.68 0.156.6GB22-100 0.6 - SU - C-T 0.68 0.15Li FR7-25-D 3.4 - SO - T7 - 0.46(2008) FR7-50-D 0.1 - SO - T7 - 0.43FR7-100-D 1.6 - SO - T7 - 0.40FR7-150-D 0.6 - SO - T7 - 0.40FR7-150-U 3.0 - SO - T7 - 0.40FR8-25-D2 1.6 - SO - T - 0.56FR8-50-D 0.7 - SO - T - 0.56FR8-100-D 4.7 - SO - T - 0.53FR8-200-D 3.9 - SO - T - 0.43Sail et al. (2011) G4-C 4.9 - SO - T - 0.23Note:1 At the end of test, from Table 5.4 or from literature.2 SU = Suffusion; SO = Suffosion, from Table 5.4, or from literature.3 For suffosive responses, using Eq. 6.1.4 See Table 5.2 and Fig. 6.20.5 Using Eq. 6.2.6 Characteristic state parameter Ψs, using Eq. 6.3, for type C-T micro-structure and Ψ f , using Eq. 6.4, fortype T and M-T micro-structures.7 See Fig. 6.20: specimens of gradation FR7 plot on the boundary between types C-T and T.175Table 6.3: Evaluation of portion of non-load bearing particles in gap-graded gradations.Gradation S f e es e f Sf,L Micro b1 b2(-) (-) (-) (-) (-) structure DEM Eq. 6.2(-) (-)G1-10 0.11 0.45 0.63 4.11 - C 0.01 0G1-20 0.21 0.32 0.67 1.54 - C 0.02 0G1-30 0.30 0.29 0.85 0.96 - C-T 0.24 0.24G1-40 0.40 0.32 1.21 0.79 0.95 M-T 0.61 0.48G2-20 0.21 0.35 0.71 1.67 - C-T 0.12 0.09Note:1 From Shire and O‘Sullivan (2013).2 Using Eq. 6.2.176Table 6.4: Comparison of select tests on glass beads and soils.R1 Test code D′15/d′852 S f 2 p′c3 ec3 Micro- ∆usu5 ∆uso5 ∆u f 5 p′su5 p′so5 p′f5 (1-b)6 Pheno-(-) (kPa) (-) (-) structure4 (kPa) (kPa) (kPa) (kPa) (kPa) (kPa) (-) menon7R 4.8GB20-50 4.8 0.20 54 0.48 C-T 0.2 - - 54 - - 0.56 SUR 4.8GB20-150 4.8 0.20 153 0.44 C-T 0.2 - - 153 - - 0.69 SUR 6.0GB20-50 6.0 0.20 54 0.45 C-T 0.2 - - 54 - - 0.66 SUR 6.0GB20-100 6.0 0.20 102 0.41 C-T 0.3 - - 102 - - 0.78 SUR 6.0GB20-150 6.0 0.20 151 0.40 C-T 0.4 - - 151 - - 0.81 SUR 6.0GB35-100 6.0 0.35 102 0.37 T - 1.2 - - 101 - 0.52 SOR 6.5GB35-50 6.5 0.35 54 0.30 M-T - 1.2 1.4 - 54 53 0.63 SOSA 5.1BT20-50 5.1 0.20 53 0.63 C-T 0.3 - - 53 - - 0.47 SUSA 5.1BT20-150 5.1 0.20 152 0.65 C-T 0.3 - - 152 - - 0.42 SUSA 5.7BT20-50 5.7 0.20 54 0.59 C-T 0.3 - - 54 - - 0.58 SUSA 5.7BT20-100 5.7 0.20 102 0.56 C-T 0.2 - - 102 - - 0.67 SUSA 5.7BT20-150 5.7 0.20 154 0.57 C-T 0.2 - - 154 - - 0.64 SUSA 5.7BT35-100 5.7 0.35 102 0.46 T - - - - - - 0.54 PCSA 7.0BT20-50 7.0 0.20 56 0.55 C-T 0.2 - - 56 - - 0.69 SUSA 7.0BT20-150 7.0 0.20 153 0.53 C-T 0.3 - - 153 - - 0.75 SUSA 7.0BT35-50 7.0 0.35 54 0.42 T - - - - - - - PCSA 8.6BT20-50 8.6 0.20 53 0.51 C-T 0.4 - - 53 - - 0.81 SUSA 8.6BT35-50 8.6 0.35 55 0.46 T 5.3 - - 52 - - - SUSA 10.4BT25-50 10.4 0.25 53 0.46 C-T 0.3 - - 53 - - 0.80 SUSA 10.4BT30-50 10.4 0.30 53 0.41 C-T 2.0 - - 52 - - 0.72 SUSA 10.4BT35-50 10.4 0.35 55 0.36 M-T - 1.8 - - 54 - 0.70 SOSA 10.4BT35-50(R) 10.4 0.35 57 0.37 M-T - 0.8 - - 56 - 0.68 SOSA 10.4BT35-50 10.4 0.35 102 0.38 M-T - 1.1 - - 102 - 0.67 SONotes:1 R = Rounded; SA = Sub-angular, from Table 3.2; 2 from Table 3.2; 3 from Tables 4.1 (GB tests) and 4.3 (BT tests); 4 from Tables 5.2 (GBtests) and 5.5 (BT tests); 5 from Tables 5.4 (GB tests) and 5.7 (BT tests); 6 using Eq. 6.2; 7 SU = Suffusion; SO = Suffosion, from Tables 5.4(GB tests) and 5.7 (BT tests).177Hypothesis No. 1 Hypothesis No. 2 Hypothesis No. 3Hypothesis No. 4Glass beads tests(this study)Glass beads tests(literature)Soil tests(this study)Soil tests(literature)Sec. 6.5Sec. 6.3 Sec. 6.4Sec. 6.2Sec. 6.6Experimental database from literatureSec. 6.2Factors controlling suffusion and suffosionFigure 6.1: Reader guide for discussion on factors controlling suffusion and suffosion.178    3DUWLFOHVL]HdPP0DVVSDVVLQJF*%)5)5*&$*677Figure 6.2: Experimental database compiled from literature.∆u∆usoεvEv1∆uf0(1)(2)(1) Onset of suffosion(2) Onset of failureFigure 6.3: Suffosion: definition of average unit rate of deformation Ev.Figure 6.4: Suffosion: variation of Ev and D′15/d′85 in suffosive responses of glass beads.179Figure 6.5: Suffosion: variation of volumetric and axial strain in tests on gradation4.8GB35.Figure 6.6: Suffosion: variation of volumetric and axial strain in tests on gradation6.0GB25.Figure 6.7: Suffosion: variation of volumetric and axial strain in tests on gradation6.0GB30.180Figure 6.8: Suffosion: variation of volumetric and axial strain in tests on gradation6.0GB35.Figure 6.9: Suffosion: variation of volumetric and axial strain in test on gradation6.5GB25.Figure 6.10: Suffosion: variation of volumetric and axial strain in tests on gradation6.5GB35.181Figure 6.11: Onset of suffosion in tests on gradation 4.8GB35.Figure 6.12: Onset of suffosion in tests on gradation 6.0GB25.Figure 6.13: Onset of suffosion in tests on gradation 6.0GB30.182Figure 6.14: Onset of suffosion in tests on gradation 6.0GB35.    0HDQHIIHFWLYHVWUHVVp′N3D'LIIHUHQWLDOSRUHZDWHUSUHVVXUHDFURVVVSHFLPHQ∆uN3D2QVHWp′c N3Dec Figure 6.15: Onset of suffosion in test on gradation 6.5GB25.Figure 6.16: Onset of suffosion and condition at failure in tests on gradation 6.5GB35.183Figure 6.17: Onset of suffosion: variation between mean effective stress and differentialpore water pressure across the specimen in tests on glass beads gradations.Figure 6.18: Suffosion: upper limit for onset of suffosion and failure in tests on glassbeads gradations FR7 and FR8 (of Li, 2008).Figure 6.19: Seepage-induced internal instability phenomena in tests on glass beads gra-dations.184Figure 6.20: Variation of micro-structure of glass beads gradations with seepage-inducedinternal instability phenomena.Figure 6.21: Suffusion: variation of relative increase in hydraulic conductivity and por-tion of non-load bearing fine particles in C-T micro-structures of glass beads.Figure 6.22: Variation of inter-coarse and inter-fine state parameters in glass beads testspecimens.185Figure 6.23: Onset of suffosion: variation of differential pore water pressure with charac-teristic state parameter in tests on glass beads gradations.Figure 6.24: Suffosion: variation of volumetric and axial strains in tests on gradation10.4BT35.Figure 6.25: Onset of suffusion in tests on glass beads and soils.186Figure 6.26: Onset of suffosion in tests on gradation 10.4BT35.Figure 6.27: Upper limit for onset of suffosion and failure in tests on soil gradation T-0 (adapted from Moffat et al., 2011, reproduced with permission from CanadianScience Publishing).187Figure 6.28: Upper limit for onset of suffosion and failure in tests on soil gradation T-5 (adapted from Moffat et al., 2011, reproduced with permission from CanadianScience Publishing).Figure 6.29: Variation between p′so and ∆uso in soil gradation GS (data extracted fromChang and Zhang, 2013).188Figure 6.30: Seepage-induced internal instability phenomena in soils.Figure 6.31: Suffusion: variation of relative increase in hydraulic conductivity and por-tion of non-load bearing fine particles in C-T micro-structures of glass beads andsoils.189Figure 6.32: Variation of inter-coarse and inter-fine state parameters in glass beads andsoil test specimens.Figure 6.33: Onset of suffosion: variation of differential pore water pressure across thespecimen with characteristic state parameter in tests on glass beads and soil grada-tions.190p'∆u p'so∆u∆usoεvEv1failure boundary suffosion boundary ∆ufp'f p'cFigure 6.34: Unified approach for characterisation of suffosion.191Chapter 7Conclusions, recommendations andimplications for practiceThe aim of this study is to establish causative relations between factors governing suffusion andsuffosion. Following a literature review of the state-of-art on seepage-induced internal instabil-ity, a new flexible wall permeameter was designed with control of effective stress and hydraulicload, and with the novel feature of measurement of total volume change. This laboratory inves-tigation comprises a parametric sensitivity study where the dependent variable is the responseof a gap-graded material, subject to isotropic consolidation to a mean effective stress rangingfrom 50 to 150 kPa and subsequently, multi-stage upward seepage flow, with head control. Thecontrol variables are the finer fraction content of the material S f , the ratio of the particle sizesof the coarse and fine fractions of the material D′15/d′85, the particle shape of the material R, themean effective stress of the specimen p′, and differential pore water pressure across the speci-men ∆u. Two commissioning tests, together with a main test program of 23 tests on eight glassbeads gradations and 16 tests on ten soil gradations were conducted. The findings of this inves-tigations are summarised in Section 7.1. The novel contributions are accentuated in Section 7.2and recommendations for future research are presented in Section 7.3. Finally, implications forpractice are discussed in Section 7.4.7.1 Conclusions7.1.1 Conclusions derived from the literature reviewA conceptual framework is developed such that a distinction can be reasonably made betweenphenomenological responses based on mass loss and volume change, which can be measureddirectly, and change in hydraulic conductivity, which can be deduced from measurement ofhydraulic gradient and flow rate. A review of the literature established three distinct phenomenaof soils subject to seepage flow:192• Migration of fine particles from a soil, termed suffusion, which is characterised as seepage-induced mass loss without change in volume, accompanied by an increase of hydraulicconductivity; and,• Collapse of the soil structure, termed suffosion, which is characterised as a seepage-induced mass loss, accompanied by a volumetric contraction and change in hydraulicconductivity; and,• Reduction of effective stress to zero in internally stable soils, termed fluidisation, whichis characterised as a seepage-induced volumetric expansion, accompanied by an increasein hydraulic conductivity, with no mass loss.7.1.2 Conclusions concerning the test methodA flexible wall permeameter comprising a double-walled triaxial cell, a seepage control system,through which unidirectional multi-stage seepage flow can be imposed, and instrumentation,has been designed and built. The following conclusions are drawn:• A novel feature in seepage-induced instability testing is the measurement of the totalvolume change, using a technique adopted from Wheeler (1986). It is derived frommonitoring the volume change of the cell fluid in the inner chamber of a double-walledtriaxial cell, that, with correction for the intrusion of the loading ram, membrane pen-etration and small calibrated volume changes of dissolving air and absorption of water,yields an accurate measure of total volume change in the test specimen.• For test specimens with a typical volume of approximately 780 cm3, the resolution of thevolume change measurement of +/- 0.1 cm3 is equivalent to εv ≈ 0.01 % and the accuracyof +/- 0.5 cm3 is equivalent to εv ≈ 0.07 %.The commissioning of the apparatus and assessment of the repeatability of the test resultsyield the following conclusions:• The resolution and accuracy of the volume change measurement technique are believedsufficient to quantify the onset and progression of deformation in specimens with a vol-ume of approximately 780 cm3.• The comparisons of three sets of companion tests indicate that the test procedure yieldsrepeatable results in test specimens of glass beads and soils with finer fraction contentsof S f = 0.20 and S f = 0.35.Additionally, the following conclusion is drawn on the method of specimen preparation:• The reconstitution of homogeneous, saturated specimens of gap-graded spherical glassbeads using the modified slurry deposition method yields test results that are repro-ducible.1937.1.3 Conclusions derived from the analysis of test resultsThe analysis of the results of the commissioning tests, and of the tests constituting the main testprogram, established the following:• A micro-structure identification diagram (Thevanayagam et al., 2002), with explicit def-initions of all boundaries, has been constructed for gap-graded gradations of glass beadsand soils, based on considerations of the experimental limits of stable particle packingarrangements of the fine fraction and coarse fraction. Analysis of the micro-structureof gap-graded materials has yielded five different types of micro-structure, in compli-ance with the original work of Thevanayagam et al. (2002): clast-supported; transitionalclast-supported; transitional; transitional matrix-supported; and matrix supported. Theability to explicitly distinguish between clast-supported and transitional clast-supportedmicro-structures, and between matrix-supported and transitional matrix-supported micro-structures yields a key improvement of the micro-structure identification diagram, be-cause of the distinction between micro-structures that exhibit potential for particle rear-rangement and micro-structures that do no exhibit this potential.• The potential for seepage-induced internal instability is qualitatively examined for eachmicro-structure type: it is inferred that clast-supported and transitional clast-supportedmicro-structures exhibit a substantial portion of non-load bearing fine particles, whereasa potential for particle rearrangement was established in clast-supported, transitionaland transitional matrix-supported micro-structures, but not in clast-supported or matrix-supported micro-structures.• Analysis of the test results has yielded conclusive evidence for the establishment of thephenomenological response in 36 of 41 tests. The following evidence was considered:thepotential for particle migration and particle rearrangement, the variation of the hydraulicconductivity, axial and volumetric strains, any mass loss inferred from the variation ofhydraulic conductivity in conjunction with volume change, and forensic observations ofmass loss.• A suffusive response, characterised by negligible or very small strains, which is in agree-ment with the strict definition established from a review of the literature, was establishedin four tests on two glass beads gradations and in ten tests on six soil gradations. Abroader definition of suffusion, associated with the non-progressive development of smallaxial or volumetric strains, was invoked to characterise the response in three tests on twoglass beads gradations and in one test on a soil gradation.• A suffosive response, associated with the progressive development of spatially variablecontractive volumetric deformations was established in 15 tests on glass beads gradations194and three tests on soil gradations, which exhibited a transitional clast-supported, transi-tional or transitional matrix-supported micro-structure. Failure, defined as continuingcontractive volumetric deformations at a constant differential pore water pressure acrossthe specimen, was identified in two glass beads tests that exhibited suffosion.• A “pre-critical” condition was reached in three tests on three glass beads gradations and intwo tests on two soil gradations; it was characterised by a constant hydraulic conductivityin the absence of axial and volumetric deformation.7.1.4 Factors governing suffusion and suffosionFactors governing suffusion and suffosion are discussed based on the evidence of glass beadsand soil tests from this study, supplemented by a limited body of evidence from six studies re-ported in the literature. The discussion of volume change as a characteristic variable of seepage-induced internal instability yields the following conclusions:• The predominant response in specimens that exhibit a suffusive response associatedwith a small, non-progressive volume change, is one of local migration of fine parti-cles, largely in the absence of volume change. Based on glass beads and soil tests inthis study, and supplemented by the findings of Crawford-Flett (2014), it is proposed tobroaden the definition of suffusion to “a seepage-induced mass loss without change involume, or a small non-progressive change in volume, accompanied by an increase ofhydraulic conductivity”.• Visual observations established that the progressive, seepage-induced contractive volumechange associated with suffosion did not occur equally throughout the specimen, but wasrather of a local nature.• The average unit rate of volumetric deformation Ev was introduced and found a useful,gradation-specific variable to quantify the volumetric deformation associated with suffo-sion. The average unit rate of volumetric deformation increases with increasing particlesize of the coarse component.• Nine glass beads tests and three soil tests that exhibited a suffosive response, experi-enced substantial volumetric deformations in the absence of substantial axial deforma-tions. Accordingly, measurement of total volume change is deemed necessary to avoidany mis-interpretation of the phenomenological response to seepage flow.Discussion of the influence of effective stress on the onset of suffusion and suffosion estab-lished the following:• The differential pore water pressure at the onset of suffosion increases with increasingmean effective stress. The findings of Chang and Zhang (2013), and a re-interpretationof tests of Moffat (2005), support this conclusion.195• The onset of suffusion is not dependent on effective stress in glass beads or soil specimensthat exhibit a transitional clast-supported micro-structure with S f ≤ 0.25.Discussion of the influence of the micro-structure on suffusion and suffosion yielded thefollowing conclusions:• A suffusive response was identified in glass beads gradations with D′15/d′85 > 5, and S f <0.25, whereas a suffosive response was identified in glass beads gradations with D′15/d′85> 5 and 0.25 ≤ S f ≤ 0.40.• A suffusive response was identified in soil gradations with 0.15≤ S f ≤ 0.35 and D′15/d′85> 5, similar to the glass beads gradations. The particle shape has a profound influenceon the susceptibility to suffosion: a suffosive response was identified in soil gradationswith 0.30 ≤ S f ≤ 0.40, and D′15/d′85 > 8.• A novel procedure is proposed to determine the portion of load-bearing fine particlesin micro-structures of gap-graded gradations. The procedure was validated using theDEM simulations of Shire and O‘Sullivan (2013), and it yielded a very good estimateof the portion of non-load bearing fine particles in clast-supported and transitional clast-supported micro-structures, and a reasonable estimate of the portion of non-load bearingfine particles in transitional matrix-supported micro-structures.• The findings indicate that: nearly all finer fraction is non-load bearing in a clast-supportedmicro-structure; a substantial portion of the finer fraction is load-bearing in a transitionalclast-supported micro-structure; and that a substantial portion of the finer fraction in atransitional matrix-supported micro-structure is non-load bearing.• The portion of non-load bearing fine particles, determined using the novel procedure,appears a useful parameter to quantify the potential for particle migration in gap-gradedgradations: a greater portion of non-load bearing fine particles is associated with a greaterincrease of hydraulic conductivity at the end-of-test.• The novel concepts of a modified inter-coarse state parameter Ψs and a inter-fine stateparameter Ψ f are introduced. The comparison of soil and glass beads tests indicates thatthe proposed state parameters are predictors of the relative susceptibility to suffosion,independent of particle shape: Ψ f is a characteristic state parameter of transitional andtransitional matrix-supported micro-structures, and Ψs is a characteristic state parameterof a transitional clast-supported micro-structure.Discussion on the influence of particle shape yields the following conclusions:• The comparison between tests on glass beads and soil gradations with S f = 0.20, estab-lished that the particle shape does not significantly affect the response of the materials196with transitional clast-supported micro-structure. It is postulated that the constrictions ofthe glass beads and soils are of a similar size in these transitional clast-supported micro-structures.• The comparison between tests on glass beads and soil gradations with S f = 0.35 demon-strates that sub-angular particles result in a transitional micro-structure that is more re-sistant to collapse than a similar micro-structure of glass beads. The greater resistanceof micro-structures of sub-angular particles, is attributed to the greater resistance againstparticle rotation, and contact slipping, of sub-angular particles.Finally, a unified approach is presented to characterise suffosion in the domain of meaneffective stress, differential pore water pressure across the specimen, and volumetric deforma-tion. The approach unifies the notion of a hydro-mechanical boundary at the onset of suffosion,and a hydro-mechanical boundary at seepage-induced failure of internally unstable gradations.7.2 Novel contributionsThe most important contributions of this investigation comprise a conceptual framework, mea-surement of total volume change in a flexible wall permeameter, and insights into the factorsgoverning suffusion and suffosion:• A conceptual framework, based on an extensive literature review, is developed such thata distinction can be reasonably made between suffusion and suffosion. Suffusion is de-fined as “a seepage-induced mass loss without change in volume, or a with small non-progressive change in volume, accompanied by an increase of hydraulic conductivity”.Suffosion is defined as “a seepage-induced mass loss accompanied by a reduction involume and change in hydraulic conductivity”.• A flexible wall permeameter comprising a double-walled triaxial cell, a seepage con-trol system; and instrumentation has been designed, built and commissioned. A novelfeature of the flexible wall permeameter, is the measurement of total volume change, us-ing a technique adopted from Wheeler (1986). Measurement of total volume change isnecessary to avoid any mis-interpretation of the phenomenological response to seepageflow. The average unit rate of volumetric deformation Ev is introduced as a variable tocharacterise suffosion.• The differential pore water pressure across the specimen at the onset of suffosion in-creases with increasing mean effective stress, which confirms the previous finding ofChang and Zhang (2013). The differential pore water pressure across the specimen at theonset of suffusion is independent of effective stress for gradations with S f ≤ 0.25, whichexpands the finding of Crawford-Flett (2014) into the domain of the soils.197• In a further advance of the concepts of Thevanayagam et al. (2002), a novel procedureis proposed to determine the portion of load-bearing fine particles in micro-structuresof gap-graded gradations. It was established that: nearly all fine particles are non-loadbearing in clast-supported micro-structures; a substantial portion of the fine particles isload-bearing in transitional clast-supported micro-structures; and a substantial portion ofthe fine particles in transitional matrix-supported micro-structures is non-load bearing.• The novel concepts of the modified inter-coarse state parameter Ψs and the inter-finestate parameter Ψ f are introduced, as another advance of the concepts of Thevanayagamet al. (2002). The state parameters are good indicators of the relative susceptibility tosuffosion.• The particle shape does not significantly affect the response of specimens with a transi-tional clast-supported type micro-structure at S f = 0.20. However, sub-angular particlesof gradations with S f = 0.35 result in a transitional micro-structure that is more resistantto collapse that a similar particle arrangement of glass beads.• A unified approach to characterise suffosion is provided.7.3 RecommendationsThe following recommendations are made to improve the flexible wall permeameter as an in-vestigative tool:• It is recommended to increase the maximum head than can be imposed, so that inherentlystable specimens can be taken to fluidisation. It is further recommended to improvecontrol of the hydro-mechanical loading path, by incorporating feedback systems in theseepage control system, the axial loading system and the cell pressure system, whichwould allow strength testing of the soil, post-seepage flow.• The range of the volume change measurement device in this study, was limited to attaina high accuracy and resolution, necessary to detect the onset of suffosion. It is recom-mended to extend the range of the volume change measurement device, whilst maintain-ing sufficient accuracy and resolution, so that the extent of failure can be investigated.The following recommendations are made concerning future studies on factors governingsuffusion and suffosion:• Given the influence of effective stress on the onset of suffosion, it is believed prudent tostudy the influence of the hydro-mechanical loading path, including anisotropic consoli-dation, on the onset of suffosion.198• This study was limited to gap-graded gradations. It is recommended to expand the inves-tigation on the factors governing suffusion and suffosion to broadly graded soils.• It is recommended to verify that the findings of this study are also valid for particlepackings prepared using different reconstitution techniques, such as moist tamping orcompaction.7.4 Implications for practiceThe findings of this investigation yield several implications for advanced laboratory testing,discrete element modelling, and dam safety management. Considering the localised nature ofthe volume change associated with suffosion, which can exhibit in the absence of axial defor-mation, measurement of the total volume change is necessary to avoid any mis-interpretationof the phenomenological response to seepage flow. Accordingly, caution is warranted whenderiving seepage-induced volumetric deformations from measurements of radial strains or pho-tographic measurements. Flexible wall permeameters, which permit independent control ofeffective stress, and hydraulic load, and measurement of total volume change, are thus recom-mended for future investigations of seepage-induced internal instability.Studies using Discrete Element Models have examined aspects of the fabric of internallyunstable materials, using spherical particles. The findings of this investigation demonstratethat the particle shape has a profound influence on the stability of the micro-structures of cer-tain gap-graded materials, which appears to govern the seepage-induced internal instabilityphenomenon. Simulations using Discrete Element Modelling should account for the effect ofparticle shape.Finally, concerning dam safety management, the distinction between the phenomena ofsuffusion and suffosion should be acknowledged in the failure mode analysis of embankmentdams and the foundation. Suffusion, which is migration of fine particles from a soil by seepageflow, in the absence of volume change, is independent of effective stress and may initiate atsmall hydraulic loads. Suffosion, which is seepage-induced collapse of the soil structure, is ofconcern in broadly graded soils, especially if the possibility of segregation exists. Suffosioninitiates at small hydraulic loads that are dependent on effective stress. Given the insights intothe micro-structures of specimens that exhibit suffusion or suffosion, it is believed prudent tore-evaluate the premise of empirical tools used to assess the susceptibility to suffusion andsuffosion.199BibliographyA˚ berg, B. (1993). “Washout of grains from filtered sand and gravel materials.” Journal ofGeotechnical Engineering Division, 119(1), 36–53. → pages 6, 17, 18, 26Adel, H., Bakker, K. J., and Klein Breteler, M. (1988). “Internal instability of minestone.”International Symposium on Modelling Soil-Water-Structure Interactions, Balkema,Rotterdam, 225–231. → pages 18, 19Altuhafi, F., O’Sullivan, C., and Cavarretta, I. (2013). “Analysis of an image based method toquantify the size and shape of sand particles.” Journal of Geotechnical andGeoenvironmental Engineering, 139(8), 1290–1307. → pages 13ASTM (2008). “Standard Guide for Reporting Uncertainty of Test Results and Use of theTerm Measurement Uncertainty in ASTM Test Methods.” Report No. E2655, ASTMInternational, West Conshohocken, Pa. → pages 208, 209ASTM (2009). “Standard Practice for Description and Identification of Soils (Visual-ManualProcedure).” Report No. D2488-09a, ASTM International, West Conshohocken, PA. →pages 13ASTM (2010). “Standard Test Methods for Measurement of Hydraulic Conductivity ofSaturated Porous Materials Using a Flexible Wall Permeameter.” Report No. D5084-10,ASTM International, West Conshohocken, PA. → pages 38ASTM (2011). “Standard Test Method for Consolidated Drained Triaxial Compression Testfor Soils.” Report No. D7181-11, ASTM International, West Conshohocken, Pa. → pages46ASTM (2014). “Standard Test Methods for Specific Gravity of Soil Solids by WaterPycnometer.” Report No. D854-14, ASTM International, West Conshohocken, PA. →pages 44, 45ASTM (2015a). “Standard Test Method for Minimum Index Density and Unit Weight of Soilsand Calculation of Relative Density.” Report No. D4254-14, ASTM International, WestConshohocken, PA. → pages 44, 45ASTM (2015b). “Standard Test Methods for Maximum Index Density and Unit Weight ofSoils Using a Vibratory Table.” Report No. D4253-14, ASTM International, WestConshohocken, PA. → pages 44, 45Bear, J. (1972). Dynamics of fluids in porous media. Elsevier, New York. → pages 14, 16, 119200Bendahmane, F., Marot, D., and Alexis, A. (2008). “Experimental parametric study ofsuffusion and backward erosion.” Journal of Geotechnical and GeoenvironmentalEngineering, 134(1), 57–67. → pages 3, 17, 18, 20, 29, 31, 40Bishop, A. and Donald, I. (1961). “The experimental study of partly saturated soils in thetriaxial apparatus.” 5th International conference on soil mechanics and foundationengineering, Paris, France, 13–21. → pages 38Bonelli, S. and Marot, D. (2011). “Micromechanical modeling of internal erosion.” EuropeanJournal of Environmental and Civil Engineering, 15(8), 1207–1224. → pages 30Burenkova, V. V. (1993). “Assessment of suffosion in noncohesive and graded soils.” Filtersin Geotechnical and Hydraulic Engineering, Balkema, Rotterdam, 357–360. → pages 2, 6,7, 19, 26Carrier, W. D. (2003). “Goodbye, Hazen; Hello, Kozeny-Carman.” Journal of Geotechnicaland Geoenvironmental Engineering, 129(11), 1054–1056. → pages 15Chang, D. (2012). “Internal erosion and overtopping erosion of earth dams and landslidedams.” Ph.D. thesis, The Hong Kong University of Science and Technology, The HongKong University of Science and Technology. → pages 174Chang, D. S. and Zhang, L. M. (2011). “A stress-controlled erosion apparatus for studyinginternal erosion in soils.” Geotechnical Testing Journal, 34(6), 579–589. → pages 17, 18,21, 31, 39, 40Chang, D. S. and Zhang, L. M. (2013). “Critical Hydraulic Gradients of Internal Erosionunder Complex Stress States.” Journal of Geotechnical and Geoenvironmental Engineering,139(9), 1454–1467. → pages 2, 3, 20, 22, 119, 153, 168, 169, 170, 195, 197Chapuis, R. P. (1992). “Similarity of internal stability-criteria for granular soils.” CanadianGeotechnical Journal, 29(4), 711–713. → pages 19Chapuis, R. P. and Aubertin, M. (2003). “On the use of the KozenyCarman equation to predictthe hydraulic conductivity of soils.” Canadian Geotechnical Journal, 40(3), 616–628. →pages 15Chapuis, R. P. and Aubertin, M. (2004). “Reply to the discussion by D . Hansen on On theuse of the Kozeny Carman equation to predict the hydraulic conductivity of soils .”Canadian Geotechnical Journal, 41(5), 994–996. → pages 15, 27Chapuis, R. P., Contant, A., and Baass, K. A. (1996). “Migration of fines in 0-20 mm crushedbase during placement, compaction, and seepage under laboratory conditions.” CanadianGeotechnical Journal, 33(1), 168–176. → pages 17, 18, 21, 27, 119Cho, G.-C., Dodds, J., and Santamarina, J. C. (2006). “Particle Shape Effects on PackingDensity, Stiffness, and Strength: Natural and Crushed Sands.” Journal of Geotechnical andGeoenvironmental Engineering, 132(5), 591–602. → pages 14Cividini, A., Bonomi, S., Vignati, G. C., and Gioda, G. (2009). “Seepage-Induced Erosion inGranular Soil and Consequent Settlements.” International Journal of Geomechanics, 9(4),187–194. → pages 17201Crawford-Flett, K. (2014). “An improved hydromechanical understanding of seepage-inducedinstability phenomena in soil.” Ph.D. thesis, The University of British Columbia, TheUniversity of British Columbia. → pages 2, 12, 13, 19, 20, 21, 22, 120, 124, 125, 153, 156,161, 163, 164, 165, 166, 168, 169, 195, 197Dybbs, A. and Edwards, R. V. (1984). “A new look at porous media fluid mechanics - Darcyto turbulent.” Fundamentals of transport phenomena in porous media, J. Bear and M.Corapcioglu, eds., Newark, Martinus Nijhoff Publishers, 199–256. → pages 14, 15Fannin, R. J. and Slangen, P. (2014). “”On the distinct phenomena of suffusion andsuffosion”.” Geotechnique Letters, 4(4), 289–294. → pages iiiFannin, R. J., Vaid, Y. P., and Shi, Y. (1994). “A Critical Evaluation of the Gradient RatioTest.” Geotechnical Testing Journal, 17(1), 35–42. → pages 46Fell, R. and Fry, J. J. (2007). “The state of the art of assessing the likelihood of internalerosion of embankment dams, water retaining structures and their foundations.” Internalerosion of dams and their foundations, R. Fell and J. J. Fry, eds., Taylor & Francis, London,1–24. → pages 1, 29Foster, M., Fell, R., and Spannagle, M. (2000). “The statistics of embankment dam failuresand accidents.” Canadian Geot, 37(5), 1000–1024. → pages 1Frydman, S., Zeitlen, J., and Alpan, I. (1973). “The Membrane Effect in Triaxial Testing ofGranular Soils.” Journal of Testing and Evaluation, 1(1), 37–41. → pages 213, 214Garner, S. J. and Fannin, R. J. (2010). “Understanding internal erosion: a decade of researchfollowing a sinkhole event.” Hydropower and Dams, 17(3), 93–98. → pages 2, 8Garner, S. J. and Sobkowicz, J. C. (2002). “Internal instability in gap-graded cores and filters.”CDA Annual Conference 2002, Victoria, BC, Canada, 6–10. → pages 6, 7, 18, 28, 119Haider, A. and Levenspiel, O. (1989). “Drag coefficient and terminal velocity of spherical andnonspherical particles.” Powder Technology, 58(1), 63–70. → pages 166Hansen, D. (2004). “Discussion of ”On the use of the KozenyCarman equation to predict thehydraulic conductivity of soils”.” Canadian Geotechnical Journal, 41(5), 990–993. →pages 15Holtz, R. D., Kovacs, W. D., and Sheahan, T. C. (2011). An Introduction To GeotechnicalEngineering. Pearson Education, Inc., Upper Saddle River, NJ. → pages 9, 14, 167Honjo, Y., Haque, M. A., and Tsai, K. A. (1996). “Self-filtration behaviour of broadly and gapgraded cohesionless soils.” Geofilters’96 , May 29-31, Montreal, Canada, J. Lafleur and A.Rollin, eds., Richmond, B.C., Canada, BiTech Publications, 227–236. → pages 17, 18, 28ICOLD (2013). “Internal erosion of existing dams, levees and dikes, and their foundations.Volume 1: Internal erosion processes and engineering assessment (pre-print).” Report No.Draft, Commission Internationale des Grand Barrages, Paris, France. → pages 1202Indraratna, B., Vo Trong, N., and Rujikiatkamjorn, C. (2011). “Assessing the Potential ofInternal Erosion and Suffusion of Granular Soils.” Journal of Geotechnical andGeoenvironmental Engineering, 137(5), 550–554. → pages 21Jones, J. (1981). The nature of soil piping - A review of research. Geo Books, Norwich. →pages 25Ke, L. and Takahashi, A. (2012). “Strength reduction of cohesionless soil due to internalerosion induced by one-dimensional upward seepage flow.” Soils and Foundations, 52(4),698–711. → pages 17, 20, 119Ke, L. and Takahashi, A. (2014a). “Experimental investigations on suffusion characteristicsand its mechanical consequences on saturated cohesionless soil.” Soils and Foundations,54(4), 713–730. → pages 20Ke, L. and Takahashi, A. (2014b). “Triaxial Erosion Test for Evaluation of MechanicalConsequences of Internal Erosion.” Geotechnical Testing Journal, 37(2), 1–18. → pages 3,17, 18, 21, 31, 40Kenney, T. C., Chahal, R., Chiu, E., Ofoegbu, G. I., Omange, G. N., and Ume, C. A. (1984).“Controlling constriction sizes of granular filters.” Canadian Geotechnical Journal, 22(1),32–43. → pages 21Kenney, T. C., Chahal, R., Chiu, E., Ofoegbu, G. I., Omange, G. N., and Ume, C. A. (1985).“Controlling constriction sizes of granular filters.” Canadian Geotechnical Journal, 22(1),32–43. → pages 10Kenney, T. C. and Lau, D. (1985). “Internal stability of granular filters.” CanadianGeotechnical Journal, 22(2), 215–225. → pages 2, 10, 17, 18, 19, 22, 26, 120Kenney, T. C. and Lau, D. (1986). “Internal stability of granular filters - Reply.” CanadianGeotechnical Journal, 23(3), 420–423. → pages 2, 19Kezdi, A. (1979). Soil physics. Elsevier, Amsterdam. → pages 2, 6, 10, 18, 19, 22, 24, 120,161Kovacs, G. (1981). Seepage hydraulics. Elsevier, Amsterdam. → pages 6, 18, 25, 120Krumbein, W. and Sloss, L. (1963). Stratigraphy and sedimentation. Freeman, San Francisco,2nd edition. → pages 13Ku, H. H. (1966). “Notes on the Use of Propagation of Error Formulas.” Journal of Researchof the National Bureau of Standards, 70C(4), 263–273. → pages 209Kuerbis, R. and Vaid, Y. P. (1988). “Sand sample preparation - the slurry depositiontechnique.” Soils and Foundations, 28, 107–118 ST – Sand sample preparation – the slurry.→ pages 46Lafleur, J., Mlynarek, J., and Rollin, A. L. (1989). “Filtration of broadly graded cohesionlesssoils.” Journal of Geotechnical Engineering, 115(12), 1747–1768. → pages 17, 18, 26, 119203Leroueil, S. and Hight, D. W. (2003). “Behaviour and properties of natural soils and softrocks.” Characterization and Engineering Properties of Soils, T. Tan, K. Phoon, D. Hight,and S. Leroueil, eds., Vol. 1, Singapore, Balkema, Rotterdam, 29–254. → pages 9Li, M. (2008). “Seepage induced instability in widely graded soils.” Ph.D. thesis, TheUniversity of British Columbia, The University of British Columbia. → pages 2, 3, 6, 7, 8,17, 18, 20, 21, 29, 119, 153, 156, 157, 158, 159, 163, 164, 168, 217Li, M. and Fannin, R. J. (2008). “Comparison of two criteria for internal stability of granularsoil.” Canadian Geotechnical Journal, 45(9), 1303–1309. → pages 19Li, M. and Fannin, R. J. (2012). “A theoretical envelope for internal instability of cohesionlesssoil.” Geotechnique, 62(1), 77–80. → pages 20, 216Luo, Y.-l., Qiao, L., Liu, X.-x., Zhan, M.-l., and Sheng, J.-c. (2012). “Hydro-mechanicalexperiments on suffusion under long-term large hydraulic heads.” Natural Hazards, 65(3),1361–1377. → pages 3, 17, 18, 19, 21, 31, 40McGeary, R. K. (1961). “Mechanical Packing of Spherical Particles.” Journal of the AmericanCeramic Society, 44(10), 513–522. → pages 9, 10, 12, 44, 120, 121, 122Mitchell, J. and Soga, K. (2005). Fundamentals of soil behavior. John Wiley & Sons, Inc. →pages 6, 9, 14, 15, 119Moffat, R. (2005). “Experiments on the internal stability of widely graded cohesionless soils.”Ph.D. thesis, The University of British Columnbia, The University of British Columnbia. →pages 2, 3, 19, 21, 153, 158, 169, 170, 195Moffat, R. and Fannin, R. J. (2011). “A hydromechanical relation governing internal stabilityof cohesionless soil.” Canadian Geotechnical Journal, 48(3), 413–424. → pages 20Moffat, R., Fannin, R. J., and Garner, S. J. (2011). “Spatial and temporal progression ofinternal erosion in cohesionless soil.” Canadian Geotechnical Journal, 48(3), 399–412. →pages 2, 3, 6, 7, 18, 20, 30, 119, 168Moffat, R. A. and Fannin, R. J. (2006). “A large permeameter for study of internal stability incohesionless soils.” Geotechnical Testing Journal, 29(4), 273–279. → pages 17, 21, 40, 41,46, 125Molenkamp, F., Calle, E. O. F., Heusdens, J. J., and Koenders, M. A. (1979). “Cyclic filtertests in a triaxial cell.” 7th European Conference on Soil Mechanics and FoundationEngineering, Vol. 2, Brighton, UK, 97–101. → pages 7, 17, 18, 25, 31Ng, C. W. W., Zhan, L. T., and Cui, Y. J. (2002). “A new simple system for measuring volumechanges in unsaturated soils.” 39(3), 757–764. → pages 18Polvino, S. (2011). “Accuracy, Precision, and Resolution in Strain Measurements onDiffraction Instruments.” Ph.d. thesis, Columbia University, Columbia University. → pages208Powers, M. (1953). “A new roundness scale for sedimentary particles.” Journal ofSedimentary Research, 23(2), 117–119. → pages 13204Rahman, M. M., Lo, S. R., and Gnanendran, C. T. (2008). “On equivalent granular void ratioand steady state behaviour of loose sand with fines.” Canadian Geotechnical Journal,45(10), 1439–1456. → pages 12Richards, K. S. and Reddy, K. R. (2007). “Critical appraisal of piping phenomena in earthdams.” Bulletin of Engineering Geology and the Environment, 66(4), 381–402. → pages 2,6, 7, 29Sail, Y., Marot, D., Sibille, L., and Alexis, A. (2011). “Suffusion tests on cohesionlessgranular matter Experimental study.” European Journal of Environmental and CivilEngineering, 15, 799–817. → pages 17, 21, 153, 156, 157, 163, 164Sanchez, R. L., Strutynsky, A. I., and Silver, M. L. (1983). “Evaluation of the erosion potentialof embankment core materials using the laboratory triaxial erosion test procedure.” ReportNo. GL-83-4, US Army Corps of Engineers. → pages 17, 31Scott, G. D. (1960). “Packing of Spheres.” Nature, 188(4754), 908–909. → pages 10, 44, 121Scott, G. D. and Kilgour, D. M. (1969). “The density of random close packing of spheres.”Journal of Physics D: Applied Physics, 2(6), 863–866. → pages 9, 12, 44, 120, 121Sharma, R. (1998). “Mechanical behaviour of unsaturated highly expansive clays.” Ph.D.thesis, University of Oxford, University of Oxford. → pages 18Sherard, J. L. (1979). “Sinkholes in dams of coarse, broadly graded soils.” Thirteenth Congresof Large Dams, ICOLD, New Delhi. → pages 2, 19, 25, 161Shire, T. and O‘Sullivan, C. (2013). “Micromechanical assessment of an internal stabilitycriterion.” Acta Geotechnica, 8(1), 81–90. → pages 2, 21, 44, 159, 160, 176, 196Shire, T., O‘Sullivan, C., Hanley, K., and Fannin, R. J. (2014). “Fabric and Effective StressDistribution in Internally Unstable Soils.” Journal of Geotechnical and GeoenvironmentalEngineering, 140(12), 1–11. → pages 2Skempton, A. W. and Brogan, J. M. (1994). “Experiments on piping in sandy gravels.”Geotechnique, 44(3), 449–460. → pages 2, 7, 10, 17, 18, 19, 22, 27, 119, 120, 122, 153, 169Sterpi, D. (2003). “Effects of the Erosion and Transport of Fine Particles due to SeepageFlow.” International Journal of Geomechanics, 3(1), 111–122. → pages 17Sun, B.-B. (1989). “Internal stability of clayey to silty sands.” Ph.D. thesis, University ofMichigan, University of Michigan. → pages 17, 19, 31, 39Sympatec (2008). “Windoxoperating instructions release 5.4.1.0. → pages 13, 44Tatsuoka, F. (1981). “A simple method for automatic measurement of volume change inlaboratory tests.” Soils and Foundations, 21(3), 104–106. → pages 43Terzaghi, K. and Peck, R. (1948). Soil Mechanics in Engineering Practice. John Wiley &Sons, Inc., New York. → pages 7, 119205Thevanayagam, S. (1998). “Effect of fines and confining stress on undrained shear strength ofsilty sands.” Journal of Geotechnical and Geoenvironmental Engineering, 124(6), 479–491.→ pages 11Thevanayagam, S. and Mohan, S. (2000). “Intergranular state variables and stressstrainbehaviour of silty sands.” Ge´otechnique, 50(1), 1–23. → pages 12, 162Thevanayagam, S., Shenthan, T., Mohan, S., and Liang, J. (2002). “Undrained fragility ofclean sands, silty sands, and sandy silts.” Journal of Geotechnical and GeoenvironmentalEngineering, 128(10), 849–859. → pages 11, 12, 22, 120, 121, 122, 123, 124, 139, 160,194, 198USACE (1953). “Filter experiments and design criteria.” Report No. AEWES-TM-3-360, ArmyEngineer Waterways Experiment Station, Vicksburg, Missisippi. → pages 17, 18, 19, 24USBR (2011). “Protective Filters.” Report No. DS-13(5)-9, United States Department of theInterior - Bureau of Reclamation, Denver, CO. → pages 1Vaid, Y. P. and Negussey, D. (1984). “A Critical Assessment of Membrane Penetration in theTriaxial Test.” Geotechnical Testing Journal, 7(2), 70–76. → pages 42, 213Vaid, Y. P. and Negussey, D. (1988). “Preparation of Reconstituted Sand Specimens.”Advanced Triaxial Testing of Soil and Rock, Donaghe, ed., Louisville, KY, AmericanSociety for Testing and Materials, 405–417. → pages 46Vaid, Y. P. and Sivathayalan, S. (2000). “Fundamental factors affecting liquefactionsusceptibility of sands.” Canadian Geotechnical Journal, 37(3), 592–606. → pages 11Vardoulakis, I. (2004). “Fluidisation in artesian flow conditions: Hydromechanically stablegranular media.” Ge´otechnique, 54(2), 117–130. → pages 8Wan, C. F. and Fell, R. (2004). “Experimental investigation of internal instability of soils inembankment dams and their foundations.” Report No. 429, University of New South Wales,Sydney, Australia. → pages 17, 18, 28, 119Wan, C. F. and Fell, R. (2008). “Assessing the potential of internal instability and suffusion inembankment dams and their foundations.” Journal of Geotechnical and GeoenvironmentalEngineering, 134(3), 401–407. → pages 2, 19Wheeler, S. (1986). “The stress-strain behaviour of soils containing gas bubbles.” Ph.D. thesis,The University of Oxford, The University of Oxford. → pages 18, 38, 43, 193, 197Wittmann, L. (1977). “Some aspects of transport processes in porous media.” 6th AustralianHydraulics and Fluid Mechanics Conference, Adelaide, Australia, 121–124. → pages 17Wittmann, L. (1978). “Phenomena and parameters of two-component-soil.” Symposium on theeffects of flow through porous media, IAHR, Greece, 68–80. → pages 2, 6, 7, 10, 22, 24, 120Xiao, M. and Shwiyhat, N. (2012). “Experimental Investigation of the Effects of Suffusion onPhysical and Geomechanic Characteristics of Sandy Soils.” Geotechnical Testing Journal,35(6), 1–11. → pages 3, 17, 18, 20, 31206Yin, J.-h. (2003). “A Double Cell Triaxial System for Continuous Measurement of VolumeChanges of an Unsaturated or Saturated Soil Specimen in Triaxial Testing.” 26(3), 1–6. →pages 18Youd, T. (1973). “Factors controlling maximum and minimum densities of sands.” Evaluationof relative density and its roll in geotechnical projects involvling cohesionless soil, E. Seligand R. Ladd, eds., Los Angeles. CA, American Society for Testing and Materials, 98–112.→ pages 14, 165207Appendix AMeasurement uncertaintyASTM E2655 (ASTM, 2008) describes uncertainty as “an indication of the magnitude of errorassociated with a value that takes into account both systematic errors and random errors associ-ated with the measurement or test process.” Accuracy is a commonly referenced when seekingto quantify the uncertainty of measured parameters. In this study, accuracy is defined as thestandard deviation of the calibration data around the linear regression line. Precision charac-terises the statistical variability of a measured value; it is defined as the standard deviation ofthe measured value around the mean measured value and thus not related to the calibration ofa measurement technique. Resolution quantifies the smallest significant change of a measuredvalue that can be detected, for which no standard mathematical expression appears to exist(Polvino, 2011). Consider two identically shaped normal distributions with standard deviations: one distribution with mean value x¯ and a second distribution with mean value x¯ + 4 s. Onlythe tail ends of 2.5 % of the values of each distribution would overlap and the values can besatisfactory distinguished within the 95% confidence intervals. Accordingly, the resolution isdefined as four times the precision. The remainder of this Appendix describes how the accuracyof the measured values is determined, and how these measured uncertainties propagate into theuncertainties of derived quantities.A.1 Uncertainty in measured quantitiesA linear regression is derived from the calibration data, including an unbiased estimate of theuncertainty, as a standard deviation, si, of the calibration data around the linear regression line:(si)2 =1n−1n∑i=1(zi− z¯i)2 (A.1)with base variable zi and z¯i = predicted value of variable ziThe uncertainties in the measured quantities, determined using Eq. A.1 or assumed values,are presented in Tables A.1 and A.2.208A.2 Propagation of uncertaintyFor this study, the method of propagation of uncertainties suggested by ASTM E2655 (ASTM,2008) is followed. The parameters reported in this study, for example p′, q, e, v, i, k, εa, εv, aretypically derived from combinations of measured quantities. Determination of the uncertaintyin the derived parameters is then a purely mathematical exercise, which can be simplified by as-suming that: 1) errors in measurements of different variables are uncorrelated; and 2) errors arerelatively small compared to measured or mean values (Ku, 1966). Based on these assumptions,the propagation of uncertainty in derived parameters can be approximated using the followingequation:(sx)2 =n∑i=1(δ fδ zi)2(si)2 (A.2)where sx = standard deviation of derived variable zx which is defined by function; and si =standard deviation of base variable zi. The uncertainty of the derived variable is evaluated atthe mean values of zi. The uncertainty in the derived quantities is presented in Table A.3.209Table A.1: Uncertainty in measured quantities (Part 1 of 2).Measured quantity Unit TypicalvalueStandarddevia-tionDescriptionSpecimen length mm 100 2 Uncertainty in determining speci-men length between top and bot-tom wire meshesSpecimen diameter mm 100 0.2 Uncertainty in determining inter-nal diameter of split mold using avernier caliperAxial deformation mm 0 to 15 0.01 Calibration uncertainty of LVDTWater level in measurement bu-rettemm NA 1 Measurement uncertainty of man-ual reading of water level in mea-surement buretteSpecimen mass g 1200to15000.5 Sum of balance resolution and po-tential for mass loss during speci-men transferMass of collected water in manualdischarge measurementg 200 to20001 Sum of balance resolution and po-tential for moisture remaining incontainer between measurementsMass of top cap g 1234 1 Measurement uncertainty of massof top capTest duration h 2-8 1/3600 Measurement uncertainty of man-ual time measurementElapsed time in manual dischargemeasurements 120 to6001 Measurement uncertainty of man-ual time measurement210Table A.2: Uncertainty in measured quantities (Part 2 of 2).Measured quantity Unit TypicalvalueStandarddevia-tionDescriptionCross sectional area of measure-ment burette #1cm2 1.27 0.02 Measurement uncertainty basedon five measurementsCross sectional area of measure-ment burette #2cm2 0.472 0.003 Measurement uncertainty basedon five measurementsCell pressure in TPT #1 kPa 0-150 0.2 Calibration uncertainty of TPT #1Pore water pressure in TPT #2 kPa 0-150 0.5 Calibration uncertainty of TPT #2Differential pressure in DPT #1 cmH20 0-160 0.1 Calibration uncertainty of DPT #1Differential pressure in DPT #2 cmH20 0-40 0.016 Calibration uncertainty of DPT #2Calibration factor of time-dependent volume change of innerchambercm3/h 0.036 0.025 Determined based on ten calibra-tion measurementsConstant for proportionality ofmembrane penetration per unitareakg/cm2 0.016 0.004 Determined from six measure-ments of membrane penetrationDensity of solids g/cm3 2.5and2.70.01 Measurement variation based onthree tests for GB and BT mate-rials, respectively.211Table A.3: Uncertainty in derived quantities.Derived quantity Unit Typical value Standard deviationmean effective stress, p′ kPa 50 0.2150 0.2deviatoric stress, q kPa 0 0.6void ratio at the end of consolidation, e - 0.50 0.03axial strain, εa % 0.00 0.012.5 0.05volume change, ∆V cm3 0.0 0.1824.0 0.24volumetric strain, εv % 0.00 0.033.00 0.07hydraulic gradient, i - 0.00 0.0110.00 0.02specific discharge, v cm/s 0.0040 0.00040.20 0.02hydraulic conductivity, k cm/s 0.020 0.002212Appendix BMembrane complianceB.1 Measurement of membrane complianceIn the experimental procedure, volume change is measured using two different techniques (seeFig. 3.3): during consolidation of the specimen, volume change is measured in a measurementburette, whereas volume change of the specimen during multi-stage seepage flow is deducedfrom measurement of volume change of the cell fluid. The effective stress in the specimenchanges during both stages, which necessitates the volume change as a consequence of mem-brane compliance be accounted for:∆Vt = ∆Vm +∆V (B.1)where ∆Vt = total measured volume change, including membrane compliance, ∆Vm = vol-ume change resulting from membrane compliance, and ∆V = volume change of the specimen.Frydman et al. (1973) measured the volumetric deformations of three uniform glass beadsmaterials in both a triaxial apparatus and hollow cylinder apparatus and deduced the membranecompliance from the difference in volume change measured in both devices. The results indi-cated that for a given particle size, the membrane penetration per unit area, ∆vm in [cm3/cm2],is directly proportional to the logarithm of the cell pressure. The constant of proportionalitywas defined as slope Sm (see Fig. B.1).The membrane penetration per unit area. ∆vm is defined as:∆vm =∆VmAm(B.2)where Am = soil surface covered by the membrane.213Vaid and Negussey (1984) proposed two methods to determine the membrane compliancein a triaxial apparatus. The first method, Method 1, requires data from tests on at least twospecimens of different diameter. Method 2, which was validated using the results of Method1, can be included during the consolidation stage of a test: assuming isotropy of the specimenduring unloading, the membrane penetration per unit area can be determined as the differencebetween the measured volume change, ∆Vt and an ‘expected’ volume change during unloading,which is calculated assuming:εv,u = 3εa,u (B.3)where εa,u = axial strain during isotropic unloading and εv,u = volumetric strain duringisotropic unloading.The constant for proportionality of membrane penetration per unit area, Sm can then bedetermined by:Sm =(∆Vt − εv,uV )/Amlog10(σc, f /σc,i)(B.4)where σc, f = final cell pressure, σc,i = initial cell pressure, and V = specimen volume.The constant for proportionality of membrane penetration per unit area has been determinedfor three tests on GB material and BT material, respectively, see Table B.1 and Fig. B.1, withparticle size d = D′50 of the coarse fraction. The variability of the membrane compliance isattributed to the binary nature of the particle size distributions. The values determined for Sm ofbinary mixtures with S f ≤ 35 % appear in reasonable agreement with the findings of Frydmanet al. (1973) if d′50 of the fine fraction is taken as the governing particle size.B.2 Effect of membrane compliance in this studyAnalysis of the parameter space of typical values of the test variables in this research, yieldinsights into the relative contribution of membrane compliance to the volume changes duringconsolidation and during multi-stage seepage flow. Consider the following materials, whichrepresent the limits of the materials tested: 1) material A has a relatively high (for this study)void ratio, not corrected for membrane compliance, e = 0.51 and d′50 = 0.15 mm; and 2)material B has a relatively low void ratio, not corrected for membrane compliance, e = 0.25 andd′50 = 0.15 mm. The typical specimen dimension is V ≈ 785 cm3. The specimen consolidatesunder the increase of the cell pressure from σc ≈ 20 kPa to 150 kPa. Using Eqs. B.1, B.2and B.4, and selecting Sm according to Fig. B.1, the effect of the membrane penetration onthe void ratio of both materials can be calculated: e = 0.511, assuming Sm = 0.002, and e= 0.250, assuming Sm = 0.002 kg/cm2, for materials A and B, respectively. Hence, the void214Table B.1: Measurement of membrane compliance.Test Sm (kg/cm2)4.8GB20-100 0.00314.8GB20-150 0.00554.8GB35-150 0.00465.1BT20-150 0.00127.0BT20-100 0.00627.0BT20-150 0.0061This study: GB gradationsBT gradationsFigure B.1: Membrane compliance: variation between particle size and constant of pro-portionality of membrane penetration per unit are (source: Frydman et al., 1973.Reprinted, with permission, from the Journal Testing and Evaluation 1(1) (1973),copyright ASTM International, 100 Barr Harbor Drive, West Conshohocken, PA19428).ratio would be estimated as 0.1% and 0.3% too low for materials A and B respectively, if themembrane compliance was ignored. The change of the principle effective stress during multi-stage seepage flow is less than 10 kPa. For materials A and B, if isotropically consolidated to50 kPa, the volume change associated with membrane compliance resulting from ∆p′ = 10 kPais approximately 0.03 cm3, equivalent to εv = 0.004 %.215Appendix CVariation of effective stress duringmulti-stage seepage flowThe test procedure adopted in this investigation commences with specimen preparation, fol-lowed by consolidation in stage 1 and multi-stage upward seepage flow in stage 2. In stage 1,the specimen is isotropically consolidated to a mean effective stress p′c. In stage 2, the speci-men is subject to an in discrete increments increasing pore water pressure at the bottom of thespecimen as a result of the inflow constant-head device (see Fig. 3.1) being raised. The eleva-tion of the outflow constant-head device remains constant, which yields a constant pore waterpressure at the top of the specimen: the difference in pore water pressure between the top andbottom of the specimen is ∆u. Accordingly, the mean effective stress at the top of the specimenremains constant throughout the test, while the mean effective stress at the bottom of the spec-imen decreases with each subsequent stage of seepage flow. The mean effective stress at thecentre of the specimen p′ is reported in the test results, which thus decreases with subsequentstages of seepage flow according to:p′ = p′c−∆u2(C.1)C.1 On the scale effect of hydraulic gradientLi and Fannin (2012) inferred the existence of a ‘scale effect’ in reporting the results of seepage-induced internal instability tests on specimens of different lengths. Consider two elements ofdifferent length (see Fig. C.1). A small element, with length ∆z is isotropically consolidatedto p′c. When subject to upward seepage flow, ultimately fluidisation of the element wouldoccur if the mean effective stress at the bottom of the element reduces to zero, which yields thedifferential pressure at the onset of fluidisation ∆u f = p′c. The corresponding hydraulic gradienti f and mean principle effective stress p′f at the onset of fluidisation of the element with length216∆z are:i f (∆z) =∆u f∆zγw(C.2)p′f = p′c−12∆u f =12p′c (C.3)Now consider a larger element with length (n ∆z) isotropically consolidated to the samevalue of p′c. Fluidisation would still occur at the same values of ∆u f = p′c and p′f = 1/2 p′c asidentified prior for the element with length ∆z. However, the corresponding hydraulic gradientacross the element with length (n ∆z) would be:i f (n∆z) =∆u fn∆zγw=i f (∆z)n(C.4)Hence, when reporting the conditions at the onset of fluidisation in terms of hydraulic gradi-ent, for elements of different length, a scale effect may be observed. Li (2008) actually reportedthis effect when comparing the conditions at failure, obtained from tests in a large and a smallrigid wall permeameter, respectively. The apparent scale effect can thus be attributed to theselection of the hydraulic gradient, which is a measure of the drag force per unit volume (seeSection 2.3), instead of the use of the differential pore water pressure across the specimen,which is a measure of the seepage-induced change of effective stress.217p'cp'f ∆uf ∆z p'cp'f ∆uf n ∆zFluidisation if: ∆uf = p'c equivalent toif(∆z) = ∆uf / (∆z γw)at p'f = p'c - ∆uf / 2zp',∆u zp',∆u Fluidisation if: ∆uf = p'cequivalent toif(n ∆z) = ∆u / (n ∆z γw) = if(∆z) / nat p'f = p'c - ∆uf / 2 Figure C.1: Scale effect of hydraulic gradient.218Appendix DForensic observationsThe forensic evidence compiled in this Appendix comprises visual observations (Section D.1)and post-test particle size analyses (Section D.2).D.1 Visual observationsVisual observations were recorded to examine the nature of the volumetric deformations, throughthe walls of the water bath and the acrylic tubes, at the end of the tests as a “front view” and a“side view”, using a digital camera. The quintessential observations presented in Figs. D.8 andD.10 have been annotated to aid in the identification of the nature of the deformations. In ad-dition, photographs were taken of the top surface of the post-test specimen during the forensicanalysis. A selection of these images is presented in this Appendix as supporting evidence forthe analysis of the test results reported in Chapter 5.219Figure D.1: 4.8GB35-50: end-of-test side view shows small sign of local distress, nearthe top cap.220Figure D.2: 6.0GB25-50: end-of-test front view shows signs of local distress at the tophalf of the specimen.221Figure D.3: 6.0GB30-50: end-of-test front view shows clear signs of local distress on theright side.222Figure D.4: 6.0GB30-150: end-of-test side view shows clear signs of local distress.Figure D.5: 6.0GB30-150: top view, after dis-assembly of the device.223Figure D.6: 6.0GB35-50: end-of-test side view shows clear signs of distress.Figure D.7: 6.0GB35-50: top view, after dis-assembly of the device, shows an abundanceof fine particles.224Seepage-induced deformationTop capBase pedestalSpecimenFigure D.8: 6.0GB35-100: end-of-test side view shows clear signs of local distress.Figure D.9: 6.0GB35-100: top view, after dis-assembly of the device, shows an abun-dance of fine particles.225Seepage-induced deformationTop capBase pedestalSpecimenFigure D.10: 6.5GB35-100: end-of-test side view.Figure D.11: 6.5GB35-100: top view, after dis-assembly of the device, shows an abun-dance of fine particles.226Figure D.12: 10.4BT25-50: end-of-test side view shows coarse particles at the top half ofthe specimen protruding under the membrane.Figure D.13: 10.4BT25-50: top view, after dis-assembly of the device, shows an abun-dance of fine particles.227Figure D.14: 10.4BT30-50: end-of-test front view shows coarse particles at the top halfof the specimen protruding under the membrane.228Figure D.15: 10.4BT30-50: top view, after dis-assembly of the device, shows an abun-dance of fine particles.229Figure D.16: 10.4BT35-50: end-of-test side view, shows signs of local distress near thetop cap.230Figure D.17: 10.4BT35-50: top view, after dis-assembly of the device, shows an abun-dance of fine particles.231Figure D.18: 10.4BT35-50(R): end-of-test side view, shows signs of local distress at thetop half of the specimen.232Figure D.19: 10.4BT35-50(R): end-of-test front view shows signs of local distress at theright side of the specimen.233Figure D.20: 10.4BT35-50(R): top view, after dis-assembly of the device, shows an abun-dance of fine particles.234Figure D.21: 10.4BT35-100: end-of-test side view, shows signs of local distress on theleft side of the specimen.235Figure D.22: 10.4BT35-100: end-of-test front view, shows signs of local distress at theright side of the specimen.236D.2 Post-test particle size analysesFor nearly all tests, a forensic particle size analysis was performed on the specimens at the endof a test. The procedure was as follows: (i) following the end of the test, the cell pressure wasgradually reduced to σc ≈ 20 kPa; (ii) all valves were closed, yielding an undrained specimenand a vacuum pressure of u ≈ - 20 kPa was applied through measurement burette #1 and thecell pressure reduced to atmospheric pressure; (iii) the triaxial cell was carefully disassembledand the water level in the water bath lowered to below the bottom of the specimen; (iv) theforming mold was then placed around the specimen and the vacuum on the specimen wassubsequently released; (v) the top cap was removed and the specimen was dessicated in threeto five layers using a siphoning device; (vi) the materials from the dessicated layers were oven-dried, manually sieved, after which the mass of the components was recorded. Occasionally,the mold could not be placed around the specimen, which is attributed to the seepage-induceddeformations or deformations that occurred during dis-assembly of the triaxial cell. In thesecases, no post-test particle size analyses were conducted.237Figure D.23: 3.3GB20-50: post-test particle size analyses.Figure D.24: 4.8GB20-150: post-test particle size analyses.Figure D.25: 4.8GB35-100: post-test particle size analyses.238Figure D.26: 4.8GB35-150: post-test particle size analyses.Figure D.27: 6.0GB20-50: post-test particle size analyses.Figure D.28: 6.0GB20-100: post-test particle size analyses.239Figure D.29: 6.0GB20-150: post-test particle size analyses.Figure D.30: 6.0GB25-150: post-test particle size analyses.Figure D.31: 6.0GB30-150: post-test particle size analyses.240Figure D.32: 6.0GB35-100(R): post-test particle size analyses.Figure D.33: 6.0GB35-150: post-test particle size analyses.Figure D.34: 6.5GB35-50: post-test particle size analyses.241Figure D.35: 6.5GB35-100: post-test particle size analyses.Figure D.36: 5.1BT20-50: post-test particle size analyses.Figure D.37: 5.1BT20-150: post-test particle size analyses.242Figure D.38: 5.7BT20-50: post-test particle size analyses.Figure D.39: 5.7BT20-100: post-test particle size analyses.Figure D.40: 5.7BT20-150: post-test particle size analyses.243Figure D.41: 5.7BT35-100: post-test particle size analyses.Figure D.42: 7.0BT20-50: post-test particle size analyses.Figure D.43: 7.0BT20-150: post-test particle size analyses.244Figure D.44: 8.6BT20-50: post-test particle size analyses.Figure D.45: 10.4BT25-50: post-test particle size analyses.Figure D.46: 10.4BT30-50: post-test particle size analyses.245Figure D.47: 10.4BT35-50(R): post-test particle size analyses.Figure D.48: 10.4BT35-100: post-test particle size analyses.246

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.24.1-0166340/manifest

Comment

Related Items