UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Core movement during processing of sandwich panels Pawson, Duncan Joseph 2019

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

Item Metadata

Download

Media
24-ubc_2019_september_pawson_duncan.pdf [ 30.65MB ]
Metadata
JSON: 24-1.0380364.json
JSON-LD: 24-1.0380364-ld.json
RDF/XML (Pretty): 24-1.0380364-rdf.xml
RDF/JSON: 24-1.0380364-rdf.json
Turtle: 24-1.0380364-turtle.txt
N-Triples: 24-1.0380364-rdf-ntriples.txt
Original Record: 24-1.0380364-source.json
Full Text
24-1.0380364-fulltext.txt
Citation
24-1.0380364.ris

Full Text

Core Movement During Processing of Sandwich PanelsbyDuncan Joseph PawsonB.Sc. Mechanical Engineering, University of Calgary, 2015A THESIS SUBMITTED IN PARTIAL FULFILLMENTOF THE REQUIREMENTS FOR THE DEGREE OFMaster of Applied ScienceinTHE FACULTY OF GRADUATE AND POSTDOCTORAL STUDIES(Materials Engineering)The University of British Columbia(Vancouver)August 2019c© Duncan Joseph Pawson, 2019The following individuals certify that they have read, and recommend to the Faculty of Graduate andPostdoctoral Studies for acceptance, the thesis entitled:Core Movement During Processing of Sandwich Panelssubmitted by Duncan Joseph Pawson in partial fulfillment of the requirements for the degree of Masterof Applied Science in Materials Engineering.Examining Committee:Go¨ran Fernlund, Materials EngineeringSupervisorAnoush Poursartip, Materials EngineeringSupervisory Committee MemberChad Sinclair, Materials EnginieeringAdditional ExamineriiAbstractCore movement is a phenomenon that plagues sandwich panel fabrication during autoclave processing.External pressure can result in deformation of the core in its weak, lateral direction. This in turn,drags the facesheet plies inward. Parts displaying notable core movement are unfit for service andmust be scrapped. Restraining plies, thereby preventing core movement, has become common practicein industrial processing of sandwich panels. Progress has been made in mitigating the risk of coremovement through artificially increasing pressure within the core as well as through development ofhigh friction resin/fiber systems. However, a complete understanding of the physics of the problem isstill lacking.The focus of this research is to understand how processing conditions affect core movement andto gain a better understanding of the fundamental nature of core movement. Parameters investigatedinclude, temperature, pressure, and a variety of structural features such as tie downs, core machiningstabilization, and core chamfer angles. The problem is investigated using novel techniques. In-situdata collection is performed through use of displacement and in-core pressure sensors. Moreover, theentire event is filmed in real time using an autoclavable camera. This allowed for identification ofthe exact processing parameters during initiation and progression of core movement. Individual plymovement is determined post-processing. It was shown that failure pressure is dictated by viscosity atlower processing temperatures, but at high temperatures this is no longer the case. This implies a shiftin the lubrication regime and alters the mechanics of the problem.A basic mechanical model describing the core movement process is outlined and the conditionsnecessary for core movement initiation are proposed. It was shown that core movement initiates at thechamfer radius and fluctuates between progressing through the core and down the chamfer edge. Theresults suggest that friction is the governing mechanism behind core movement initiation, which, if true,has important ramifications for sandwich panel design.This work builds off previous research and lays the foundations for future work.iiiLay SummaryComposite sandwich panels are comprised of a low density core material surrounded by stiff, compositefacesheets. They are important for structural engineering applications, particularly in aerospace. Theyare typically processed in industrial-scale autoclaves, which apply temperature and pressure to the part.Unfortunately, the external pressure can often cause the core material to collapse inward; known ascore movement. This body of work investigates the basic parameters that influence core movement andrelates this to the underlying physics of the phenomenon. Novel techniques are implemented, whereinthe core movement process is measured in-situ through use of embedded sensors and a video camera.This allows for an accurate understanding of the problem and the conditions necessary to initiate coremovement.ivPrefaceThis dissertation titled “Core Movement During Processing of Sandwich Panels” is original work con-ducted and written by Duncan Pawson under the supervision of Professor Go¨ran Fernlund.All experiments and analysis outlined in Chapter 4, Chapter 5, and Chapter 6 were designed andperformed by Duncan Pawson while under the supervision of Professor Go¨ran Fernlund. The aims andobjectives described in Chapter 3 as well as Figure 3.1 were developed by Duncan Pawson while underthe supervision of Professor Go¨ran Fernlund. Each chapter was written by Duncan Pawson under thesupervision of Professor Go¨ran Fernlund. Materials for the research were provided by Mark Sheadat Boeing Canada - Winnipeg. The sensors and autoclavable camera were provided by ConvergentManufacturing Technologies.Results from Section 5.1 were presented at the 75th International Society for the Advancement ofMaterial and Process Engineering (SAMPE) in Charlotte, North Carolina, in May 2019. A conferencepaper was published in the proceedings with the following details. Pawson, Duncan and Fernlund,Go¨ran. (2019). “Core Movement During Processing of Sandwich Panels”, Proceedings of the Soci-ety for the Advancement of Material and Process Engineering - Seventy-fifth Technical Conference,Charlotte, 2019. The paper was written by Duncan Pawson under the supervision of Professor Go¨ranFernlund. The presentation was given by Duncan Pawson.Results from Section 5.3 were presented at the 11th Canadian - International Conference on Com-posites (CANCOM) in Kelowna, British Columbia, in July, 2019. A conference paper was published inthe proceedings with the title, “Layup effects on core movement in sandwich panels”. The listed authorsare Duncan Pawson and Go¨ran Fernlund. The conference paper was written by Duncan Pawson underthe supervision of Professor Go¨ran Fernlund. The presentation was given by Duncan Pawson.vTable of ContentsAbstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iiiLay Summary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ivPreface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vTable of Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viList of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ixList of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xList of Symbols . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xviiList of Abbreviations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xxiGlossary . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xxiiAcknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xxv1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.1 Composites and sandwich panels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.2 Core movement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22 Literature Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52.1 Honeycomb mechanics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52.1.1 Terminology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 52.1.2 Relative density . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72.1.3 Uniaxial loading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 72.1.4 Linear elastic bending . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92.1.5 Cell collapse . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102.1.6 Biaxial loading . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 112.2 Inter-ply friction of prepreg . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122.2.1 Pre-yield behaviour . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 13vi2.3 Core Movement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142.3.1 Physics of core movement . . . . . . . . . . . . . . . . . . . . . . . . . . . . 142.3.2 Mitigation methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162.3.3 Gas flow within the core . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183 Aims and Objectives . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 204 Methodology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 234.1 Experimental breakdown . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 234.2 Materials . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 244.3 Sensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 244.4 Layup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 264.4.1 Standard layup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 264.4.2 Specific layups . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 274.5 Cure cycle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 284.5.1 Standard cure cycle . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 294.5.2 Specific cure cycles . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 294.6 In-situ data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 304.7 Post cure data . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 324.8 Dry core test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 355 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 365.1 Effect of temperature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 365.1.1 Failure pressure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 365.1.2 Ply movement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 405.1.3 Deformation pattern . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 455.2 Effect of pressure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 495.2.1 Failure temperature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 495.2.2 Ply movement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 515.2.3 Extent of core movement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 545.3 Effect of layup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 565.3.1 Failure pressure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 565.3.2 Ply movement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 585.3.3 Extent of core movement . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 615.4 Dry core test . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 645.4.1 Failure pressure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 645.5 Core movement mechanics . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 685.5.1 Initiation of core movement . . . . . . . . . . . . . . . . . . . . . . . . . . . 685.5.2 Progression of core movement . . . . . . . . . . . . . . . . . . . . . . . . . . 695.5.3 In-core pressure . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89vii6 Summary and Improved Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 926.1 Drawback of current models . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 926.2 Core movement process . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 976.2.1 Conditions for initiation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 976.2.2 Conditions for progression . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1016.3 Processing window . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1036.4 Design improvements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1086.4.1 Recommendations . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1107 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1137.1 Summary of results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1137.2 Future work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 115Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117Appendix A    Data Acquisition . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  . 121A.1 LVDTs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121A.1.1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121A.1.2 AC conditioner calibration . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121A.2 In-core pressure sensors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123A.3 Autoclavable camera . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123Appendix B    Additional Post-processing Images . . . . . . . . . . . . . . . . . . . . . . .  . 124B.1 Post-processed panels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124B.2 Slope gradient maps . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 125B.3 Through-thickness section cuts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131B.4 In-plane section cuts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 138Appendix C    Additional Data Sets . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .  . 147C.1 120◦Corig results . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147viiiList of TablesTable 4.1 Summary of experiments. Note that the 120◦C and “Standard” experiment are thesame test and represent the baseline test. . . . . . . . . . . . . . . . . . . . . . . . 24Table 5.1 Extent of deformation in the ribbon and non-ribbon direction associated with differ-ent processing temperatures. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46Table 5.2 Difference in displacement between the ribbon and non-ribbon edges associated withdifferent processing temperatures. . . . . . . . . . . . . . . . . . . . . . . . . . . . 46Table 5.3 Extent of deformation in the ribbon and non-ribbon direction associated with differ-ent processing pressures. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55Table 5.4 Extent of deformation in the ribbon and non-ribbon direction associated with variouslayup features. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63Table 5.5 Average panel height within the crush zone compared with the degree of core move-ment in the L direction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 83Table 5.6 Average panel height within the crush zone compared with the degree of core move-ment in the W direction. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86Table 5.7 In-core pressure at onset and collapse. . . . . . . . . . . . . . . . . . . . . . . . . 89Table 6.1 Main parameters involved in the initiation of core movement. . . . . . . . . . . . . 109ixList of FiguresFigure 1.1 (A) Sandwich panel after layup and prior to cure. (B) Same panel following cure;drastic core movement has occured. . . . . . . . . . . . . . . . . . . . . . . . . . 3Figure 1.2 Processed sandwich panel on tool. Grit strips surrounding the part are used to re-strain plies (tie-downs). Indents in the panel are from pressure sensors. . . . . . . 4Figure 2.1 Hexagonal honeycomb structure. Glue lines are present in the ribbon direction (L)as indicated by the red lines. The non-ribbon direction (W) is perpendicular in-planeto this. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6Figure 2.2 Honeycomb unit cell, where ` represents length, h height, t thickness, c cell size,α the angle between each side, and θ the cell wall angle. For regular honeycombstructures h = `, α = 120◦, and θ = 30◦. The out-of-plane thickness is denoted as b. 6Figure 2.3 Typical compressive stress-strain curve for an elastomeric honeycomb. I: Bending -linear elastic bending of cell walls, II: Collapse (plateau region) - constant stress ascells fail, III: Densification - cell walls touching. Figure adapted from [22]. . . . . 8Figure 2.4 Effect of increasing nominal relative density on the compressive stress-strain re-sponse of honeycombs. I: Bending - linear elastic bending of cell walls, II: Collapse(plateau region) - constant stress as cells fail, III: Densification - cell walls touching.Figure adapted from [22]. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8Figure 2.5 Stribeck Curve demonstrating three regions of lubrication. I: Boundary lubrication,II: Mixed lubrication, III: Hydrodynamic lubrication. Figure adapted from [34]. . . 13Figure 2.6 FBD of sandwich panel demonstrating forces involved in core movement. Fsti f f nessincludes both the material stiffness of the skins and core in the lateral direction aswell as the in-core pressure. Figure adapted from [28]. . . . . . . . . . . . . . . . 16Figure 3.1 Core movement map. Major factors known to influence core movement are dis-played. Green arrows indicate when one property influences another. Propertieswith an asterisk are suspected to influence core movement, but their relation to coremovement has not yet been documented nor are they investigated in this body of work. 21xFigure 4.1 Test setup showing in-situ sensors. Top left is an LVDT sensor, top right is anin-core pressure sensor, and the bottom is the full setup in the autoclave with thecamera mounted above the panel. . . . . . . . . . . . . . . . . . . . . . . . . . . 25Figure 4.2 Standard sandwich panel layup. First layer against the tool is surfacing film, fol-lowed by 4 toolside plies. Core with pressure sensor and film adhesive layers isthen added, followed by 4 filler plies. Next, 4 bagside plies are draped over thecore. PTFE release film, breather, and vacuum bagging are then added. The LinearVariable Displacement Transducers (LVDTs) are placed after vacuum bagging iscomplete. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 26Figure 4.3 Core machining stabilization - 76.2 mm wide picture frame film adhesive strip iscured along the toolside of the core. (A) Toolside, (B) Bagside. . . . . . . . . . . . 28Figure 4.4 Representative viscosity profile for three different temperatures during cure. Adaptedfrom RAVEN data received from M. Shead (Boeing Winnipeg). . . . . . . . . . . 30Figure 4.5 Data output from in-situ sensors. The colour scheme is as follows: black - auto-clave pressure, blue – LVDT sensor, purple – in-core pressure sensor, green - vac-uum pressure. The region between onset and collapse represents the critical zone,wherein core movement initiates. . . . . . . . . . . . . . . . . . . . . . . . . . . 31Figure 4.6 Polished through-thickness section cut. The region between the panel edge and coreedge is referred to as the laminate edge. . . . . . . . . . . . . . . . . . . . . . . . 32Figure 4.7 Cross sectional image of laminate edge showing individual plies and where theyterminate after core movement. Numbers on the right identify the individual plieswhile the red Xs indicate where the associated ply terminates. . . . . . . . . . . . 33Figure 4.8 Polished in-plane section cut showing the deformation profiles in L and W. . . . . 34Figure 5.1 Net pressure acting on the sandwich panels at onset and collapse for the given (a)processing temperature and (b) associated resin viscosity. . . . . . . . . . . . . . . 37Figure 5.2 25◦C, 95◦C, 120◦C, and 180◦C tests post-processing, showing the extent of coremovement. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39Figure 5.3 Schematic showing the expected thickness changes along the laminate edge of asandwich panel that (a) experiences no core movement and (b) experiences coremovement. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41Figure 5.4 Change in thickness across laminate edge of processed panels in the L and W direc-tions at varying temperatures. The “No movement” case represents the test with tiedown plies, as no visible core movement was seen. . . . . . . . . . . . . . . . . . 42Figure 5.5 Microscope image of laminate edge showing surfacing film seeping through inter-stitial zones between tows. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43Figure 5.6 Specific ply movement profile of processed panels in L and W at varying tempera-tures. Initially, the core edge is 50.8 mm from the panel edge. . . . . . . . . . . . 44xiFigure 5.7 Core movement in ribbon and non-ribbon directions for (a) low versus (b) hightemperature processing. Autoclave pressure is absolute external pressure. Vacuumpressure is not shown. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47Figure 5.8 (a) Temperature at which onset and collapse occur for sandwich panels processed ata specific pressure and (b) the associated resin viscosity at failure. . . . . . . . . . 50Figure 5.9 Change in thickness across laminate edge of processed panels in the L and W di-rections at varying pressures. The “No movement” case represents the test with tiedown plies, as no visible core movement was seen. . . . . . . . . . . . . . . . . . 52Figure 5.10 Specific ply movement profile of processed panels in L and W at varying pressures.Initially, the core edge is 50.8 mm from the panel edge. . . . . . . . . . . . . . . . 53Figure 5.11 325 and 600 kPa tests post-processing, showing the extent of core movement. . . . 54Figure 5.12 Core movement in ribbon and non-ribbon direction for high pressure (600 kPa)processing. Autoclave pressure is absolute external pressure. Vacuum pressure isnot shown. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 55Figure 5.13 Minor deformations in cellular pattern of full tie down test. . . . . . . . . . . . . . 56Figure 5.14 Net pressure acting on the sandwich panels at onset and collapse for varying layups.The data for half tie downs represents that of the unrestrained edges. . . . . . . . . 57Figure 5.15 In-situ data from 45◦ core test. Collapse occurs immediately after venting of vac-uum, following a one minute hold at higher pressure. . . . . . . . . . . . . . . . . 57Figure 5.16 Change in thickness across laminate edge of processed panels in the L and W di-rections for varying layups. The full tie down cases show no externally visible coremovement. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 58Figure 5.17 Specific ply movement profile of processed panels in L and W for varying layups.Initially, the core edge is 50.8 mm from the panel edge. . . . . . . . . . . . . . . . 60Figure 5.18 Microscope images showing termination of plies 5-6 (red circle), ply 7 (yellowcircle), and ply 8 (blue circle) in the W and L directions for the Core MachiningStabilization (CMS) sample. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61Figure 5.19 Microscope images showing termination of plies 5-6 (red circle), ply 7 (yellowcircle), and ply 8 (blue circle) in the W and L directions for the 45◦ chamfer sample. 62Figure 5.20 Standard, full tie down, half tie down, CMS, and 45◦ chamfer tests post-processing,showing extent of core movement. . . . . . . . . . . . . . . . . . . . . . . . . . . 63Figure 5.21 In-situ results from dry core test. Cells deform upon loading and recover almostentirely upon unloading. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65Figure 5.22 Snapshots taken during autoclave test for dry honeycomb. 1: Prior to pressure ap-plication. 2: Moment of collapse. 3: Maximum deformation. 4: Fully unloadedhoneycomb core following depressurization. . . . . . . . . . . . . . . . . . . . . . 66Figure 5.23 Plastic deformation in W following autoclave pressurization. (A) Image of entirehoneycomb. (B) Magnified view of deformation in W. . . . . . . . . . . . . . . . 67xiiFigure 5.24 Change in surface roughness from onset of core movement through to collapse.Transitioning from 1-4, a notable change in surface roughness is seen within thegreen circle. The bagside plies conform to the honeycomb structure as it deforms,observable through the breather cloth. . . . . . . . . . . . . . . . . . . . . . . . . 69Figure 5.25 In-plane initiation of core movement showing the progression of cell bending through-out the honeycomb. Images are from different tests showing (a) no core movement,(b) minor bending, (c) widespread bending, and (d) early stages of collapse. . . . . 70Figure 5.26 Slope angle of processed panels. Dark red indicates a vertical slope with an angleof 90◦. Dark blue indicates a flat plane with an angle of 0◦. For most samples, aslope angle greater than 20◦ indicates crush has occurred along the chamfer edge.For the 45◦ chamfer, a slope angle larger than 45◦ indicates crush. . . . . . . . . . 71Figure 5.27 (a) Standard and (b) CMS crush patterns, displaying growth of the crush zone fromimage 1-4. Red arrows show the crush front in the L direction while blue arrowsshow the crush front in the W direction. Elapsed time from the application of pres-sure is displayed. Note the change in orientation between samples. . . . . . . . . . 72Figure 5.28 Crush pattern for 600 kPa sample, displaying growth of the crush zone from image1-4. W crush front (blue arrow) progresses at a faster rate than the L crush front(red arrow). Elapsed time from the application of temperature is displayed. . . . . 74Figure 5.29 Rate of deformation of the top and bottom of a 20◦ chamfer edge witnessed through(a) in-situ sensors and (b) in-situ camera footage. Red arrows points to the bottomof the chamfer edge while yellow arrows points to the top. Each square in (b) is12.7 x 12.7 mm, allowing for the crush rate to be determined. . . . . . . . . . . . 75Figure 5.30 Rate of deformation of the top and bottom of a 45◦ chamfer edge witnessed through(a) in-situ sensors and (b) in-situ camera footage. Red arrows points to the bottomof the chamfer edge while orange arrows points to the top. Each square in (b) is12.7 x 12.7 mm, allowing for the crush rate to be determined. . . . . . . . . . . . 76Figure 5.31 Schematic showing the manner of crush progression through the core. A: Crushinitiates at chamfer radius and progresses inward - bottom and top of chamfer moveat same rate. B: Crush initiates at chamfer radius and progresses down chamferedge - bottom of chamfer displaces quicker than top creating a concave surface. C:Actual manner of crush progression through core - combination of A-B-A. 1 - Crushinitiates at chamfer radius and progresses inward (similar to A). 2 - Chamfer edgecrushes quicker than cells beyond radius (similar to B). 3 - Crush switches again tocollapse cells further within core (similar to A). . . . . . . . . . . . . . . . . . . . 78Figure 5.32 Through-thickness section cuts of processed samples. Purple indicates the crushzone, blue the densified regions, and red the bagside wrinkles. Additional featuresare outlined in black. (a) The 120◦C, standard layup and (b) 45 chamfer samplesare provided here. Other samples are shown in the following images. . . . . . . . 79Figure 5.33 (a) CMS and (b) half tie-down samples. . . . . . . . . . . . . . . . . . . . . . . . 80xiiiFigure 5.34 Full tie-down and room temperature samples. . . . . . . . . . . . . . . . . . . . . 81Figure 5.35 General crush pattern through panel in the W direction. Image is taken from the120◦C, standard layup experiment. Purple indicates the crush zone, red the bagsidewrinkles, and blue the densified regions. . . . . . . . . . . . . . . . . . . . . . . . 83Figure 5.36 Buckling of plies nearest the core (blue arrows) on the (a) bagside and (b) toolsidesurfaces within the crush zone. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84Figure 5.37 Map of panel height extracted from Coordinate Measuring Machine (CMM) data for(a) no core movement and (b) drastic core movement. White bars in (b) representthe length of the crush zone in L and W. There is a notable increase in height alongthe crush zone. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85Figure 5.38 In-plane section cuts of processed panels showing densification along the L andW directions. The orange line represents the L crush front whereas the red linerepresents the W crush front. Yellow arrows point to regions of low densification inthe non-ribbon direction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87Figure 5.39 In-core pressure sensor response to core movement taken from the standard, 95◦Csample. Pressure recovery is captured upon release of vacuum. No difference inin-core pressure is observed between onset and collapse. Following collapse, con-siderable noise is seen due to pulling of the sensor wires. . . . . . . . . . . . . . . 90Figure 6.1 Schematic showing the tension (F1-F4) developed in bagside plies as core deforms.Friction (F12-FBag) at the interfaces is provided. The plies can be approximated assprings slipping relative to one another. . . . . . . . . . . . . . . . . . . . . . . . 95Figure 6.2 Schematic showing the first zone of resistance against core movement (i.e. thechamfer radius). In order for the cells to collapse, they must either slide relative tothe under and overlying plies or the plies must buckle with the cells. . . . . . . . . 96Figure 6.3 Schematic showing the second zone of resistance against core movement (i.e. thelaminate edge). In order for the plies to move, they must either stretch and/or slipacross an interface. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98Figure 6.4 Core movement initiation for standard and high temperature processing. Each modelincorporates the use of bagside Polytetrafluoroethylene (PTFE) release film and atacky surfacing film across the tool. . . . . . . . . . . . . . . . . . . . . . . . . . 100Figure 6.5 Schematic showing friction across the core-ply interface. . . . . . . . . . . . . . . 102Figure 6.6 Empirical failure envelope outlining the processing conditions necessary for coremovement. The curves are based on experimental results for a chamfer angle of20◦. Dotted lines are assumed values. Additional experimental results concerning achange in layup are provided at 120◦C. . . . . . . . . . . . . . . . . . . . . . . . 104Figure 6.7 Expected failure envelope for a chamfer angle of 45◦. The experimental results at120◦C are also provided. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105Figure 6.8 (a-b) General evolution of resistive forces with temperature and (c) Degree of Cure(DOC). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106xivFigure 6.9 Effect of altering the laminate edge length on prepreg-prepreg friction at 120◦C. Reddots represent core movement initiation. In-core pressure is assumed to be equal to1 atm. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109Figure 6.10 Effect of increasing the (a) core height and (b) chamfer angle for various laminateedge lengths at 120◦C. In-core pressure is assumed to be 1 atm. . . . . . . . . . . 111Figure A.1 LVDT wiring to conditioner module. Taken from [14]. . . . . . . . . . . . . . . . 122Figure A.2 Video footage synchronized with in-situ sensor data. 95◦C experiment. . . . . . . 123Figure B.1 Top-down view of processed panels. . . . . . . . . . . . . . . . . . . . . . . . . . 125Figure B.2 Slope gradient map for 95◦C sample. Although not shown, the colour bar appliesto each of the following images. Noise is seen in some of the samples which can bedisregarded. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126Figure B.3 Slope gradient map for 120◦Corig sample. . . . . . . . . . . . . . . . . . . . . . . 126Figure B.4 Slope gradient map for 120◦C sample. . . . . . . . . . . . . . . . . . . . . . . . . 127Figure B.5 Slope gradient map for 180◦C sample. . . . . . . . . . . . . . . . . . . . . . . . . 127Figure B.6 Slope gradient map for 325 kPa sample. . . . . . . . . . . . . . . . . . . . . . . . 128Figure B.7 Slope gradient map for 600 kPa sample. . . . . . . . . . . . . . . . . . . . . . . . 128Figure B.8 Slope gradient map for full tie-down sample. . . . . . . . . . . . . . . . . . . . . 129Figure B.9 Slope gradient map for half tie-down sample. . . . . . . . . . . . . . . . . . . . . 129Figure B.10 Slope gradient map for CMS sample. . . . . . . . . . . . . . . . . . . . . . . . . 130Figure B.11 Slope gradient map for 45◦ chamfer sample. . . . . . . . . . . . . . . . . . . . . . 130Figure B.12 Through-thickness section cut of room temperature sample. . . . . . . . . . . . . 132Figure B.13 Through-thickness section cut of 95◦C sample. . . . . . . . . . . . . . . . . . . . 132Figure B.14 Through-thickness section cut of 120◦Corig sample. . . . . . . . . . . . . . . . . . 133Figure B.15 Through-thickness section cut of 120◦C sample. . . . . . . . . . . . . . . . . . . 133Figure B.16 Through-thickness section cut of 180◦C sample. . . . . . . . . . . . . . . . . . . 134Figure B.17 Through-thickness section cut of 325 kPa sample. . . . . . . . . . . . . . . . . . . 134Figure B.18 Through-thickness section cut of 600 kPa sample. . . . . . . . . . . . . . . . . . . 135Figure B.19 Through-thickness section cut of full tie-down sample. . . . . . . . . . . . . . . . 135Figure B.20 Through-thickness section cut of half tie-down sample. . . . . . . . . . . . . . . . 136Figure B.21 Through-thickness section cut of CMS sample. . . . . . . . . . . . . . . . . . . . 136Figure B.22 Through-thickness section cut of 45◦ chamfer sample. . . . . . . . . . . . . . . . 137Figure B.23 In-plane section cut of room temperature sample. . . . . . . . . . . . . . . . . . . 138Figure B.24 In-plane section cut of 120◦C sample. . . . . . . . . . . . . . . . . . . . . . . . . 139Figure B.25 In-plane section cut of 180◦C sample. . . . . . . . . . . . . . . . . . . . . . . . . 140Figure B.26 In-plane section cut of 600 kPa sample. . . . . . . . . . . . . . . . . . . . . . . . 141Figure B.27 In-plane section cut of full tie-down sample. . . . . . . . . . . . . . . . . . . . . . 142Figure B.28 In-plane section cut of half tie-down sample between two restrained edges. . . . . 143Figure B.29 In-plane section cut of half tie-down sample between two unrestrained edges. . . . 144xvFigure B.30 In-plane section cut of CMS sample. . . . . . . . . . . . . . . . . . . . . . . . . . 145Figure B.31 In-plane section cut of 45◦ sample. . . . . . . . . . . . . . . . . . . . . . . . . . . 146Figure C.1 Core movement in ribbon and non-ribbon direction for 120◦Corig sample. Autoclavepressure is absolute external pressure. Vacuum pressure is not shown. . . . . . . . 147xviList of SymbolsAcore Area of core affected by lateral component of pressure [m2].α Angle between honeycomb cell members ` and h [rad].b Out-of-plane thickness of honeycomb cell [m].c Honeycomb cell size [m].d Film thickness [m].dγdtShear strain rate [s−1].E In-plane Young’s modulus of regular honeycomb [Pa].E1 Young’s modulus of honeycomb in X1 direction [Pa].E2 Young’s modulus of honeycomb in X2 direction [Pa].Es Young’s modulus of solid material from which honeycomb walls are comprised [Pa].Eply Young’s modulus of ply [Pa].(ε∗el)2 Strain at cell collapse under elastic buckling.ε In-plane compressive strain for regular honeycomb.η Dynamic viscosity [Pa·s].Ff Resistive force due to core material [N].Ff Coloumb friction force [N].Fg Force due to gravity (weight) [N].Fp−b Friction force between prepreg and bag [N].Fp−c Friction force between prepreg and core [N].Fp−p Friction force between prepreg and prepreg [N].xviiFp−t Friction force between prepreg and tool [N].Fsti f f ness Combined resistive force due to skin-core stifness and in-core pressure [N].Ftop Friction force due to bagside ply interactions [N].Fbtm Friction force due to toolside ply interactions [N].FT Tension force developed in plies due to strain [N].G12 In-plane shear modulus of honeycomb [Pa].Gply Inter-ply shear modulus [Pa].γ Shear strain.H Hershey Number.h Length of vertical honeycomb cell members [m].hcore Height of core [m].I Second moment of area [m4].j Shear joint strength [Pa].K Column effective length factor of plies undergoing buckling.L Ribbon direction.` Length of angled honeycomb cell members [m].µ Coefficient of friction.N Normal force [N].n Number of plies experiencing strain.np Rotational stiffness of honeycomb cell nodes.ν12 Poisson’s ratio of honeycomb describing strain in X2 direction resulting from stress in X1direction.ν21 Poisson’s ratio of honeycomb describing strain in X1 direction resulting from stress in X2direction.P Applied Pressure [Pa].Pcrit Critical load to cause Euler buckling [N].Phorizontal Horizontal component of applied pressure [Pa].xviiiPload Applied load in Euler buckling scenario [N].Pvertical Vertical component of applied pressure [Pa].ρ Density of cellular structure [kg/m3].ρs Density of solid material from which cell walls are comprised [kg/m3].ρρsRelative density.σ In-plane, applied compressive stress for regular honeycomb [Pa].σ1 Applied compressive stress in X1 directionv [Pa].σ2 Applied compressive stress in X2 direction [Pa].σ∗ Plateau stress of regular honeycomb [Pa].σ∗1 Plateau stress of honeycomb in X1 direction [Pa].σ∗2 Plateau stress of honeycomb in X2 direction [Pa].(σ∗el)2 Stress to cause cell collapse by elastic buckling [Pa].σn Normal Stress [Pa].(σ∗pl) Stress to cause cell collapse by formation of plastic hinges. Plastic collapse stress [Pa].σys Yield stress of solid material from which honeycomb walls are comprised [Pa].t Thickness of honeycomb cell member ` [m].Tg Glass transition temperature [◦C].τc Critical yield stress of plies [Pa].τint Inter-ply shear stress [Pa].τhyd Shear stress developed under hydrodynamic friction [Pa].(τ∗pl)12 Shear stress required to cause plastic deformation of honeycomb. Plastic shear strength[Pa].θ Cell wall angle of honeycomb [rad].θcore Angle of core chamfer edge [rad].v Relative speed between sliding plates [m/s or rad/s].W Non-ribbon direction.xixwcore Width of core [m].X1 First axis of honeycomb core. Corresponds to the in-plane, non-ribbon direction.X2 Second axis of honeycomb core. Corresponds to the in-plane, ribbon direction.X3 Third axis of honeycomb core. Corresponds to the through-thickness or transversedirection.xxList of AbbreviationsAC Alternating CurrentAISI American Iron and Steel InstituteBMS Boeing Material SpecificationCFRP Carbon Fiber Reinforced PolymerCMM Coordinate Measuring MachineCMS Core Machining StabilizationDOC Degree of CureGFRP Glass Fiber Reinforced PolymerLCM Liquid Composite MoldingLVDT Linear Variable Displacement TransducerMRCC Manufacturer Recommended Cure CycleNT Never TwistedOX OverexpandedPTFE PolytetrafluoroethyleneST Standard TwistUT Un-TwistedVBO Vacuum-Bag OnlyxxiGlossaryBagside Top side of the core. Nearest the bag.Bending Linear, elastic bending of honeycomb cell walls. Precedescollapse.Boundary lubrication Combination of coloumb friction and hydrodynamic frictionarising from contact of two surfaces seperated by a viscous layerof molecular thickness.Chamfer edge Inclined edges of the core.Chamfer radius Top of the chamfer edge.Core edge outside edge of the core.Core machining stabilization Method to stabilize core for machining purposes and to mitigatecore movement. Film adhesive precured to toolside of core.Core movement Process induced deformation of sandwich panels resulting fromcompression of the core material due to external pressure.Collapse (honeycomb mechanics) Failure of the honeycomb cells. Stress plateaus while strainincreases significantly. Proceeds bending and precedesdensification.Collapse (in-situ data) Point where deformation of core is obvious, and beyond which,continuous, and significant. Corresponds to collapse ofhoneycomb cells.Coloumb friction Classical friction arising due to contact between two surfaces.Refer to Equation 2.16.Critical Zone Range of pressure between onset and collapse from the in-situdata. Represents the range of pressure over which core movementinitiates.Crush front The edge of the crush zone that is propagating across the panel.xxiiCrush zone Region of the core that experiences densification during coremovement.Densification Cell walls begin to touch each other. Proceeds collapse.Dry core Honeycomb core without plies.Edgeband Region of panel laminate where all plies are located.Hydrodynamic friction Friction arising from the presence of a viscous film layer ofconsiderable thickness. Refer to Equation 2.15.Laminate edge Region of the panel where no core is present. Between the paneledge and core edge.Neat resin Resin in absence of fibers.NOMEX R© A popular honeycomb material for composite sandwich panels.Comprised of phenolic resin cured with aramid fibers.Non-ribbon direction Direction of honeycomb expansion. Perpendicular in-plane to theribbon direction. Refer to Figure 2.1.Onset First point of observable movement in sandwich panel. LVDTsensors deviate from breather deformation pattern. Corresponds tobending of honeycomb cells.Overexpanded honeycomb Honeycomb that has been expanded past the regular configuration.The cell wall angle is less than 30◦.Panel edge Outside edge of the panel.Pillowing Intrusion of plies into the core creating wrinkles.Prepreg Fibers/fabric pre-impregnated with resin.Regular honeycomb Honeycomb with cell wall angles of 30◦.Relative density Ratio of cell density to that of the solid material from which thecell is comprised. Refer to Equation 2.1.Ribbon direction Direction of the glue lines from which honeycomb is expanded.Refer to Figure 2.1.Sandwich panel A structure comprising a low density core with stiff over andunderlying facesheets for the purpose of increasing bendingrigidity.xxiiiTie-down plies Plies that are restrained from moving.Toolside Bottom side of the core. Nearest the tool.xxivAcknowledgmentsI would first and foremost like to thank my supervisor Professor Go¨ran Fernlund. His knowledge andguidance made this thesis a possibility and has helped shape my professional outlook moving forward.I enjoyed our many discussions both in relation to this research and outside of the university.I would also like to extend a special thanks to Professor Anoush Poursartip for filling in as supervisorwhile Go¨ran was away. His insight and wisdom helped bring focus to this research. An additional thanksto Professor Reza Vaziri for the thoughtful discussions.Thank you to The Boeing Company and especially Mark Shead who has been involved in the projectsince the beginning. Without his insight and experience, this project would not have been made a reality.Thank you also to the folks of Convergent Manufacturing for providing me with the tools necessaryto complete my research. In particular, I would like to thank Malcolm Lane and Trevor Fisher for theirtraining and technical expertise.I am grateful for all the help from each member of the Composite Research Network, particularlythe professional staff. Dr. Casey Keulen for his consistent help on many aspects of the project andfor training me on a variety of equipment, Dr. Navid Zobeiry for providing guidance and acting as aninterim co-supervisor while Go¨ran was away, Dr. Christophe Mobuchon for training me on the autoclaveand the many enjoyable discussions we had, and Roger Bennett for assisting in technical matters. Thankyou also to Ross Mcleod for helping with tooling for my project. I would like to extend a special thanksto Suzana Topic for her continual support and advice throughout my degree. I will miss Tuesday coffeebreaks.To my current and past friends and colleagues at the Composite Research Network, Hubert Courteau-Godmaire, Caitlin Duffner, Scott Nesbitt, Cheng Chen, Gavin Tao, Shayan Fahimi, Shaghayegh Kiafer,Sahar Abouali, Mohammad Mohseni, Nasser Arbabi, Shayan Fahimi, Nima Bakhshi, Erfan Forghani,Margarita Galper, Dr. Andrew Stewart, Dr. Janna Fabris, Dr. Kamyar Gordnian, and John Park, thankyou and welcome. Your friendship and fruitful discussions have made my time at UBC an enjoyableone. I would especially like to give a shout out to those students who began their degrees with me inSeptember 2016, Hubert Courteau-Godmaire, Caitlin Duffner, and Cheng Chen. I would also like toacknowledge Dr. Johannes Reiner and Professor Martine Dube as well as visiting students JohannesWiedemann, Alexandre Looten, Manish Nagaraj, Yuta Kumagai, Vincent Gill, Oliver Herb, and DavidZimmerman. It was a pleasure to get to know you all and I hope to continue our friendships. To allcurrent CRN students, I wish you all the best in the future.Thank you to my funding sources, NSERC-CGSM. The funding received helped greatly offset thexxvcost of living in Vancouver while completing a Master’s Degree.To all my friends I made while at UBC, thank you for the good times. From spending many daystogether in the mountains to pub trivia and heated board games matches, I enjoyed my time here andhope to continue seeing all of you around. I would especially like to thank Hubert Courteau-Godmaire,Stephen Ji, and Emma Kazic for being unreal roommates and all the huge plays.I would also like to thank my parents, Mark and Michelle Pawson for their unwavering support andinspiring me to follow my passions. To my siblings, Kail, Danica, and Kenny Pawson, suhhh duh.Finally, I would like to thank my partner Valentina Fazio for challenging me and helping keep thingsin perspective. I am beyond grateful for your continuing devotion and support.xxviChapter 1Introduction1.1 Composites and sandwich panelsA composite material is a multi-constituent material, wherein each component has differing physicaland chemical characteristics. Typically a composite is comprised of a stiff, load-bearing component(reinforcement) coupled with a surrounding matrix material for force distribution. In advanced compos-ite materials such as Glass and Carbon Fiber Reinforced Polymers (GFRP and CFRP respectively), thefibers act as the load-bearing component and the polymer as the matrix. A major benefit of compositesand particularly CFRPs is that they offer exceptional specific strength characteristics. That is, they arestrong and light-weight. This is advantageous in design, particularly in aerospace applications. Fur-thermore, the anisotropic nature of composites allows for tailoring of mechanical design to account forspecific load paths. This lends itself to the importance of fiber architecture in composites for structuralperformance. There are a wide array of possible fiber orientations, including unidirectional or wovenfabrics, which must be carefully chosen to suit the intended application.Advanced composites are manufactured in several ways. Such methods include Liquid CompositeMolding (LCM) processes whereby fibers are pre-placed in a mold and resin is injected into the mold viaa pressure gradient and allowed to cure. Another common method is the use of fibers pre-impregnatedwith resin, known as prepreg, developed through a hot melt or solvent dip process. Prepreg manufactur-ing involves placing each individual layer on a tool which is then typically placed in an oven or autoclaveto cure. This layer-by-layer construction is known as“laying up”. Due to the anisotropy of composites,the pattern of layup greatly affects the mechanical performance of the final part and must be considered.Polymer composites can be broken into two major categories, thermosets and thermoplastics, withthe former being more common for prepreg applications. Thermosets begin in a viscous liquid stateand undergo an irreversible transition through to a solid upon curing. Curing is achieved by subjectingthe part to heat in order to allow for chemical rearrangement of the polymer chains into a cross-linkedstructure. Typically, the curing process is performed in an autoclave with the part placed under a vacuumbag to allow for external pressurization. This is done to achieve necessary consolidation of the part andprevent defects such as porosity and poor resin distribution.1Two common types of composite structures that are used in design are laminates and sandwichpanels. The former refers to a simple stack up of the composite in question (i.e. CFRP or GFRP)whereas the latter adds in an additional component, a core material. Sandwich panels typically consist oftwo constituent materials, a low-density core sandwiched between stiff under and overlaying facesheetsor skins. This construction allows for increased bending stiffness without compromising on weight. Theprinciple is similar to an I-beam, wherein the core represents the web and the facesheets represent theflanges. The additional thickness offered by the core increases the moment of inertia, with the locationsof maximum bending moment (i.e the top and bottom surface) reinforced by the stiff facesheets. Incomposite panels, the facesheets will typically be a GFRP or CFRP material. Metal face-sheets arecommon outside of composites [5]. The core is generally a foam or honeycomb design, again with awide range of materials used for each. The World War II military aircraft, the de Havilland Mosquitois often credited with being the first major structure to incorporate the use of sandwich panels [4, 5].Both its wings and fuselage were made of a sandwich construction using plywood skins and a balsawood core. Nowadays honeycomb core is becoming increasingly popular in sandwich panels, withboth aluminum and NOMEX R© used in engineering applications [4]. NOMEX R© is the brand name fora textile made from aramid fibers, first developed by DuPontTM in the 1960s [18]. Aerospace gradeNOMEX R© honeycomb core is manufactured from sheets of NOMEX R© paper adhesively bonded andthen expanded to give the honeycomb pattern. The expanded structure is coated with phenolic resin andthen cured to provide a stable shape.1.2 Core movementDuring processing of sandwich panels, the high pressure of the autoclave may actually result in defor-mation of the core in its weak, lateral direction. This problem is well known in the field of compositesand is often referred to as core crush. However, core movement is becoming a preferred term in industryso to prevent confusion with core crush in the through-thickness direction. The term core movementis used in this body of work, although descriptive words such as “crush” may be used to illustrate thedeformation process.Core movement is often extreme and visible by the naked eye as seen in Figure 1.1. Deformedpanels are irreparable and must be scrapped, making it one of the costliest manufacturing problems incomposites fabrication [28]. The problem is so pervasive that processing methods are often altered toprevent core movement. This may include processing the parts at lower pressures than a traditionallaminate. Moreover, specific tools and layup methods are used to mitigate the risk. Tie-downs are acommon procedure wherein specific plies are restrained from moving, so to prevent deformation of thecore itself (Figure 1.2). The idea being that if the core and immediate plies are well bonded, then byrestraining ply movement, core movement is also prevented. This method has proved successful and isused extensively by major aerospace companies such as The Boeing Company. The problem, however,is that implementing tie-downs increases tooling and labour cost as well as production time. Moreover,because the physics of core movement is not well understood, the use of tie-downs is empirically basedand not necessarily optimized.2ABFigure 1.1: (A) Sandwich panel after layup and prior to cure. (B) Same panel following cure;drastic core movement has occured.Much of the research on core movement focuses on the material aspect of the phenomena. That is,looking at the frictional resistance of various prepreg systems. However, there is a lack of fundamentalknowledge regarding the mechanisms involved in the core movement process. Moreover, the processingconditions under which the phenomena occurs are not well defined. As such, methods to mitigate coremovement are often based on empirical notions. There have been attempts to develop basic mechanicalmodels for core movement [8, 28], however, these have yet to be validated. As of now, whether or notcore movement will initiate during a given processing window remains largely unknown. This calls fora science based approach to understanding the driving factors involved in core movement during theprocessing of sandwich panels. By understanding the physics of the phenomenon, process maps can be361 cmFigure 1.2: Processed sandwich panel on tool. Grit strips surrounding the part are used to restrainplies (tie-downs). Indents in the panel are from pressure sensors.created to allow for prediction of core movement. This will provide manufacturers with the knowledgenecessary to design for core movement. That is, mitigation methods can be tailored towards the panelsand their processing window rather than employ a one-size-fits-all solution.4Chapter 2Literature Review2.1 Honeycomb mechanicsHexagonal honeycombs are the most common form used. The structural material can be a metal, poly-mer, or even ceramic. Polymer honeycombs, particularly NOMEX R© honeycomb, is of importance inthe context of this body of work and will be the focus of this section. Furthermore, as it pertains tocore movement, in-plane deformation due to compression will be discussed as opposed to out-of-planedeformation and deformation due to tension.2.1.1 TerminologyMany honeycombs are manufactured by expanding sheets of glued paper strips or ribbons. This givesdirectionality to the honeycomb structure as seen below in Figure 2.1. The in-plane dimensions of ahoneycomb structure refer to the X1-X2 plane. The out-of-plane or through-thickness dimension is theX3 direction as shown in the figure [22, 49]. The X2 direction is known as the ribbon (L) direction since itfollows the direction of the glued ribbons. Correspondingly, the X1 direction is known as the non-ribbon(W) direction. For the sake of consistency the terms L and W will be used to refer to the X2 and X1direction respectively. Regular honeyomb are expanded such that the unit honeycomb cell has equal sidelengths with the angle between each side being 120◦. Overexpanded (OX) honeycomb is also availablewherein the cellular structure is no longer regular, so to provide increased mechanical benefits in theW direction and improve formability [10]. Finally, honeycomb may also be manufactured by bondingcorrugated sheets as opposed to expanding glued sheets. This is common in aluminum honeycomb. Ahexagonal honeycomb unit cell is shown in Figure 2.2.Hexcel is a common supplier of honeycomb materials. Their aramid lineup is given the followingdesignation [25]:Material - Cell Size - DensityWhere, material refers to the honeycomb type, cell size refers to the size of the honeycomb unit cell (asshown in Figure 2.2), and density refers to the nominal density of the honeycomb material in pounds per5X1(Non-Ribbon Direction -W)X2(Ribbon Direction -L)X3Figure 2.1: Hexagonal honeycomb structure. Glue lines are present in the ribbon direction (L) asindicated by the red lines. The non-ribbon direction (W) is perpendicular in-plane to this.tcαℓhθA AB BC CD DE EF FG GH HJ JK KL LM MN NP PR RT T242423232222212120201919181817171616151514141313121211111010998877665544332211DRAWNCHK'DAPPV'DMFGQ.AUNLESS OTHERWISE SPECIFIED:DIMENSIONS ARE IN MILLIMETERSSURFACE FINISH:TOLERANCES:   LINEAR:   ANGULAR:FINISH: DEBURR AND BREAK SHARP EDGESNAME SIGNATURE DATEMATERIAL:DO NOT SCALE DRAWING REVISIONTITLE:DWG NO.SCALE:10:1 SHEET 1 OF 1A0WEIGHT: HoneyCombUnitCellFigure 2.2: Honeycomb unit cell, where ` represents length, h height, t thickness, c cell size, α theangle between each side, and θ the cell wall angle. For regular honeycomb structures h = `,α = 120◦, and θ = 30◦. The out-of-plane thickness is denoted as b.6cubic foot. An example of a NOMEX R© -based honeycomb product designation from Hexcel is shownbelow.E.g.: HRH-10 - 1/8 - 3.02.1.2 Relative densityRelative density is a highly important factor in determining the properties of a foam or honeycombmaterial [7, 22, 49]. It is defined as the ratio of cell density to that of the solid material, or simply:Relative Density =ρρs(2.1)Where, ρ is the density of the cellular structure and ρs is the density of the solid material from whichthe cell walls are comprised. Cell density is sometimes displayed with an asterisk [22].Relative density is a function of the cell wall thickness (t) and edge-lengths (` and h). It can, ingeneral, be written as follows:ρρs=t/`(h/`+2)2cosθ(h/`+ sinθ)(2.2)For regular honeycomb where h = ` and θ = 30◦, then:ρρs=2t√3`(2.3)Therefore, it can be said that honeycombs with thicker cell walls have higher relative density, whereashoneycombs with longer edges, and therefore a larger cell size exhibit lower relative density, as wouldbe expected.It should be noted that the expansion method of honeycomb manufacturing results in a doubling ofthe cell wall thickness for members of length h in Figure 2.2. This does not affect the in-plane Young’sModuli nor Poison’s ratios, however it does affect the in-plane shear modulus and relative density. Thatsaid, if the ratio of t/` is small (i.e. t << `), meaning the relative density is low, then it can be assumedthat thickness is constant amongst all cell walls [22]. This is assumed in the calculations that follow.2.1.3 Uniaxial loadingThe functional loading direction of honeycomb cores in sandwich panels is along their stiff, through-thickness direction (the X3 direction as shown in Figure 2.1) [49]. However, it is the relatively largein-plane, compressive loads that lead to the deformation of the cellular structure that is core movement.In-plane compression of honeycomb results in three distinct regions. They are, in order of progression,bending, collapse, and densification [7, 11, 22, 38, 39]. Bending constitutes a linear elastic regimewherein there is an initial constant rise in the stress strain diagram. Cellular collapse may be the result ofelastic buckling, formation of plastic hinges, or fracture of the cell walls. The manner in which collapseoccurs is material dependent. It is characterized by a plateau region on the stress-strain curve, whereindeformation occurs under relatively constant stress due to failure of the cellular structure. Finally,7Stress, σStrain, εI II IIIFigure 2.3: Typical compressive stress-strain curve for an elastomeric honeycomb. I: Bending -linear elastic bending of cell walls, II: Collapse (plateau region) - constant stress as cells fail,III: Densification - cell walls touching. Figure adapted from [22].Stress, σStrain, εIIIIII(ρ/ρs)instant = 0.5Figure 2.4: Effect of increasing nominal relative density on the compressive stress-strain responseof honeycombs. I: Bending - linear elastic bending of cell walls, II: Collapse (plateau region)- constant stress as cells fail, III: Densification - cell walls touching. Figure adapted from[22].densification occurs when opposing cell walls begin to touch. This is marked by an exponential risein stress, as the cells close up and the structure increases in density. The magnitude of each regime isdependent on the relative density. Higher relative densities result in a stiffer initial response, a shorter(or less severe) plateau section, and earlier onset of densification. A typical stress-strain curve foran elastomeric honeycomb is shown in Figure 2.3. Similarly, Figure 2.4 shows how the stress strainresponse is affected by an increase in the nominal relative density. It should be noted that relativedensity increases as the cellular structure is compressed, due to closing of the cell walls. In fact, theonset of densification is defined as the point when the instantaneous relative density becomes 0.5 [22].8NOMEX R© honeycomb has been shown to exhibit a similar stress-strain curve as that shown inFigure 2.3 under compression in both the ribbon and non-ribbon direction [23]. That said, the manner inwhich collapse occurs is different for loads applied in the ribbon versus non-ribbon direction. Initially,in both cases the cell members marked ` in Figure 2.2 bend while the L direction members (markedh in Figure 2.2) remain straight. Once the elastic limit is reached, loads applied in the L directionresult in a homogeneous collapse of the cellular structure by folding of the h members. In contrast,loads applied in the W direction result in a row-by-row cellular collapse pattern, wherein h members areforced together. The first row(s) to collapse appear arbitrary [23]. In the case of aluminum honeycombs,local row-by-row collapse occurs in both the ribbon and non-ribbon direction [38, 39]. In general, forelastic honeycombs, such as NOMEX R© or those with high hardening properties, homogeneous collapsedominates (at least in the ribbon direction). This gives a nonlinear, monotonically increasing stress-strain response [31, 38]. It should be noted that for NOMEX R© the slope of the plateau region is stillincreasing slightly. This effect is more obvious in the L direction due to uniform cell collapse as opposedto localized failure. The general shape of the stress-strain response of NOMEX R© is independent of strainrate for both L and W loading, although higher strain rates result in an overall stiffer response, especiallyin the L direction due to the aforementioned uniform failure pattern [23].2.1.4 Linear elastic bendingSimilar to relative density, Young’s modulus and shear modulus for honeycombs are given relative tothe modulus of the solid material from which the cell walls are constructed. Due to their anisotropy,honeycombs have four independent moduli (E1, E2, G12, ν12) and two plateau stresses (σ∗1 , σ∗2 ), relatingto loading in the W and L direction. However, for regular honeycombs σ∗1 = σ∗2 = σ∗ and E1 = E2 = E,resulting in only three independent moduli and one plateau stress [22]. For the purpose of being conciseand as it pertains to this body of work, only the equations for regular honeycombs will be discussed.The equations describing linear elastic bending of regular honeycombs and their importance are shownbelow [22]:Young’s Modulus:EEs= 2.3( t`)3(2.4)Since the Young’s modulus is only a function of t/`, just as relative density is, it can be said that aone-fold increase in relative density results in a cubic increase in compressive stiffness.Poisson’s Ratio:ν12 = ν21 = 1 (2.5)Equation 2.5 has important implications as it means that in-plane compression of honeycomb in onedirection results in an equivalent magnitude of expansion in the other in-plane direction. The through-9thickness (X3) direction does not experience any deformation, however. In a solid material, a Poisson’sRatio of 1 is impossible as the material behaves in a three-dimensional manner; deformation along anyaxis results in deformation along the other two axes. However, honeycomb behaves as a hinged, geo-metrical material, allowing it to act as a two dimensional structure in-plane.Shear Modulus:G12Es= 0.57( t`)3≈ E4Es(2.6)Similar to compressive stiffness, it can be said that a one-fold increase in relative density results ina cubic increase in shear stiffness.It should be noted that when polymeric honeycombs approach their glass transition temperature,Tg, the cell walls no longer bend in a linear-elastic fashion but rather undergo viscoelastic deformation[22]. In the case of NOMEX R© honeycomb the glass transition temperature occurs around 300◦C [9].Since this body of work only deals with NOMEX R© honeycomb, at a maximum processing temperatureof 180◦C, viscoelastic deformation of honeycomb is not discussed.2.1.5 Cell collapseFollowing bending, as mentioned above, there are three modes in which the cell structure may fail. Thatis, elastic buckling, plastic collapse, and brittle failure [22]. Elastic buckling and plastic collapse areshown below as they relate to polymers. Note that an asterisk denotes properties pertaining to that ofthe plateau region.Elastic buckling only occurs in the ribbon direction, since members h are aligned with the appliedload. That is to say that at a critical load (i.e. Pload = Pcrit), the members parallel to the loading directionwill buckle like an Euler column [11, 22]. This critical load is displayed as follows:Pcrit =n2ppi2EsIh2(2.7)Where, I = bt3/12 for a simple beam and np refers to the rotational stiffness of the cell nodeswhere the walls meet. Pload = 2σ2`bcosθ (where σ2 is the applied stress) and for regular hexagonalhoneycombs, np = 0.69 [22]. Therefore:(σ∗el)2Es= 0.22( t`)3(2.8)Combining Equation 2.8 and Equation 2.4 yields:(ε∗el)2 =110(2.9)Note that the subscript “2” in Equation 2.8 and Equation 2.9 denotes the loading direction, as buck-ling only occurs in the L (i.e. X2) direction.In the case of plastic collapse, plastic hinges form at the nodes, meaning further deformation of10the cell structure is unrecoverable. This occurs when the the bending moment reaches the fully plasticmoment, giving a plastic collapse stress of σ∗pl [22, 31].Once again, in the case of regular honeycombs, the plastic collapse stress is the same for both the Land W direction and is a function of the solid material yield stress. It is denoted as [22]:σ∗plσys=23( t`)2(2.10)Similarly, a shear load can be applied beyond the plastic limit, such that honeycomb will undergoplastic shear deformation. For regular honeycombs the plastic shear strength is denoted as [22]:(τ∗pl)12σys=12√3( t`)2(2.11)It is possible for both elastic buckling and plastic collapse to occur with the former proceeding thelatter. It has been shown that this occurs at a critical value of t/` (thereby relating to the relative densityof the material). For regular honeycombs this value is defined as follows [22]:( t`)crit= 3σysEs(2.12)2.1.6 Biaxial loadingLinear elasticity under biaxial loading follows much of the same response as that for uniaxial load-ing. Collapse however changes, as the two stresses now influence one another and the buckling modechanges. This complicates the calculations of elastic buckling and is therefore not shown. The patternof deformation seems to be a combination of that seen in uniaxial loading of the L and W direction [22].E and ν12 hold the same relation as those laid out in 2.4 and 2.5. However, axial deformation ofthe cell walls must now be considered. It was previously ignored, as bending was the predominantmode of deformation. Taking this into account and assuming that the two applied stresses are equal(σ1 = σ2 = σ ), then the strain response in the L and W direction are equivalent and can be written asfollows [22]:ε =√3σEs(t/`)(2.13)In the case of plastic collapse for a regular honeycomb where the applied stresses are equal, thefollowing relation can be obtained [22, 31]:σ∗plσ=43( t`)2(2.14)112.2 Inter-ply friction of prepregInter-ply friction of carbon/epoxy prepreg systems refers to the friction between plies in a stack up.Not to be confused with intra-ply friction, which refers to friction between tows within a single ply.When the inter-ply shear force exceeds that of the inter-ply friction, relative movement of plies willoccur. Generally speaking, there are two main modes of friction between articulating surfaces. That is,hydrodynamic friction and coulomb (or dry) friction. Hydrodynamic friction is present when a viscousfilm of appreciable thickness exists between two surfaces. It is a function of film thickness, viscosity, andsliding rate according to Equation 2.15. Coulomb friction is the classical friction description betweentwo surfaces and exists when no fluid interface is present. It is governed by the normal force andcoefficient of friction according to Equation 2.16.Hydrodynamic Friction :τhyd =ηdv (2.15)Where, τhyd = hydrodynamic shear stress, η = viscosity, d = film thickness, and v = relative speedbetween plates.Coloumb Friction :Ff = µN (2.16)Where Ff = coulomb friction force, µ = coefficient of friction, and N = normal force.Of course, it is possible for both modes to be present. If the film separating two surfaces is of molec-ular thickness, it is considered to be under boundary lubrication and follows a behaviour similar to thatof coloumb friction [34]. If both boundary (or pure coloumb friction) and hydrodynamic lubrication arepresent, it is referred to as mixed lubrication [34, 45]. In the case of prepreg there will be points wherethe two surfaces are in direct contact, separated only by a thin film, and areas where they are separatedby a film of considerable thickness, resulting in mixed lubrication. The Stribeck Curve provides a visualfor lubrication friction regimes, wherein the coefficient of friction is mapped against the Hershey Num-ber, as shown in Figure 2.5 [34, 44]. The curve allows for the governing friction mechanisms at differentprocessing conditions to be determined, wherein processing conditions are represented by the HersheyNumber. That is to say, depending on the friction coefficient at a given Hershey Number one can tellwhether they are in a state of boundary, mixed, or hydrodynamic lubrication. The Hershey Numberis a function of the viscosity, sliding rate (as presented above), and applied pressure (P), according toEquation 2.17. Note that v typically denotes rotational motion when referring to the Hershey Number.Hence, the Hershey Number is often presented as dimensionless.H =ηvP(2.17)Inter-ply friction between prepreg layers changes throughout the cure cycle. Initially, when there isa considerable film thickness, hydrodynamic friction dominates. As temperature is ramped up, viscosity12I II IIIFriction, log(μ)Hershey Number, HFigure 2.5: Stribeck Curve demonstrating three regions of lubrication. I: Boundary lubrication, II:Mixed lubrication, III: Hydrodynamic lubrication. Figure adapted from [34].decreases and inter-ply friction thus also decreases. However when the viscosity is low enough, thiseffect is no longer seen and, in fact, inter-ply friction is shown to increase [19, 20, 34, 35]. This increasein friction at low viscosity has been associated with an insufficient buildup of film thickness, allowingfor mechanical interaction between fibers. In other words, the friction response transitions from a hy-drodynamic regime to boundary lubrication, resulting in a coloumb dominated friction response at lowviscosity. However, some material systems seem apathetic to changes in viscosity at viscosities above250 Pa·s, with a friction behaviour that remains coloumbic in nature [34].It has also been shown that as normal force increases, friction initially decreases but eventuallystabilizes in accordance with boundary lubrication [30, 34]. This has been attributed to the reduction ofsurface roughness under pressure until the prepreg can no longer be consolidated further and roughnessremains constant.2.2.1 Pre-yield behaviourWhen the shear stress acting on a laminate exceeds the inter-ply frictional resistance, yielding occurs.That is, slip or relative movement between plies occur. It is this inter-ply slip that allows for the pro-gression of core movement. As such, pre-yield behaviour is of importance.Erland et al [19] describe inter-ply shear using a viscoelasto-plastic model, wherein the viscoelasticcontribution pertains to the pre-yield behaviour. The relation is given as:τint = Gplyγ+ηdγdt(2.18)Where τint = inter-ply shear stress, γ = shear strain, Gply = inter-ply shear modulus, anddγdt= shearstrain rate. Yield occurs when the inter-ply shear stress exceeds the critical yield stress (i.e. τint ≥ τc).The critical yield stress is given as a function of the coloumb and hydrodynamic friction, according toEquation 2.19.13τc = µσn+ j (2.19)Where σn = normal stress and j = shear joint strength. Shear joint strength is similar to tack andrepresents the hydrodynamic portion of the friction response. As mentioned earlier, inter-ply frictiondecreases upon heating to a minimum value after which it increases again. This is marked by a changein friction response from one dominated by hydrodynamic lubrication to one dominated by boundarylubrication. A similar response is shown when it comes to the critical yield stress [19]. That is, τc de-creases upon heating, reaches a minimum value, and then increases again slightly. Slip is most dominantwhen the critical yield stress is at its minimum. The changes in τc with temperature can be attributed tothe affects of µ and j. The latter, j, highest within the hydrodynamic regime, begins at a maxima anddecreases with an increase in temperature. Conversely, µ , highest during boundary lubrication, beginsat a minima and increases with an increase in temperature [19]. These contradictory affects are relatedto the amount of surface resin present [19, 28]. Initially, the surface resin layer is thick and resistanceto shear is dominated by properties of the resin matrix. As temperature is ramped up, the softened fric-tion response is due to a decrease in the shear modulus of the resin. With continual heating, the resinmigrates to the core of the plies [16, 43, 50], allowing for improved fiber intermingling and an increasein dry friction. Therefore, the shear response at low temperature processing is dominated by the resin,whereas at high temperature, the shear response is dominated by fiber-fiber contact. The magnitudesconstituting “high” versus “low” temperature processing are dependent on the material system [34].2.3 Core Movement2.3.1 Physics of core movementCore movement occurs when the lateral force acting on the panel exceeds the resistive forces of thepanel. The lateral force manifests through the horizontal component of pressure acting along the cham-fer edge of the panel. Brayden and Darrow [8] hypothesized that the resistive forces are a combinationof the frictional resistance of the prepreg, the compressive strength of the core and plies, and the forceexerted by entrapped gas filling the core volume. The latter variable was identified as the predominantfactor resisting core collapse. This model, however, wrongly assumes frictional resistance to be negli-gible without evidence supporting the notion. Alteneder et al [6] expanded this work by claiming that itis the difference in autoclave pressure and internal core pressure that is the driving force for core move-ment. They correlated the extent of core deformation with the level of in-core pressure and showed thatin-core pressure could be manipulated to delay or prevent the onset of core movement.In a multivariate design of experiment, Renn et al [42] further showed that manipulation of the in-ternal core pressure greatly influenced core movement. Vacuum pressure during layup was identified asthe primary factor responsible for dictating in-core pressure during autoclave processing. However, theresin system was found to be the most important factor influencing core movement resistance; although,they were unsure how exactly how.14Martin et al [35, 36] noted that in order for core movement to occur, plies must slip across aninterface. They developed a method for measuring frictional resistance of woven prepregs and showedthat inter-ply friction does in fact play a key role in dictating honeycomb core movement [35]. As such,the magnitude of friction is dictated by temperature according to the properties of inter-ply friction asdescribed above. In the hydrodynamic regime, viscosity dictates inter-ply friction. However, if littlesurface resin is present (i.e the film layer is thin), fiber intermingling occurs and boundary lubricationdominates. In a first-of-its-kind study, Martin et al [35] showed that both viscosity and surface resindistribution affect frictional resistance, which ultimately influences core movement. Moreover, in aseparate study [36], they hypothesized that the impregnation level of the prepreg should affect coremovement by influencing the resin distribution. It was shown that, for unidirectional fibers, fiber tensionwas the only impregnation condition that had any significant affect on core movement. Prepregs withlow-tensioned fibers displayed a higher concentration of fibers near the prepreg surface. This allowed forgreater mechanical interaction with adjacent plies and thus increased the inter-ply friction and reducedthe severity of core movement. Martin and Seferis [37] then further showed that resin systems could betailored to improve their frictional characteristics by adjusting the concentration of elastomer content.Doing so, would effectively alter the amount of surface resin present during cure.Pelton et al [40] developed a similar testing rig to quantify frictional values associated with coremovement for specific GFRP and CFRP systems. They showed that for carbon fabrics, fiber type, inaddition to the resin system, plays a large role in influencing friction and thus dictating core move-ment. Similarly, Hsiao et al [28] demonstrated that fiber architecture, plays a crucial role in determiningfrictional resistance and the extent of core movement.Hsiao et al [28] built upon the initial free body diagram laid out by Brayden and Darrow [8] toinclude frictional resistance at the ply level, as seen in Figure 2.6. This includes friction between thebottom ply and the tool (Fp−t), friction between plies (Fp−p), and friction between the top ply and the bag(Fp−b). In this model, the applied force is transferred through the plies by a series of shear interactions,wherein the plies nearest the core experience the greatest shear force. Assuming the core and prepregare well bonded, then in order for the core to deform, slippage must occur along a bagside and toolsideinterface [28, 36]. Hsiao et al [28] postulate that slippage occurs across the interfaces of least friction,wherein the prepreg-bag interaction (Fp−b) presents the highest friction and the prepreg-tool interaction(Fp−t) presents the lowest friction. The following expression can be made [28]:Fp−b > Fp−p > Fp−t (2.20)Therefore, bagside slippage is assumed to occur across a prepreg-prepreg interface (Fp−p), whereastoolside slippage occurs across the prepreg-tool interface.Hence, the condition for core movement becomes [28]:Fp−p+Fp−t +Fsti f f ness < Phorizontal ·Acore (2.21)Or more specifically:15PTool SideBag SideFtopFtopFtopFtopFtopFbtm FbtmFbtmFbtmFbtmPly 1Ply 2Ply 3Ply 4Ply 5Ply 6FstiffnessFtopFbtmFgNPPhorizontalPverticalFstiffnessCorePly 3 and 4Figure 2.6: FBD of sandwich panel demonstrating forces involved in core movement. Fsti f f nessincludes both the material stiffness of the skins and core in the lateral direction as well as thein-core pressure. Figure adapted from [28].Fp−p+Fp−t +Fsti f f ness < Psinθcore ·hcore (2.22)Where Fsti f f ness encompasses the material stiffness of the core and uncured plies, as well as the thein-core pressure. θcore represents the chamfer angle of the core and hcore is the height of the core withunit width into-the-page.This model represents the most up-to-date core movement model. That said, it isn’t without lim-itations. The manner in which ply slippage occurs according to Equation 2.20 is speculative. It alsoassumes this condition holds true for all material systems, which may not be the case considering theirfrictional values can be quite different [34, 35]. Furthermore, forces arising from in-core pressure plystiffness, and interface friction evolve during cure and are therefore transient. This makes the evaluationof such a condition as presented in Equation 2.22 quite challenging. It also adds a further conditionthat if the frictional forces are transient then the assumption laid out in Equation 2.20 must hold truethroughout the cure cycle for the model to be valid. These limitations have not yet been discussed orexplored.2.3.2 Mitigation methodsVarious methods to mitigate core movement have been suggested over the years. Simple methods in-clude reducing the core chamfer angle (20◦ is standard) and using cores with higher relative density16[10]. Owing to the severity of core movement, it is common for manufacturers to use a reduced pres-sure cycle of 310 kPa (45 psi) gauge [8, 10, 35]. This is in contrast to traditional laminates which aretypically processed around 621 kPa (90 psi) gauge. In addition to reducing the external pressure, man-ufacturers often implement tie-down plies [12, 26, 29, 37, 40, 41], wherein specific plies are restrainedfrom moving. This is one of the most common ways to prevent core movement. In this setup, generallyplies nearest the core are tied-down. The idea being that if ply movement is prevented, then so is coremovement - assuming the core does not deform relative to the over and underlying facesheets. Pliesare often restrained through the use of grit strips (sharp strips along the tool that plies are adhere to)[37, 41], although adhesives have also been used [26]. Restraining occurs along the plies’ outer edge,past the trim point of the panel.Another industrial technique to combat core movement includes mechanical stabilization of the core[10, 13, 17, 35]. Many variations on the process exist, with the idea being to increase rigidity of thecore. In one scenario [13, 17], a foam adhesive is applied to the core chamfer and a film adhesive isapplied along the flat underside of the core. The structure is then cured. The process, however, maytake several hours and must be done prior to the sandwich panel layup [13]. Alternatively, it is commonto implement a film adhesive layer only along the underside of the core and pre-cure it. Processes suchas these may also be used to improve rigidity for the purpose of machining and are often referred toas Core Machining Stabilization (CMS). Septumization of the core has also been used to mitigate coremovement [10, 37] by increasing core rigidity; although this technique has largely been used to improveacoustical dampening properties of the honeycomb. Adding a septum involves sectioning the core andbonding pre-cured plies (typically glass fiber) within the center of the core. Finally, while potting istypically used to allow for mechanical fastners to be attached to the sandwich panel, it has also beemployed as a means to strengthen the core in order to prevent core movement [10]. In the latter case,potting agent is added to select honeycomb cells and then cured to effectively increase the local coredensity.Altering the cure cycle to allow for a build up of in-core pressure has also been shown to be aneffective solution to managing to core movement [6, 8, 42]. The gas pressure entrapped within the coreduring cure increases with temperature according to the ideal gas law. Therefore, during the heatingphase of the cure cycle resistivity to core collapse increases (assuming pressure is kept constant). Hence,core movement is most likely to occur early in the cure cycle. As such, Brayden and Darrow [8] suggestkeeping the autoclave pressure low initially - 69 to 103 kPa (10-15 psi) gauge - and increase it to 310kPa once the temperature has reached a desired limit and in-core pressure has built up; typically the firsthold temperature. In this design, vacuum pressure is initially reduced and then vented as pressure israised. This allows for volatiles to escape during heating but prevents over pressurizing the panel onceexternal pressure is increased. Working off this theory, Alteneder et al [6] showed that you could inducean initial pressure state in the core which prevents core movement from occurring early in the cure cycle.This is achieved by early pressurization of the vacuum bag to match the autoclave pressure. As long asthe consolidation pressure is less than 125 kPa, the gas within the vacuum bag will permeate through theprepreg to the core. After some time, bag pressure is reduced to atmospheric conditions while autoclave17pressure increases. This effectively seals off the core by increasing the consolidation pressure acrossthe prepreg. Autoclave pressure can then be increased at a rate proportional to the expansion of gaswithin the core (during heating) such that the pressure differential between autoclave and core remainsrelatively low.Aside from early work on core movement, much of the research has focused on improving thematerial system. It has been shown that the both resin viscosity and surface resin distribution affect coremovement [28, 35, 37]. Although low viscosity systems may exhibit lower frictional resistance in thehydrodynamic regime, the resin can easily flow into interstitial zones, thus decreasing the amount ofsurface resin present and improving fiber intermingling. High viscosity systems present higher frictionin the hydrodynamic regime but don’t flow as well and are, therefore, typically coupled with highamounts of surface resin. As temperature increases, viscosity decreases and remaining surface resinmay, therefore, act as a lubricant. As such, it has been suggested to avoid prepreg systems exhibiting thecombination of high resin viscosity and high initial surface resin content [35]. Furthermore, reducingelastomer content in prepregs allows for a reduction in surface resin content during cure, leading to asubsequent increase in frictional resistance [37].Pelton et al further showed that by twisting fiber tows, one can significantly improve the frictionalresistance of the system [40]. Three kinds of tows were tested, namely, Never Twisted (NT), Un-Twisted (UT), and Standard Twist (ST). The UT tows represented tows that were twisted and then un-twisted. Both UT and ST tows demonstrated significant improvement in frictional resistance over NTtows, and eliminated or greatly reduced the severity of core movement. The use of ST tows has becomecommon practice in some industrial settings. Hsiao et al [28] also showed that prepreg with roundertows exhibit higher friction. Tow rounding is typically achieved by twisting or untwisting the tow. Itcan also be done by applying a chemical sizer or naturally occurs due to cross sectional conformationsof the fibers. Rounder tows result in larger undulations and thus have a rougher surface. This in-turncreates a more open fabric allowing for improved fiber interlocking. Moreover, rounder tows showed ahigher degree of impregnation, meaning less surface resin is present to act as a slip plane. Prepregs withvarying tow aspect ratios and thicknesses were used in their experiments, with ubiquitous agreementthat lower aspect ratio and greater thickness (i.e. a more round shape) hinders core movement. It wasalso shown that prepreg with larger tows (12k versus 3k) were more susceptible to movement, but couldbe mitigated through tow rounding. Patents have been filed in utilizing such methods to create crushresistant fabric and prepreg for composite sandwich panels [27].2.3.3 Gas flow within the coreWhile in-core pressure is important for reducing core movement, the application of vacuum will reduceentrapped gas unless specifically accounted for [8]. Honeycomb is generally considered impermeable inits in-plane dimension, if not perforated [46]. Therefore, gas must travel through or within the prepregto escape [33, 46–48]. Gas is vented from the core in two ways. Either through the bagside laminate inthe transverse (through-thickness) direction or out the edges of the panel in the in-plane direction. Thelatter occurs if edge breathing techniques are used and the former occurs if the release film is permeable18or perforated [33]. Depending on the permeability of the adhesive and the compaction pressure alongthe panel edgeband, it can take anywhere from minutes to hours for gas to flow in and out of the core[6, 33, 47, 48]. Kratz and Hubert [33] showed that gas transport out of the core can be broken into threemain events, namely: delay time, linear increase in air permeability, and final in-core pressure. Delaytime refers to the time until in-core pressure begins to drop. The permeability then increases linearlyas pressure continues to drop, until the pressure eventually stabilizes. The magnitude of each of theseregimes showed large variations amongst equivalent experiments. In addition, different weave patternsgreatly influenced permeability. Plain weave fabric showed the highest permeability in comparison withunidirectional and 5 harness satin weaves. The final in-core pressures of plain weave fabric panelsheld under vacuum for 24 hours varied between 5-20 kPa. It was also shown that permeability wassignificantly greater in-plane rather than in the through-thickness direction.Travares et al [47] demonstrated that the delay time is a result of the film adhesive being imper-meable prior to cure. Panels without film adhesive showed an immediate reduction of in-core pressureupon application of vacuum. Evacuation of in-core gas is significantly improved if the film adhesiveis perforated. Traveres et al [48] also showed that the optimal range of in-core pressure for skin-coreadhesion is 40-70 kPa. Similarly Kratz et al [32] showed that pressure fluctuations arise if the in-corepressure reaches 1 atm and may act to increase porosity. However, these experiments are for Vacuum-Bag Only (VBO) processing of sandwich panels. Therefore, they do not necessarily contrast the desirefor high in-core pressure to mitigate core movement during autoclave processing of sandwich panels.Both Taveres et al [46, 47] and Kratz et al [32, 33] used a similar setup to measure in-core pressureand related this to permeability. The setup involves the use of a cavity under the core connected toa pressure transducer. As a result, plies are only placed on the bagside of the core. The transducermeasures the pressure in the cavity prior to cure and as cure develops. This is synonymous with in-corepressure for this setup. However, without the presence of toolside plies, each cell of the honeycombcore is effectively connected. This behaves similar to a core with perforated cell walls. Thus a pressuregradient cannot develop across the honeycomb, as may be the case in a “real” sandwich panel.Alteneder et al [6] measured in-core pressure during autoclave processing of sandwich panels fora material system qualified to BMS8-256. In their experiments, bag-core pressure stabilization wasachieved in around 15 minutes. The exact time depended on the adhesive system implemented andthe compaction pressure. It was shown that the rate of gas flow across the panel edgeband achieveda steady-state value within 500 seconds for consolidation loads less than 125 kPa. Above 125 kPa ofcompaction pressure the prepreg became impermeable in a matter of minutes, trapping gas within thecore.19Chapter 3Aims and ObjectivesFactors influencing core movement and their relation to each other were identified and a core movementmap was created. Figure 3.1 provides an overview of the map. The layout is such that there are threemain categories that influence core movement, namely: materials, core geometry, and processing pa-rameters. Within these there are sub-categories that directly influence core movement. The map is notintended to provide an exhaustive list of all potential factors involved in core movement, but rather tosummarize the major components along with those of interest to this body of work.Much of the past research on core movement focuses on the materials aspect of the phenomenon. Inparticular, a great deal of work has been done on correlating the in-core pressure and frictional resistanceof the prepreg system to core movement. However, there is a lack of knowledge in understanding thephysical process of core movement. As a result, the impact of altering processing parameters in regardsto core movement is not well defined. This similarly extends to alterations in geometry and materials aswell.The objectives of this body of work can be broken into two broad categories. Namely, to betterunderstand the factors involved in core movement during processing and to construct an improved modelof the phenomenon based on the experimental findings. This is not an attempt to provide a completeanalysis of all potential factors in core movement, but rather provide clarity to the problem of coremovement itself as well as analyze the effect specific parameters have on core movement.The objectives of this research can be broken down as follows:1. Develop a better understanding of the factors influencing core movement during processing ofsandwich panels including:• Effect of temperature/viscosity• Effect of pressure• Effect of various layups2. Build a simple mechanical model based on experimental observations including:• Initiation/progession of core movement20 CORE MOVEMENT Processing Parameters Autoclave Temperature External Pressure Vacuum Tie Downs Core adjustments Core Machining Stabilization Perforated Cell Walls Septum Internal Temperature Internal Pressure  Gas flow Materials Consumables Surfacing Film Release Film Film Adhesive Honeycomb Core Relative density Material Prepreg  Material Fibers  Architecture Weave Pattern Twist Tow roundness Resin System Resin Distribution Flow Rate Viscosity Geometry Core Chamfer angle Cross Sectional Area Height Width Plies  Orientation* Moisture content*Figure 3.1: Core movement map. Major factors known to influence core movement are displayed.Green arrows indicate when one property influences another. Properties with an asterisk aresuspected to influence core movement, but their relation to core movement has not yet beendocumented nor are they investigated in this body of work.• Specific ply movement• Effect of processing conditionsThe aim of this research is to shed light on the core movement process so that manufacturers canimplement science-based methods to combat the problem. The primary questions that are posed andaddressed include:• How does core movement occur?– How/where does it initiate?– What are the sequence of events?• When in the processing cycle does core movement occur?21– Is it temperature dependent?– Does it occur below a certain viscosity?• Do current models capture everything?– What are the major forces at play?Other specific questions that are addressed include:• How effective are tie downs?• How does core machining stabilization affect core movement?• How does core type affect processing?– Chamfer angle– Ribbon vs non-ribbon direction22Chapter 4Methodology4.1 Experimental breakdownIn order to effectively answer the questions posed in Chapter 3, a design of experiment was performedand ten individual tests were conducted. They can be broken into the following three broad categories:1. Effect of TemperatureTemperature is increased to a set value, after which pressure is increased relatively slowly.Layup is equivalent among tests; only the set temperature differs.– Pressure at core movement initiation is recorded2. Effect of PressurePressure is increased rapidly to a set value, after which temperature is increased relativelyslowly. Layup is equivalent among tests; only the set pressure differs.– Temperature at core movement initiation is recorded3. Effect of LayupAspects of the layup are changed for each test. During processing, temperature is increasedto 120◦C and held constant. Thereafter, pressure is increased relatively slowly.– Pressure at core movement initiation is recordedIn each test, the material system is kept constant. A summary of the experiments is listed in Ta-ble 4.1. Further details regarding the testing procedure is provided in the sections that follow.23Table 4.1: Summary of experiments. Note that the 120◦C and “Standard” experiment are the sametest and represent the baseline test.1. Effect of Temperature 2. Effect of Pressure 3. Effect of LayupStandard layup Standard layup 120◦CSet temperature Set pressure Set layupIncrease pressure Increase temperature Increase pressure25◦C 325 kPa (47 psi) Standard95◦C 600 kPa (87 psi) Full single tie-down plies120◦C Half single tie-down plies180◦C Core machining stabilization45◦ core chamfer4.2 MaterialsThe prepreg used comprised of Cycom 970 epoxy resin with T300 3k fibers from Toray. The fibers areplain weave, standard twist, carbon fabric. The nominal resin content is 40% by weight and containsadded toughners. The material system is qualified to BMS 8-256. It is a low flow, high temperaturecure (177◦C) material and is non self-adhesive, meaning a film adhesive must be placed between theprepreg and core. The film adhesive used was Metlbond 1515-4 which is co-cured with the prepreg andhas the same Manufacturer Recommended Cure Cycle (MRCC). The core itself was HexWeb HRH-10-1/8-3.0 from Hexcel [25]. It is a regular hexagonal Nomex honeycomb core with a cell size of 3.175mm, density of 48.06 kg/m3, and approximate cell thickness of 0.08 mm. The chamfer angle of the corewas 20◦, except in the noted experiment where the chamfer angle was 45◦. Finally, Surface Master 905from Solvay was used as the surfacing film on the tool, to provide a good surface and improve resindistribution on the base of the panel. It is also co-cured with the prepreg and has the same MRCC. A 12mm thick, A36 steel tool was used for the experiments. The tool included AISI 321 stainless steel gritstrips adhesively bonded and riveted along the tool’s perimeter to allow for plies to be tied down.4.3 SensorsTwo DS1000SUH Linear Variable Displacement Transducers (LVDTs) were used to track deformationof the panel in-situ. These were coupled with a 5M30 AC LVDT Conditioner to provide a calibratedanalog output signal for displacement. The sensors were placed normal to the chamfer edge or radiusof the panels in the L and W direction. For the 120◦C and 45◦ chamfer experiments (described later),the LVDTs were placed along a single panel edge at the top and bottom of the chamfer in order to trackthe pattern of deformation of the chamfer itself. In each experiment, embedded pressure sensors weresurgically placed in the core at the chamfer radius to track in-core pressure fluctuations throughout theentire process. This included measuring pressure in the core following vacuum application, as wellas during core movement. Early experiments featured two pressure sensors. However, due to lack ofresources, later experiments only featured one pressure sensor. Both bagside and toolside locations weretested for pressure sensor placement with no differences seen. Finally, an autoclavable camera was setup24Figure 4.1: Test setup showing in-situ sensors. Top left is an LVDT sensor, top right is an in-corepressure sensor, and the bottom is the full setup in the autoclave with the camera mountedabove the panel.to allow for real-time visualization of the core movement process. Due to the size of the panels and fieldof view of the camera, the recording was restricted to one corner of the panel. Both the pressure sensorsand camera were developed by Convergent Manufacturing Technologies Inc. The sensor setup is shownin Figure 4.1. More information on the in-situ sensors and camera setup are provided in Appendix A.For the Dry Core test (described in Section 4.8) a GoPro was used to record the process, allowingfor the entire core to be captured. The same setup could not be implemented in tests involving heatingdue to operating restrictions of the GoPro.25LVDTCorePly 1Ply 12Surfacing FilmTool= Pressure Sensor= Film AdhesiveVacuum bagRelease flimBreather clothFigure 4.2: Standard sandwich panel layup. First layer against the tool is surfacing film, followedby 4 toolside plies. Core with pressure sensor and film adhesive layers is then added, fol-lowed by 4 filler plies. Next, 4 bagside plies are draped over the core. PTFE release film,breather, and vacuum bagging are then added. The LVDTs are placed after vacuum baggingis complete.4.4 LayupAs laid out in Table 4.1, the experiments consisted of three distinct categories, 1 - effect of temperature,2 - effect of pressure, and 3 - effect of layup. In category 1 and 2, the layup for each experiment wasidentical and is referred to as the standard layup. In category 3, however, the layup was varied for eachexperiment in order to study its effect on core movement.4.4.1 Standard layupFigure 4.2 provides an illustration of the general layup used. Each panel consisted of a layer of surfacingfilm, twelve individual prepreg plies, one core, and two layers of film adhesive. The surfacing film wasthe first layer against the tool, followed by four prepreg plies (plies 1-4). Film adhesive was thencentered on the laminate and the core was centered on the film adhesive. Pressure sensors were insertedin the core and then another layer of film adhesive was draped over the core. Four layers of filler plies(plies 5-8) were placed around the core in a pinwheel design, to allow for a smooth edgeband transition.The first two filler plies (plies 5-6) were butted against the core with the following two plies (plies 7-8)extending 6.35 mm and 12.7 mm up the core edge respectively. Four, full length prepreg plies (plies9-12) were draped over the whole structure. The film adhesive layers extended 12.7 mm past the coreedge. Ply 1 was held under vacuum for five minutes after being laid down to prevent shifting duringlayup.Figure 4.2 shows the pressure sensor on the bagside surface of the core. As mentioned, the toolsidelocation, directly under the radius of the core, was also tested with no observed difference of in-core26pressure as compared with the bagside location. For pressure sensors placed on the bagside surface, Asmall cutout in the upper film adhesive around the sensors were made to prevent intrusion of the filminto the sensors during cure. For both toolside and bagside locations, the walls of the honeycomb cellswere cut along the length of the pressure sensor cables to allow the cables to sit firmly in the core, thuspreventing shifting of the sensor itself and the allowing for plies to be laid flat.Once the layup was finished, Polytetrafluoroethylene (PTFE) release film was placed on the prepregand breather cloth surrounded the entire panel. The part was then vacuumed bagged with pleats extend-ing along the edgeband transitions to prevent bridging. Exact debulk time was not measured, but waslimited to a few minutes.As mentioned in Section 4.3, the 120◦C, standard layup and 45◦ chamfer experiment (describedbelow) featured both LVDTs on the same panel edge, with the second sensor positioned against the baseof the chamfer. This is not shown in Figure 4.2. Moreover, a 12.7 mm x 12.7 mm grid pattern wasdrawn on the breather cloth to track the rate of deformation in the video footage. These two panels werechosen as representative tests for investigating the rate of deformation in panels with varying chamferangles, namely 20◦ and 45◦.4.4.2 Specific layupsFour specific layups were tested to observe their effects on core movement. They include, full singletie-down plies, half single tie-down plies, CMS, and use of a core with a 45◦ chamfer angle. Thelayups follow the same gross procedure as for the standard layup, with minute differences in each. Thedifferences are described below.Full single tie-down plies:The term “single” refers to the number of plies above and below the core that are tied-down. Singletie-down, therefore, refers to a layup in which one toolside and one bagside ply are restrained. In thisexperiment, ply 4 and ply 9, as shown in Figure 4.2, are restrained. These represent the nearest fullplies (i.e. not filler plies) to the core. The plies are restrained by tacking the plies onto the grit stripssurrounding the tool. All four edges of each ply are restrained, hence the term “full tie-down”. Sinceply 4 is placed first, it extends 25.4 mm onto the grit strips. To allow ply 9 to have an equivalent “fresh”section of grit strip, it extends a further 25.4 mm past ply 4, or 50.8 mm total, onto the grit strip. Theremainder of the layup is identical to that described in Section 4.4.1.Half single tie-down plies:This layup follows the exact same procedure as that for full single tie-down plies, with one difference.Only two edges of plies 4 and 9 are restrained. Two adjacent edges are restrained such that a ribbon andnon-ribbon edge are restrained and the other ribbon and non-ribbon edges are free to move. A pressuresensor was placed on both a restrained and unrestrained non-ribbon edge.27A BFigure 4.3: Core machining stabilization - 76.2 mm wide picture frame film adhesive strip is curedalong the toolside of the core. (A) Toolside, (B) Bagside.Core Machining Stabilization (CMS):Prior to layup, CMS is performed (see Figure 4.3). A 76.2 mm wide picture frame strip of film adhesiveis cut out and placed around the toolside perimeter of the core. This is then cured under reduced vacuum(27.1 - 33.9 kPa) according to the MRCC, with no external pressure applied to prevent core movement.A layer of PTFE release film was placed under the film adhesive during cure in order to roughen upthe surface and provide a good bond during layup. The center of the core remained unstabilized or“neked”. Once stabilized, the core is then used in the layup as described in Section 4.4.1. During layup,film adhesive is still placed under the stabilized core, to provide the necessary bond with the underlyingprepreg facesheet.45◦ chamfer:In this scenario, the layup follows that of the standard layup described in Section 4.4.1. However, ahoneycomb core with a chamfer angle of 45◦ was used instead of one with a 20◦ chamfer angle.4.5 Cure cycleSimilar to layup, the cure cycle varied depending on the specific experiment. The general procedure wassuch that the part was heated until reaching a set temperature after which pressure was applied, with theexception of category 2 experiments according to Table 4.1. The cure cycle for category 3 experimentsis referred to as the standard cure cycle. Category 1 experiments largely follow this cure cycle with themain difference being that the hold temperature varies for each experiment. It should be noted that thecure cycles presented in 4.4 are different from the MRCC, as the purpose of these cycles are to isolatetheir effects in relation to core movement. RAVEN models [3] were used to calculate the resin viscosityof each sample throughout the cure cycle.284.5.1 Standard cure cycleHeating was applied at a rate of 5◦C/min, until the temperature reached 120◦C, after which pressureapplication began. External pressure was set to increase at a rate of 21 kPa/min, however the autoclavecontroller could not achieve this. As a result, a stepwise pressure cycle was introduced to attain a stableresponse. Autoclave pressure was ramped up in steps of 103 kPa (≈ 1 atm) up to 600 kPa (87 psi)gauge. After each step, pressure was held for one minute to artificially induce slow pressurization.Once maximum pressure was achieved, it was held for three hours while the part cured. At 103 kPagauge, the vacuum bag was vented to atmospheric pressure to prevent additional force acting on thepanel and reduce the chance of core movement occurring too early. 120◦C was chosen as the standardprocessing temperature as it is around this temperature when Cycom 970 reaches minimum viscosityunder the MRCC.4.5.2 Specific cure cyclesRoom temperature testA single test was conducted at room temperature to observe the effects on core movement when noheating is applied. This is listed as the 25◦C experiment under Table 4.1. Since this was the first testconducted, the pressure cycle was not altered to achieve slow pressurization. As a result, pressure ini-tially increased around two times faster than the standard cure cyle described above. Once the externalpressure reached 420 kPa gauge, however, the rate of pressurization slowed considerably. No heatingwas applied, however the temperature within the autoclave rose to approximately 30◦C due to pressur-ization. The remainder of the cure cycle followed the standard cure cycle outlined above.Effect of temperature:Three panels were built according to the standard layup. Each were processed at different temperatures,namely 95◦C, 120◦C, and 180◦C. The 120◦C sample followed the standard cure cycle described above.The other two samples also followed the same cure cycle but with different hold temperatures, i.e. 95◦Cand 180◦C respectively. In these tests, after one minute at maximum pressure, the hold temperature waschanged to 120◦C for final cure.The 180◦C test was chosen as that is similar to the hold temperature listed under the MRCC. More-over, 180◦C is representative of high temperature processing for sandwich panels, while 120◦C is in-dicative of low temperature processing. A test at 95◦C provides an additional data point outside thetypical processing window. For both 95◦C and 120◦C tests, the viscosity remains low after reaching itsrespective minimum; thereby maximizing the potential for core movement. Additionally, as shown inliterature, frictional resistance reaches a minimum before viscosity does. 120◦C represented the temper-ature at minimum viscosity under the MRCC. Therefore, it was hypothesized the frictional resistancecould be lower at 95◦C. The viscosity profiles of the three samples are shown in Figure 4.4. It shouldbe noted that at 180◦C, the sample begins to cure rapidly after reaching minimum viscosity.In an early attempt to prevent rapid pressurization, an experiment was performed at 120◦C following29180°C120°C95°CTimeTemperature/Viscosity180°C – viscosity curve120°C – viscosity curve95°C – viscosity curveFigure 4.4: Representative viscosity profile for three different temperatures during cure. Adaptedfrom RAVEN data received from M. Shead (Boeing Winnipeg).the standard layup. It is herein referred to as 120◦Corig. In this test, a pressure hold of one minute wasintroduced at an autoclave gauge pressure of 325 kPa. Rapid pressurization still occurred however,paving the way for the introduction of the standard cure cycle outlined in Section 4.5.1. Other thanextracting patterns in the deformation of the L and W edges, the results of this test are dismissed. Theyare presented in Appendix C.Effect of pressure:Two tests were conducted at different pressures, namely 325 kPa (47 psi) gauge and 600 kPa (87 psi)gauge. Pressure was applied rapidly (≈ 172 kPa/min) until the set pressure, after which heating began.Heating and the remainder of the cure cycle followed that outlined in Section 4.5.1. It should be notedthat rapid pressurization induced notable heat in the system prior to external heating.These tests are more indicative of standard sandwich panel processing in the sense that pressure isapplied before the part is heated. 325 kPa is typical for processing of sandwich panels, while 600 kPa isrepresentative of pressures used in the processing of laminates.4.6 In-situ dataData from the LVDT and pressure sensors were recorded real time using LabVIEW and later analyzedwith MATLAB. A typical output is shown in Figure 4.5. Note that only one LVDT and pressure sensor isshown for clarity. The sudden early rise in vacuum pressure corresponds to venting of the vacuum bag.This is followed by a rise of in-core pressure. Considerable noise is often seen in the pressure sensorsonce deformation is apparent due to pulling of the wires as the core moves inward. In-situ sensor dataand video footage were synchronized to allow for an understanding of the exact processing conditionsduring the entire core movement window.The test involving full tie downs displayed no visible core movement. As such, the deformationseen in this test was taken to represent deformation of the breather cloth. It was assumed this breatherdeformation was indicative of the breather deformation for all the tests. This allowed for a decoupling30Onset Collapse2.543.815.081.2708006004002000-200-400-600-80000:14:00Pressure (kPa)Displacement (mm)00:15:00 00:16:00 00:17:00 00:18:00 00:19:00Time (hh:mm:ss)LVDT deformation deviates from breather (orange) deformation at onset point (red circle). Slight compaction noticeable in video footage.Onset:Core/plymotion is continuous and obvious (black arrow).Collapse:Figure 4.5: Data output from in-situ sensors. The colour scheme is as follows: black - autoclavepressure, blue – LVDT sensor, purple – in-core pressure sensor, green - vacuum pressure.The region between onset and collapse represents the critical zone, wherein core movementinitiates.of core movement from breather displacement for each experiment.Core movement was broken into two stages, onset and collapse. These correspond to elastic bendingand collapse of the honeycomb structure, respectively, as shown in Figure 2.3. Onset was defined as thefirst point of observable movement in both the video footage and LVDT data. This is represented in theLVDT data as the location where the displacement pattern of the panel deviates from that of the breathercloth. This manifests either as an increase in rate of panel deformation compared to the breather clothor continued panel deformation while the breather deformation flatlines. In the video footage, the panelsurface becomes notably rougher. Collapse was defined as the point of obvious, continuous deformation.That is, it is undeniable from video footage and LVDT data that core movement is occurring. Followingcollapse, the deformation proceeds rapidly at a relatively constant rate, independent of pressure holds.The transition is marked by a sharp increase in slope in the LVDT data and movement of core and pliesare apparent in the video footage.The absolute net pressure acting on the panels at the moment of onset and collapse were determinedfor each sample. This was in turn correlated with the viscosity of each sample at the exact moment ofonset and collapse. This was done by matching the location on the viscosity profile, Figure 4.4, with theknown time at which core movement occurs from the in-situ data. The range of pressure between onsetand collapse is referred to as the “critical zone”, as it represents the range over which core movement31initiates. Pressures exceeding the upper bound of the critical zone are likely to result in drastic coremovement, whereas core movement may be prevented if pressure is kept below the lower bound.4.7 Post cure dataFollowing cure, panels were scanned using a Coordinate Measuring Machine (CMM) to obtain thesurface topography. MATLAB was used to analyze the generated Point Cloud Data files. From this, thechamfer angle and panel height after core movement could be obtained. Furthermore, thickness changesacross the samples were tracked to get an idea of ply movement. This was done for each sample. Thefull tie down test is represented as the control test, where no obvious ply movement is observed. Variousregions along the panels were identified and labeled. Specifically, the panel edge refers to the outsideedge of the panel, the core edge refers to the outside edge of the core at the base of the chamfer, and theedgeband refers to the plies immediately surrounding the core. It is representative of the region whereall plies are located and is the thickest region of the panel excluding the core. Figure 4.6 provides avisual reference. The entire region between the panel edge and core edge is herein referred to as thelaminate edge, as the plies in this area are effectively stacked as a laminate.In addition to scanning, the panels were sectioned through-thickness along the L and W directionusing a vertical bansaw with a diamond blade cutter. The cut samples were polished with sandpaper ofvarying grit. This allowed for individual plies along the laminate edge to be observed via microscopy.Several images of each sample were taken using a Keyence digital microscope and stitched togetheralong the entire laminate edge. From these images, the location of each ply was tracked, so that individ-ual ply movement could be determined as shown in Figure 4.7. It should be noted that individual pliesoften seem to disappear only to reappear later. This is due to plies moving in and out of the plane of thecut, either due to the cut not being perfectly straight or waviness associated with the tows themselves.In these cases, opposing tows (i.e. those in the out-of-the page direction) are used to help track the plies.Through-thickness cuts also allowed for specific features of interest to be identified and labeled, suchas the “crush zone”. The crush zone represents the region of the core that undergoes deformation duringcore movement. It is best observable from a through-thickness section cut as shown in Figure 4.6.Additionally, wrinkle patterns in the bagside and toolside plies were more obvious from a through-thickness perspective.Crush ZoneCore EdgePanel EdgeEdgebandBagside WrinklesFigure 4.6: Polished through-thickness section cut. The region between the panel edge and coreedge is referred to as the laminate edge.3257462318Surfacing Film910111210119876435Direction of pull Core EdgePanel EdgeFigure 4.7: Cross sectional image of laminate edge showing individual plies and where they ter-minate after core movement. Numbers on the right identify the individual plies while the redXs indicate where the associated ply terminates.Several in-plane section cuts were also made using a vertical bansaw with a diamond blade cutter.The cut sections were polished with sandpaper of varying grit to allow for the cell deformation patternto be observed. The difference in deformation pattern associated with compression in the L and Wdirection is of particular interest. Figure 4.8 provides a visual reference.33WLDeformation in WDeformation in LFigure 4.8: Polished in-plane section cut showing the deformation profiles in L and W.344.8 Dry core testA separate test was run where dry core (i.e. without plies) was placed in the autoclave. In this setup,breather cloth was only placed around the edges of the core. Release film was placed under the core toreduce friction effects with the tool. Other than the vacuum bag, nothing was placed on the bagside ofthe core to allow for visualization of the structure as it deformed. Only one LVDT was implemented asthe other failed.Pressure was applied according to the standard cure cycle outlined in Section 4.5.1. No heating wasapplied, however temperature rose to 38◦C due to pressurization. A GoPro was mounted to the ceilingof the autoclave to film the process.35Chapter 5Results and Discussion5.1 Effect of temperatureThe following section details the effect temperature has on core movement. This includes both theinitiation of core movement as well as the manner in which it progresses. Tests conducted include thoselisted under “Effect of Temperature” in Table 4.1.5.1.1 Failure pressureThe absolute autoclave pressure at the moment of onset and collapse (as defined in Figure 4.5) wasrecorded for each specimen. The associated vacuum pressure at these times was subtracted from theautoclave pressure to give the pressure differential across the vacuum bag. This is the net pressureacting on the panels and represents the driving force for core movement initiation.Figure 5.1(a) shows the net pressure at onset and collapse for each of the samples. The room tem-perature test displayed no core movement and is therefore not included in the figure. This indicates, thatsome level of heat application is required to cause core movement. Specifically, the external temperaturemust be greater than 25◦C to result in noticeable core movement.It is not temperature itself that is the root cause of core movement initiation. Rather, temperatureinfluences properties of the resin, such as viscosity and ability to flow, which in turn influence coremovement. Moreover, the length of time at which the sandwich panel is held at a given temperature alsoinfluences these properties. As such, the resin viscosity at onset and collapse is shown in Figure 5.1(b)to give a direct indication of the resin properties at the time of failure. It should be noted that fromthe time of pressure application through to initiation of core movement, the viscosity does not changefor the 95◦C and 120◦C samples. However, the degree of cure (DOC) advances quickly for the 180◦Csample, resulting in a rapid increase in viscosity. At the time of pressure application the 180◦C samplehad a viscosity of 12 Pa·s. A few minutes later, the viscosity rose to 68 Pa·s during the onset of coremovement, and was around 368 Pa·s by the time collapse occurred. These values are based on thenumerical simulations from the RAVEN software and are best visualized in Figure 4.4.The net pressure at onset is similar amongst the 95◦C and the 180◦C samples. The 95◦C sample361001201401601802002202402602803000 25 50 75 100 125 150 175 200Net pressure (kPa)Temperature (°C)OnsetCollapse95°C120°C180°C(a) Net pressure at failure for various processing temperatures1001201401601802002202402602803001 10 100 1000Net pressure (kPa)Viscosity (Pa s)120°C95°C180°C(b) Resin viscosity at failureFigure 5.1: Net pressure acting on the sandwich panels at onset and collapse for the given (a)processing temperature and (b) associated resin viscosity.displayed onset at 192 kPa and the 180◦C sample at 189 kPa. However, at 120◦C the pressure to causeonset drops slightly to 170 kPa. This difference becomes more apparent at collapse, where the 95◦C and180◦C samples fail at 265 and 269 kPa respectively, yet the 120◦C sample fails at 200 kPa. In additionto the lower failure pressure, the 120◦C sample shows a much tighter critical zone of 30 kPa. Pressurefluctuations are, therefore, of concern as even small deviations of pressure could push the operatingpressure from below the onset point up to collapse. In contrast, the critical zone of the 95◦C and 180◦Csamples are 73 and 80 kPa respectively. This is important for manufacturers as a larger critical zoneallows for less stringent operating conditions.The divergent behaviour of the 120◦C sample is associated with a drop in viscosity below 10 Pa·s(as seen in Figure 5.1(b)). Interestingly, the 95◦C and 180◦C samples display markedly similar results inregards to failure pressure despite significant differences in their viscosity. This suggests that viscosity37does not play as much of a role in resisting core movement at high temperatures; likely as a result of thesurface resin distribution.Since the prepreg system is constant amongst samples, the initial surface resin distribution is equiv-alent. Therefore, differences in surface resin distribution must come about during cure. Two factorscontributing to the amount of surface resin are the ability for the resin to flow out under pressure andthe rate at which the resin migrates from the surface to the core of the tows. The former is influencedby the resin viscosity while the latter is a product of the cure cycle. Since the 120◦C sample displayeda lower resin viscosity at failure, it is reasonable to assume the surface resin content would similarlybe reduced due to the ease of flow. However, this would act to increase the frictional resistance not de-crease it. Therefore, this is perhaps not the case. The resin viscosity actually reaches a minimum in the180◦C sample, although this occurs prior to the application of pressure. Therefore, it is possible that thesurface resin content of the 180◦C sample was reduced in comparison with the other two samples due toresin migration into the core of the tows. This may explain its higher resistance to crush in comparisonwith the 120◦C sample. It would also explain why viscosity has no apparent affect in regards to failurepressure between the 95◦C and 180◦C samples, because the 180◦C sample is in a state of boundary lu-brication dominated by dry, coloumb friction. The 95◦C sample on the other hand displayed the highestresin viscosity throughout heating and therefore likely also contained the highest surface resin content.In short, the sample remained in a state of hydrodynamic lubrication. Therefore, its relatively “high”viscosity increased its frictional resistance to crush. It is possible that the 95◦C sample exhibited hydro-dynamic friction equal in magnitude to a boundary lubrication mode presented by the 180◦C sample;hence their apparent similar resistance to core movement. The 120◦C sample would then lie in betweenthe others, i.e. considerable surface resin content coupled with a low resin viscosity resulting in reducedfrictional resistance.Figure 5.1(a) implies that resistance to core movement follows a similar trend to inter-ply friction.That is, early on friction decreases with viscosity but eventually fiber-fiber contact dominates the re-sponse and friction stabilizes or increases slightly. The transition to a coloumb dominated responsetypically occurs before minimum viscosity [35]. However, as seen in Figure 5.1(b), the lower failurepressure exhibited during the 120◦C test seemingly corresponds to minimum viscosity, indicating hy-drodynamic lubrication remains dominant up to this point. It is suspected, however, that beyond thispoint, viscosity would continue to decrease yet resistance to movement would in fact increase due tofiber intermingling. This would support the notion that the 180◦C response is dominated by coloumbfriction and thus match that of typical inter-ply friction results; therefore, suggesting that resistance tocore movement is proportional to the inter-ply friction of the prepreg.It should be noted however, that with a higher Degree of Cure (DOC) as exhibited in the 180◦Csample, the contribution of ply resistance to core movement increases. This increase in ply stiffnessmay, in part, be responsible for the increase in failure pressure between 120◦C and 180◦C. This isfurther elaborated in later sections.3895°C120°C 180°CWL25°CFigure 5.2: 25◦C, 95◦C, 120◦C, and 180◦C tests post-processing, showing the extent of core move-ment.Extent of core movementAlthough initiation of core movement occurs at similar levels of pressure for the 95◦C and 180◦C sam-ples, the extent of core movement varies drastically amongst tests. In the case of the 95◦C and 120◦Csamples, the extent of core movement is far greater than for the 180◦C sample as seen in Figure 5.2. Thisis due to advancement in the DOC during pressurization. In the case of the 180◦C test, the panel curesrapidly, effectively solidifying the panel and preventing further movement. Recall that temperature isincreased to the set point before pressure is applied. Therefore, for the 180◦C test, once pressuriza-tion occurs, not much time is allowed for crushing to take place before the cure is quite advanced. Atonset and collapse the DOC is approximately 0.5 and 0.6 respectively, based on RAVEN simulations.Therefore, gelation has likely already occurred by the time core movement begins for the 180◦C test.39The 95◦C and 120◦C samples, on the other hand, remain at low viscosity during pressurization andtheir DOC does not advance significantly. This allows for core movement to continue, rather than halt-ing shortly after initiation as in the 180◦C case. This is represented by the viscosity profiles given inFigure 4.4. As mentioned previously, the 25◦C experiment displays no core movement.5.1.2 Ply movementGross ply movementSurface topography CMM scans of through-thickness cuts as shown in Figure 4.6 allow for thicknesschanges across the panel surface to be tracked. As core movement progresses, plies are dragged inwardto different extents. Changes in thickness across the panel edge gives an indication of the gross move-ment of the plies. That is, the number of plies at a given location can be inferred from the laminate edgethickness. From this, the extent of ply movement can be deduced. Figure 5.3 provides representativecurves of an ideal sample, showing both no core movement and a core movement scenario. In the nocore movement scenario (Figure 5.3a), the thickness of the laminate edge remains constant. Moreover,the edgeband encompasses the entirety of the laminate edge as no ply movement occurs. In the coremovement case (Figure 5.3b), plies are pulled inward with the core, thus moving the edgeband. Individ-ual plies move to different extents creating a transition in thickness toward the panel edge as plies dropoff from the edgeband.Figure 5.4 shows the thickness changes across the laminate edge of the processed panels in theribbon and non-ribbon direction for the three samples (CMM scans of the room temperature samplewere not performed). A copy of Figure 4.6 is included as a reference of the scanned sections. Nearthe panel edge some noise is seen which can be disregarded. Measurements of thickness include thatof the paint layer, which is not necessarily even among samples. The magnitude of thickness should,therefore, be used as an indicator for the number of plies as opposed to an absolute value.The edgeband is represented by the relatively flat region at the end of each of the curves. It terminatesat the core edge as the thickness begins to rise steeply. In almost every case, excluding the no movementcase, the thickness decreases just before the core edge. This indicates a higher level of compactionnearest the core. Alternatively, it could indicate that the core is deforming relative to the plies; howeverthis was not observed when looking at the specific ply movement, as outlined in the next section.The “no movement” case displays the ideal thickness profile, taken from the full tie-down test. Thatis, the edgeband extends over the entire laminate edge of the panel with no change in thickness. A minordrop off in thickness is seen nearest the panel edge. This is, however, likely due to plies being slightlyoffset during layup. Prior to cure, the edgeband represents the entirety of the laminate edge. Followingcore movement, the plies then move and the edgeband shifts. The 180◦C sample displays far less plymovement than the other two samples. This is due to the fact that it experienced a lesser degree of coremovement as explained in Section 5.1.1. What is of particular interest, however, is the shape of thethickness curves.Both the 120◦C and 95◦C samples display a gradual increase in thickness until the edgeband transi-40Core EdgePanel EdgeEdgebandDistance from panel edgePanel thicknessEdgebandCore edge(a) No core movementCore EdgePanel EdgeEdgebandDistance from panel edgePanel thicknessEdgebandCore edge(b) Core movementFigure 5.3: Schematic showing the expected thickness changes along the laminate edge of a sand-wich panel that (a) experiences no core movement and (b) experiences core movement.4100.511.522.533.540 20 40 140 160 180Panel thickness (mm)60 80 100 120 Distance from panel edge (mm)180°C - L180°C - W95°C - L95°C - WNo movement120°C - L120°C - WEdgebandArtifact NoiseCore edgePanel edgeEdgebandCore edgeFigure 5.4: Change in thickness across laminate edge of processed panels in the L and W direc-tions at varying temperatures. The “No movement” case represents the test with tie downplies, as no visible core movement was seen.tion, whereas the 180◦C case transitions rapidly to the edgeband at some distance from the panel edge.This indicates that the lower temperature tests display more relative movement of plies. The gradualchange in thickness is the result of plies being offset from one another due to differences in the extentof individual ply movement. The 180◦C panel shows less relative movement amongst plies, but rather abulk movement of all plies. This is evident upon visual inspection of the cured panel (Figure 5.2), wherethe surfacing film is clearly seen up to the edgeband of the 180◦C sample, indicating all plies have slidon this layer. While the surfacing film is apparent in the other samples, it is not as clear and is stillmostly covered by carbon fabric. In fact, surfacing film was seen to seep through interstitial spacingsbetween tows, as shown in Figure 5.5. When few plies are present it can be seen from the surface of thelaminate edge.Specific ply movementWhile Figure 5.4 provides insight into the gross movement of plies, it does not indicate the individualplies that were pulled in with the core. As demonstrated in Figure 4.7, specific ply movement wastracked using digital microscopy. This is shown in Figure 5.6. Again, ply movement is displayedbetween the panel edge and core edge.42Surfacing filmPlySurfacing film seeping throughFigure 5.5: Microscope image of laminate edge showing surfacing film seeping through interstitialzones between tows.In each case the plies nearest the core (recall Figure 4.2) pull in the furthest, as would be expected.However, only in the 180◦C case does the first ply (i.e. the ply directly against the surfacing film)pull in. It does so in both the ribbon and non-ribbon direction. Moreover, in the 180◦C sample, themagnitude of individual ply movement does not differ greatly amongst plies. There is notable relativemovement between ply 1 and the other plies in the ribbon direction, however this is the only occurance ofconsiderable relative movement for the 180◦C case. In general, the pattern of movement is such that theplies shift as a unit. This supports the aforementioned notion that bulk motion of plies is favourable athigher temperatures. Similarly, as processing temperature is reduced, relative ply movement increasesbetween plies near the core edge and those further from the core edge. The 95◦C sample displaysconsiderable relative ply movement on the toolside, while on the bagside inter-ply slip is largely isolatedto the ply 11/12 interface. In the case of the 120◦C sample, the number of toolside interfaces presentinginter-ply slip is less, while on the bagside it is greater. In no case did the surfacing film move with theplies.In order for core movement to initiate, one bag-side and one-toolside ply must slip. If the plies aretreated as rigid bodies, slippage is dictated by the interfaces of least friction. It is impossible for pliesfurther from the core to move without their counterparts nearer the core also moving, as inter-ply shearoriginates from the core. On the other hand, it is entirely possible that plies near the core could slipwhile those further away remain stationary. If the weakest interface were between the prepreg and peelply (bag), then slippage of this layer would result in all bagside plies moving equivalently. Similarly, ifthe prepreg-surfacing film (tool) interaction represented the weakest toolside interface, then all toolsideplies would slip equivalently. Conversely, if a prepreg-prepreg interface exhibited the lowest frictionalresistance it would result in relative movement between the plies, where those nearest the core wouldexhibit the greatest movement. The problem can be thought of as deck of cards where the center of thedeck is being pulled. If the top and bottom of the deck are not well supported (i.e. low friction at thetool/bag interface), the whole deck may move. However, if the top and bottom are well supported (i.e.43Distance from panel edge (mm)0 20 40 60 80 100 120 140 160 1801234567891011120 20 40 60 80 100 120 140 160 1801234567891011120 20 40 60 80 100 120 140 160 1801234567891011120 20 40 60 80 100 120 140 160 1801234567891011120 20 40 60 80 100 120 140 160 1801234567891011120 20 40 60 80 100 120 140 160 18012345678910111295°C120°C180°CLWPly number Core edge before movementCore edge after movementPly movementFinal ply positionFigure 5.6: Specific ply movement profile of processed panels in L and W at varying temperatures.Initially, the core edge is 50.8 mm from the panel edge.high friction at the tool/bag interface), a cascading pattern will be observed with cards near the centerpulling further than those towards the exterior. The results suggest that high temperature processingpromotes the former scenario, whereas low temperature processing results in the latter case.This trend fits with the previous assertion that the 180◦C sample exhibited boundary lubricationat the time of core movement, while hydrodynamic friction is still prevalent in the lower temperaturesamples. In this scenario, coloumb friction dominates the 180◦C test. Inter-ply friction would thereforeincrease as a result of fiber intermingling. However, friction between the prepreg-bag and prepreg-toolinterface would likely remain unchanged, due to the lack of fiber interactions. It is, therefore, possiblethat inter-ply friction would increase above that of the prepreg-bag and prepreg-tool interactions. Hence,44yielding of a prepreg-prepreg interface would not have been the root cause for slip in the 180◦C case.Rather, the peel ply and surfacing film would have presented the weakest bagside and toolside interfaces,respectively, resulting in all plies sliding relative to these layers. This is opposite to that of the lowertemperature samples, where inter-ply friction would remain low due to an inability for fiber interlocking.Another possible explanation for the 180◦C movement pattern is that because this sample exhibited ahigher DOC at the time of pressure application, the resin would have been in a near glassy state. Theplies would have been effectively solidified as one mass, demoting the chance for inter-ply movement.At the same time, the friction coefficient between the tool and bagside interface would have decreasedas the plies lost their tack, thus promoting slippage across these layers.The above explanation assumes the plies do experience any strain. It is possible that the extent of plymovement is a combination of pure friction as well tension developed through deformation of the plies.In this case, slippage does not necessarily originate at the interface of least friction but also dependson the number of engaged plies at the interface in question. That is, slip is dictated by the magnitudeof friction of the interface as well as the extent of ply tension. If friction is considerably higher thanthe level of ply tension, then interfacial slip will occur across the interface of least friction regardless.Moreover, At high DOCs, as the plies become stiffer they are likely to slip rather than strain. Therefore,tensile resistance of the plies is only of concern at low DOCs. This explained in more detail later.It should also be noted that relative motion of plies increases hydrodynamic shear according toEquation 2.15. This is of particular concern for the 95◦C and 120◦C samples. It is possible that slip-page occurred across a single ply interface, after which shear between neighbouring layers increased,resulting in slippage of further layers. The result would be a cascading pattern of ply movement as seenin Figure 5.6. This may also explain the discrepancy in movement between ply 1 and the other pliesin the 180◦C ribbon direction test. It is possible that, although largely dominated by coloumb friction,hydrodynamic shear still contributed to inter-ply friction in 180◦C test. In this sense, all plies may haveinitially slid across the ply 1 interface, after which shear across ply 1 exceeded the frictional limit of theply 1-surfacing film interface, resulting in slippage across that layer as well.Relation to mechanical modelThe patterns exhibited in Figure 5.2 and Figure 5.4 indicate that the mechanical nature of ply motionduring core movement is influenced by temperature. The interfaces of slip change depending on theprocessing temperature. Therefore, cure cycle plays a significant role in dictating the manner in whichcrush occurs. Although, according to Figure 5.1, this does not necessarily influence resistance to coremovement. Nevertheless, current mechanical models do not address this. Regardless of failure pressure,it is important to understand the nature of ply movement as that may dictate core movement mitigationprocedures. For example, which plies to tie down.5.1.3 Deformation patternSurprisingly, in all three samples the ribbon direction exhibits more deformation than the non-ribbondirection. This is best seen in Figure 5.2 and Figure 5.4, with the magnitude of displacement presented45Table 5.1: Extent of deformation in the ribbon and non-ribbon direction associated with differentprocessing temperatures.Test Displacement in L (mm) Displacement in W (mm)95◦C 108 104120◦C 100 86180◦C 41 16Table 5.2: Difference in displacement between the ribbon and non-ribbon edges associated withdifferent processing temperatures.Test Difference in displacement between L and W edges (mm)95◦C 4120◦C 14180◦C 25in Table 5.1. Moreover, the discrepancy in magnitude is more apparent the higher the processing tem-perature, with the 95◦C sample displaying only minor differences in deformation. This is displayedin Table 5.2. The values provided are measured from a single edge. The total discrepancy in ribbonversus non-ribbon deformation would, therefore, be twice the amount shown, assuming the deformationpattern is symmetric.The section cuts used to obtain the individual ply movement profiles were not always the exact centerof the panel. Extent of crush was, however, directly measured along the center line of the panels. As aresult, the extent of core movement displayed in Figure 5.6 may not align exactly with that provided inTable 5.1.Interestingly, in-situ data indicates onset of failure occurs in both directions simultaneously. Al-though, at higher processing temperatures, crush in the ribbon direction progresses at a notably fasterrate following collapse. This is shown in Figure 5.7. For the 95◦C sample, the two LVDTs measurea similar rate of deformation. However, for the 180◦C sample, the LVDT in the L direction deformsat a quicker rate than its counterpart in the W direction. Unfortunately, data on the rate of progressionof core movement in the L and W directions is not available for the 120◦C sample, since LVDTs wereplaced on equivalent sides of the panel for this test. However, a similar pattern to the 180◦C sample wasobserved in the 120◦Corig experiment (refer to Section 4.5.2 for a description of the test). The resultsof this test are provided in Appendix C. It should be noted that the LVDTs do not necessarily capturethe full extent of core movement as they are limited by their stroke length. Hence, when the LVDTsindicate no further movement is occurring, this may not actually be the case.The difference in magnitude between crush in the ribbon versus non-ribbon direction may be at-tributed to the manner in which cell collapse occurs in the two directions. Recall that under uniaxialloading in the L direction, NOMEX R© honeycomb displays homogenous collapse of the cellular struc-ture by buckling of the cell walls along the load path. Conversely, loading in the W direction results ina row-by-row collapse pattern. The homogenous buckling of cells in the L direction may present itselfas a quicker or preferential mode of deformation once collapse initiates. Hence, for high temperature4600:15:00 00:18:00 00:21:00 00:24:00Time (hh:mm:ss)0100200300400500600700800Autoclave Pressure (kPa)-25-20-15-10-50Displacement (mm)Autoclave PressureLVDT 1 - WLVDT 2 - LNoiseCollapse(a) 95◦C00:24:00 00:27:00 00:30:00Time (hh:mm:ss)0100200300400500600700800Autoclave Pressure (kPa)-25-20-15-10-50Displacement (mm)CollapseAutoclave PressureLVDT 1 - WLVDT 2 - L(b) 180◦CFigure 5.7: Core movement in ribbon and non-ribbon directions for (a) low versus (b) high tem-perature processing. Autoclave pressure is absolute external pressure. Vacuum pressure isnot shown.47processing, where the period of core movement is short due to the rapidly advancing DOC, deforma-tion in the W direction does not have time to progress significantly. Conversely, for low temperatureprocessing, the extended core movement window allows for the slower W deformation to catch up inmagnitude to the deformation exhibited in the L direction.485.2 Effect of pressureIt is beneficial to understand the range of pressure necessary to cause core movement at a given temper-ature. However, this begs the question, for a given pressure at what temperature does core movementoccur? This is more indicative of a “real-world” processing scenario, as often pressure is set and tem-perature is then increased according to a specific cure cycle. The following section aims to answer thisquestion. Tests conducted include those listed under “Effect of Pressure” in Table 4.1.5.2.1 Failure temperatureThe two experiments conducted are representative of typical processing pressures seen in processing oflaminates and sandwich panels. The former pressure being 600 kPa gauge and the latter 325 kPa gauge.Recall that net pressure is the absolute external pressure minus the pressure within the vacuum bag.Failure of the panels occurred after venting of the vacuum bag (i.e. pressure within the bag was 1 atm),therefore the net pressure acting on the panels is also 600 and 325 kPa. Onset of the 600 kPa sample,however, occurred while pressure was still increasing due to pressure-induced heating. That said, forsimplicity, the two samples are referred to by their set processing pressures, 325 and 600 kPa.The temperature that the two samples failed at is displayed in Figure 5.8a. For the 325 kPa sam-ple, onset and collapse occurred at 54.6◦C and 82◦C respectively. The 600 kPa sample showed slightlydifferent results as onset occurred before maximum pressure was achieved. As pressure increases, tem-perature also increases according to the ideal gas law. In this case, the temperature reached a criticalpoint sufficient to result in the onset of core movement. This occurred at a net pressure and tempera-ture of 497 kPa and 34.2◦C. Collapse occurred later after the set pressure (600 kPa) was achieved andconvection heating began. This occurred at a temperature of 42.6◦C. The fact that onset occurred nearroom temperature and without any external heating is testament to the necessity for strict processingconditions in order to prevent core movement. The room temperature sample presented in the previoussection displayed no core movement, yet underwent a processing cycle that was, initially, very similarto that of 600 kPa sample. That is, pressure was rapidly increased to 600 kPa with no external heatingapplied during this step. In the room temperature sample however, the maximum temperature withinthe autoclave was only 30◦C due to pressurization and remained for a relatively short period of time -approximately two minutes. The 600 kPa sample presented here, had been held above 30◦C for aroundfour minutes before onset was observed. This further demonstrates the stringent tolerances involved incore movement, as an apparent 4◦C rise in temperature held over a longer time period can be the differ-ence as to whether or not core movement initiates. Moreover, the room temperature sample had actuallyreached a higher pressure (600 kPa) without core movement initiating. Here, however, the sample dis-played onset at a lower pressure (497 kPa) due to a slight increase in temperature. This stresses theimportance of hydrodynamic friction in resisting core movement during the early stages of processing.If the resin viscosity remains high enough during the hydrodynamic regime, then core movement willnot initiate for pressures up to 600 kPa. However, slight increases in temperature will decrease the resinviscosity which may tip the scales in favour of core movement as shown here.Figure 5.8b shows the resin viscosity at failure for the two samples. The values presented are4901002003004005006007000 10 20 30 40 50 60 70 80 90Net pressure (kPa)Temperature (°C)OnsetCollapse600 kPa325 kPa(a) Temperature at failure for various processing pressures01002003004005006007000 200 400 600 800 1000 1200Net Pressure (kPa)Viscosity (Pa·s)600 kPa325 kPa(b) Resin viscosity at failureFigure 5.8: (a) Temperature at which onset and collapse occur for sandwich panels processed at aspecific pressure and (b) the associated resin viscosity at failure.50estimates based off RAVEN viscosity curves as demonstrated in Figure 4.4. The initial pressure-inducedheating occurs at an approximate rate of 3◦C per minute, whereas the RAVEN viscosity curves arebased off a 5◦C per minute heating cycle. Further, as the pressure approaches the set value, the rateof pressurization slows, thus decreasing the rate of pressure-induced heating. As such, the viscosityvalues are approximates. In this set of experiments viscosity decreases between onset and collapse.This is because onset and collapse occur during the heatup stage, where gelation has not yet occurred.Therefore, as temperature is increased, viscosity decreases. It is likely the friction regime remainshydrodynamic throughout the core movement process for both samples.The resin viscosity at onset and collapse for the 325 kPa sample is approximately 165 and 30 Pa·srespectively. Viscosity at onset and collapse for the 600 kPa sample is much higher, approximately966 and 404 Pa·s. This is in line with the expected response. At higher pressure, the temperature,and associated resin viscosity, necessary to allow for core movement is reduced. Interestingly, thetemperature to cause core movement decreases linearly with an increase in pressure. Onset at 325 kPaoccurs at a temperature of 55◦C and onset at 497 kPa occurs at a temperature of 34◦C. Similarly, collapseat 325 kPa occurs at a temperature of 82◦C and collapse at 600 kPa occurs at a temperature of 43◦C.In both scenarios the temperature at failure is inversely proportional to the increase in pressure. Asevidenced through the room temperature sample, however, there is a minimum temperature requirementto cause core movement for the range of pressures discussed. Following the above relation, at a pressureof 600 kPa, one would assume core movement should initiate above a temperature of 29◦C. That wasnot observed to be the case for the room temperature sample though, where net pressure was 600 kPaand the autoclave temperature reached 30◦C.In these set of experiments, the critical zone is no longer a range of pressures but rather a range oftemperatures. At a processing pressure of 325 kPa the critical zone is 27.4◦C, whereas at a processingpressure of 600 kPa the critical zone is reduced to 8.4◦C. Since onset in the latter case actually occurredat a pressure around 500 kPa, it is expected that the true critical zone at 600 kPa is even less, assuming theminimum temperature requirement is satisfied. Therefore, at high pressures, temperature fluctuationswithin the autoclave may bear significant consequences with regards to core movement.5.2.2 Ply movementGross ply movementJust as in Section 5.1 panel thickness across the laminate edge was measured in order to gain an under-standing of the scale of ply movement. The results are displayed in Figure 5.9. Recall Figure 5.3 wheninterpreting the figure. Note also that the magnitude of thickness includes that of the paint layer. As aresult, some error is introduced.The pattern of movement among the two samples is fairly similar with the major difference lyingin the extent of movement. At higher pressure, core movement is more severe as expected. In bothcases, there is a gradual increase in panel thickness until the edgeband. This indicates that in the twoexperiments, plies are sliding relative to one another resulting in plies dropping off at different locations5100.511.522.533.540 20 40 60 80 100 120 140 160 180Panel thickness (mm)Distance from panel edge (mm)No movement600 kPa - L600 kPa - W325 kPa - L325 kPa - WEdgebandArtifact NoiseCore EdgePanel EdgeEdgebandCore EdgeFigure 5.9: Change in thickness across laminate edge of processed panels in the L and W direc-tions at varying pressures. The “No movement” case represents the test with tie down plies,as no visible core movement was seen.along the laminate edge. In the previous section, low temperature processing showed similar results.Since movement occurs during the heating phase, resin has likely not infiltrated the core of the tows,meaning a considerable amount of surface resin is present. As a result, the friction mode would behydrodynamic in nature. Hence, the frictional resistance to crush would be dominated by properties ofthe resin. Ply drop offs are, therefore, expected. The shear force between plies increases as neighbouringplies slip, according to Equation 2.15. Once plies nearest the core begin to slip, additional plies wouldfollow thereafter, resulting in a cascading pattern.Specific ply movementThe individual movement of each ply is presented in Figure 5.10. For the most part, the pattern ofmovement among the high and low pressure samples are similar; with only the extent of movementdiffering. Moreover, considerable relative ply movement is seen which matches that of a hydrodynamicfriction regime. For the 600 kPa sample, ply 1 moves slightly in the L and W directions. The only otheroccurrence of this happening is in the 180◦C sample. Unlike in the 180◦C test, however, the plies donot move as a bulk. It is therefore likely that with a predominant hydrodynamic friction regime and arelatively high viscosity at failure (Figure 5.8(b)), the hydrodynamic shear stress acting on ply 1 was520 20 40 60 80 100 120 140 160 1801234567891011120 20 40 60 80 100 120 140 160 1801234567891011120 20 40 60 80 100 120 140 160 1801234567891011120 20 40 60 80 100 120 140 160 180123456789101112Distance from panel edge (mm)325 kPa600 kPaLWPly number Core edge before movementCore edge after movementPly movementFinal ply positionFigure 5.10: Specific ply movement profile of processed panels in L and W at varying pressures.Initially, the core edge is 50.8 mm from the panel edge.sufficient to result in slippage across the ply 1-surfacing film interface. The same phenomenon is notobserved in the lower pressure experiment; likely due to the decrease in applied pressure and subsequentreduction in resin viscosity at failure. While the measured deformation of ply 1 in the L direction samplefor the 600 kPa test was around 6 mm, the opposing L edge showed an approximate deformation of 16mm for ply 1. The two W edges however, showed similar levels of displacement for ply 1. Generallyspeaking the deformation pattern is fairly symmetric amongst equivalent edges, however some varianceis occasionally seen.In a sense, because the friction mode does not change throughout the time period over which coremovement occurs, managing core movement becomes simpler. Higher processing pressures result in areduction of temperature necessary for core movement initiation. The mechanical progression of coremovement does not change however. Since most cure cycles involve rapid pressurization coupled withslower heating, this represents the general relation for most situations in which core movement wouldoccur. Although high temperature processing affects the mechanical nature of core movement (presentedin the previous section), if heating is simultaneously performed with pressurization, then core movementis likely to occur before the processing temperature is achieved. Thus, the effects of high temperatureprocessing can be negated in this scenario.53325 kPa 600 kPaWLFigure 5.11: 325 and 600 kPa tests post-processing, showing the extent of core movement.5.2.3 Extent of core movementAs before, deformation in the ribbon direction is more extensive than in the non-ribbon direction for the325 kPa sample. However, for the 600 kPa sample, deformation in the non-ribbon direction is slightlygreater than in the ribbon direction. This is the only instance of this occurring among the experimentsconducted. Images of the two samples, post-processing, are shown in Figure 5.11 with their respectivemagnitudes of displacement presented in Table 5.3. Recall that extent of crush is measured directly alongthe center line of the panels and do not necessarily match the core position presented in Figure 5.10.Near a given edge, the collapse pattern is largely uniaxial in nature. That is, cells fail according tothe ribbon or non-ribbon collapse mode (i.e. buckling versus plastic bending). However, towards thecenter of the panel, as adjacent edges are pushed together, the collapse pattern takes on a biaxial state.A combination of buckling and plastic bending is seen. It is possible that the reason deformation in theL direction is higher for all samples except one, is that homogeneous, buckling is a favourable failuremode and therefore will usually dominate the biaxial response. However, as suggested by the 600 kPasample, plastic bending may “win” the response on occasion, allowing for higher deformation in the Wdirection. This is further shown in Section 5.5.2.Interestingly, despite showing greater overall displacement in the W direction, the 600 kPa sampleexhibited a slightly faster rate of displacement in the L direction initially. This is shown in Figure 5.12.54Table 5.3: Extent of deformation in the ribbon and non-ribbon direction associated with differentprocessing pressures.Test Displacement in L (mm) Displacement in W (mm)325 kPa 78 52600 kPa 85 9400:05:00 00:10:00 00:15:00 00:20:00Time (hh:mm:ss)-400-2000200400600Pressure, Temperature (kPa, °C)2520151050Displacement (mm)TemperatureAutoclave PressureLVDT 1 - WLVDT 2 - LCollapseOnsetFigure 5.12: Core movement in ribbon and non-ribbon direction for high pressure (600 kPa) pro-cessing. Autoclave pressure is absolute external pressure. Vacuum pressure is not shown.Recall that the LVDTs only capture the initial stages of core movement before their stroke lengths arereached. This indicates that the initial rate of deformation does not necessarily correspond to the finalmagnitude of deformation. Rather, the rate of progression of the crush front is important as the crushzone that projects further into the core quicker, will likely dominate the failure mode. In this case, theW collapse front “won” the biaxial response. This is explained in Section 5.5.2. Unfortunately, rate ofprogression of the two edges is not available for the 325 kPa sample, as one of the LVDTs failed.555.3 Effect of layupThe following section details the effect various layup features have on core movement. This includesboth the initiation of core movement as well as the manner in which it progresses. Tests conductedinclude those listed under “Effect of Layup” in Table 4.1.5.3.1 Failure pressureSimilar to Section 5.1.1 the net pressure at onset and collapse was determined for each of the experi-ments conducted. The results are presented in Figure 5.14. Unlike the previous tests, viscosity does notvary among samples since all tests were conducted at the same processing temperature of 120◦C. Thefull tie down test displayed no notable core movement and so is not presented in Figure 5.14. That said,minor distortions in the cellular pattern were seen upon sectioning the panel (Figure 5.13). However,these were not externally visible and likely were not picked up by the LVDTs.Not surprisingly, restraining only two edges (i.e. half tie downs) of the plies allows for movementalong the unrestrained edges while, at the same time, restricting movement along the restrained edges.Interestingly, the resistance to movement increases along the unrestrained edges in comparison with thestandard layup, indicating that core movement across adjacent edges are somewhat coupled. The criticalzone also increases in size, providing a larger buffer between onset and collapse. Onset and collapseoccur at 183 and 252 kPa respectively for half tie downs, as opposed to 170 and 200 kPa for the standardlayup.CMS slightly increases the resistance to core movement, with onset and collapse occurring at 189and 225 kPa respectively. The critical zone remains similar in size however. It was observed fromvideo footage that crush initiates beyond the stabilized zone which extends past the chamfer radius. Forall other samples crush initiates around the radius of the core. CMS, therefore, prevents crush fromoccurring along the stabilized zone while providing resistance to crush in the unstabilized zone. This isfurther detailed in Section 5.5.2.As would be expected, a steeper chamfer angle reduces the force required for core movement withonset and collapse occurring at 152 and 183 kPa respectively. Again, the size of critical zone does notchange however. Interestingly, in the 45◦ core test, collapse occurs immediately following venting ofFigure 5.13: Minor deformations in cellular pattern of full tie down test.56100120140160180200220240260280300Net pressure (kPa)StandardHalf tie downsCMS45° coreOnsetCollapseFigure 5.14: Net pressure acting on the sandwich panels at onset and collapse for varying layups.The data for half tie downs represents that of the unrestrained edges.CollapseBag ventedNet pressure after vent (129 kPa)Net pressure before vent (183kPa)Onset5.083.811.2702.54Displacement (mm)Figure 5.15: In-situ data from 45◦ core test. Collapse occurs immediately after venting of vacuum,following a one minute hold at higher pressure.the vacuum bag. In this instant, the net pressure is actually reduced to 129 kPa. The pressure prior toventing (183 kPa) had been held for one minute. It is likely that the higher pressure state prior to ventingwas responsible for core movement and is therefore the value used in Figure 5.14. This suggests a timedependent response as the core was likely on the verge of collapse and reducing the net pressure did notdelay failure. This is seen in Figure 5.15.57Core edgePanel edgeEdgeband00.511.522.533.540 20 40 60 80 100 120 140 160 180Panel thickness (mm)Distance (mm)Full tie downs - LFull tie downs - WHalf tie down - LHalf tie down - WCMS - LCMS - W45° chamfer - L45° chamfer - WStandard - LStandard - WArtifact NoiseEdgebandCore edgeFigure 5.16: Change in thickness across laminate edge of processed panels in the L and W direc-tions for varying layups. The full tie down cases show no externally visible core movement.5.3.2 Ply movementGross ply movementAs in the two previous sections, the panel thickness across the laminate edge for each of the samples isdisplayed in Figure 5.16. Recall Figure 5.3 when interpreting the figure. The full tie down case providesa reference for a sample that does not exhibit core movement. Some artifact noise is seen which can beignored. Again, measurements of thickness include that of the paint layer, which is not necessarily evenamong samples, resulting in variability between tests.Thickness changes are similar amongst samples, indicating a similar pattern of core movement. The45◦ chamfer test is an exception, however. Several step-like changes in thickness are observed along thelaminate edge, indicating a cascading effect has occurred. This is associated with considerable relativeply movement due to slippage across multiple prepreg-prepreg interfaces.In each case a depression in thickness is seen just in front of the core edge. A similar patternwas observed in the previous tests, further suggesting that ply compaction is greatest at the edgebandtransition. The depression observed in the 45◦ chamfer test is signifantly larger than those of the othersamples,. Microscope images revealed that this difference is the result of the core moving relative to theplies to the extent that filler plies are no longer in contact with the core. This creates a ply drop off in58front of the core edge.Specific ply movementIndividual ply movement maps are presented in Figure 5.17. In all cases, ply 1 remains stationary whileall other plies slide to varying degrees. This matches that seen in Section 5.1 for samples processed at120◦C.Half tie downs show similar results to the standard layup despite a greater resistance to crush aspreviously shown. For the CMS sample, the bagside plies move more than the equivalent toolside plies.This is particularly true in the ribbon direction. This may be due to a weaker bond between plies andcore along the toolside due to stabilization. Typically as the panel is heated, the film adhesive will formfillets with the core cells. This acts to provide a good bond between core and ply. Similarly pillowingof plies into the core may occur. While not ideal for mechanical properties [10], it likely increasesthe friction between core and ply. Therefore, if the core moves, the plies will likely move with it.However, if the toolside is stabilized, fillets cannot form and pillowing will not occur. Therefore, lessfriction acts on the toolside plies as the core moves inward. In short, CMS may result in a decrease infrictional resistance between ply and core along the toolside. In fact, in the non-ribbon direction, thecore actually displaces 7 mm relative to the underlying plies. This is evidenced by the fact that fillerplies 5-6 no longer butt up against the core. Interestingly, filler plies 7-8, which extend slightly up thecore, don’t seem to move much, if at all, in the microscope images (Figure 5.18). This confirms that thetoolside frictional bond between core and prepreg has been weakened through stabilization. Therefore,for unstabilized layups, the frictional bond between core and prepreg can be assumed to be considerablyhigher than other frictional interfaces. That is, friction between prepreg layers, friction between theprepreg and bag (release film), and friction between the prepreg and tool (surfacing film). This raisesthe question that if tie downs were used in combination with CMS, would resistance to core movementincrease or decrease as compared with only using tie downs? On one hand, the core chamfer is stiffer.However, on the other hand, the plies are now restrained yet the frictional resistance between ply andcore on the toolside is less. In this scenario the core may simply deform relative to the plies.The 45◦ core sample displays considerable relative movement between plies. The plies immediatelysurrounding the core (plies 4-9) experience the largest displacement. It is possible that plies 4-9 initiallyslipped across the ply 9/10 and 3/4 interfaces. As these plies slid, the hydrodynamic friction acting on theneighbouring plies would have increased; thus dragging further plies inward, resulting in a cascadingeffect. Just as in the CMS-W sample, a gap is seen between the core and filler plies 5-6, indicatingdeformation of the core relative to the plies. In this sample however, the core moved relative to bothtoolside and bagside plies. This is observed in the microscope images (Figure 5.19), as plies 7-8 are nolonger at their original position along the core (6.35 and 12.7 mm up the chamfer edge respectively). Infact, in the L direction, the relative displacement of the core with ply 7 is such that the ply no longercontacts the core. Ply 8 only just maintains contact with the edge of the core (Figure 5.19-L). Similarly,in the W direction ply 7 only just maintains contact with the core. The displacement of the core edgerelative to the plies was approximately 15 mm in the L direction and 8 mm in the W direction. Such590 20 40 60 80 100 120 140 160 1801234567891011120 20 40 60 80 100 120 140 160 1801234567891011120 20 40 60 80 100 120 140 160 180123456789101112Half ply tie downsStandard0 20 40 60 80 100 120 140 160 180123456789101112Distance from panel edge (mm)0 20 40 60 80 100 120 140 160 1801234567891011120 20 40 60 80 100 120 140 160 1801234567891011120 20 40 60 80 100 120 140 160 1801234567891011120 20 40 60 80 100 120 140 160 180123456789101112Ply number Core machining stabilization45°chamferW LCore edge before movementCore edge after movementPly movementFinal ply positionFigure 5.17: Specific ply movement profile of processed panels in L and W for varying layups.Initially, the core edge is 50.8 mm from the panel edge.60Core EdgeCore EdgeWLFigure 5.18: Microscope images showing termination of plies 5-6 (red circle), ply 7 (yellow cir-cle), and ply 8 (blue circle) in the W and L directions for the CMS sample.relative deformation was not observed in any of the other samples except in the case of CMS. Henceit can be said that while the frictional interface between core and prepreg can be ignored for chamferangles of 20◦ and less, the interface must be considered for steeper chamfer angles.5.3.3 Extent of core movementImages of each sample post processing are shown in Figure 5.20. The extent of core movement for eachsample is outlined in Table 5.4. This represents the distance the core moves in from a single side. Thisis determined by subtracting the original position of the core from its final, post-cure, position.Generally speaking, samples that exhibited a higher onset pressure as compared with the standardlayup also exhibited a lower degree of core movement and vice versa. The half tie down test is anexception, where onset occurs at a higher pressure than the standard test, yet the extent of core movementis also greater. Unlike in the previous section, DOC does not play a role in dictating the extent of crush.Therefore, propagation of the crush front is allowed to continue until a high enough level of densification61Core EdgeCore EdgeLWFigure 5.19: Microscope images showing termination of plies 5-6 (red circle), ply 7 (yellow cir-cle), and ply 8 (blue circle) in the W and L directions for the 45◦ chamfer sample.is achieved that the resistive forces outweigh the applied force. In the case of the half tie down test, crushis only occurring along adjacent edges. Hence, the crush front can continue further in a single directionsince the cells far away are not simultaneously being crushed from the opposite direction, as would bethe case for the other tests. In other words, the edges experiencing core movement must compensate bycrushing further to achieve a similar level of densification that would be present if opposing edges werecrushing together.In the CMS sample, it was observed that the stabilized region experienced little to no core movement.CMS effectively limited the area that is able to crush to that of the unstabilized region. This resulted ina smaller area to densify, thus reducing the extent of movement.With a steeper chamfer angle, the horizontal component of pressure is much greater. Therefore ahigher level of densification is required to arrest the crush front, resulting in a higher degree of coremovement. Hence, why the 45◦ sample displayed the highest level of crush.Recall that the section cuts used to obtain the individual ply movement profiles were not always theexact center of the panel. Therefore, the extent of core movement displayed in Figure 5.17 may notalign exactly with that provided in Table 5.4.Once again, the ribbon direction exhibits greater crush than the non-ribbon direction. This furthersupports the notion that homogeneous, buckling of cells in the L direction is preferential to plasticcollapse in the W direction. As shown previously, a homogeneous collapse pattern as seen in the Ldirection, seems to progress through the panel at a faster rate; thus dominating the response. Unfortu-62WLStandard CMS45° chamferFull tie downs Half tie downsFigure 5.20: Standard, full tie down, half tie down, CMS, and 45◦ chamfer tests post-processing,showing extent of core movement.Table 5.4: Extent of deformation in the ribbon and non-ribbon direction associated with variouslayup features.Test Displacement in L (mm) Displacement in W (mm)Standard 100 86Half tie downs 109 99CMS 88 8245◦ chamfer 128 124nately due to failure of one or more of the LVDTs during processing, the rate of progression of coremovement in the L and W directions could not be compared.635.4 Dry core testThe purpose of this section is to gain a basic understanding of the behaviour of honeycomb core move-ment on its own, in the absence of plies (dry honeycomb). The test results come from the ”Dry CoreTest” outlined in Section 4.8.5.4.1 Failure pressureMovement of the core was noticeable almost immediately after external pressure application, at a netpressure of 111 kPa. This is labeled as the onset point. Collapse occurs at 147 kPa, after which therate of deformation increases further. This is shown in Figure 5.21. While the numbers are indicativeof the resistance to core movement offered by the core, the experiment does not mimic a compressiontest. Since the core is sliding across the tool under external pressure, there is a friction force that resistsmovement. That said, the PTFE release film under the core reduces the magnitude of friction withthe tool. However, the maximum deformation observed was less than the typical deformation in thesandwich panels tests, suggesting that the core-tool friction is still considerable. The core and releasefilm slide together over the tool (as observed from the video footage). The core was not see to crushsolely under vacuum pressure.Surprisingly, the core recovers almost entirely upon depressurization (Figure 5.22). In fact, theL direction displayed perfect elasticity with only minor plastic deformation observed in W along thechamfer radius (Figure 5.23). This indicates that the core cells can experience large strains (up todensification) while remaining elastic in nature. Therefore, the plateau stress (stress at cell collapse) isnot synonymous with an ultimate compressive strength. It follows then that honeycomb stiffness ratherthan strength is of importance in regards to core movement. It should be noted that, in sandwich panels,core movement is non-recoverable due to curing of the plies. This raises the notion that perhaps coremovement initiation should not be thought of in terms of a critical force, but rather in terms of a failurestrain.Unfortunately, in-plane properties of the honeycomb are not provided on data sheets by Hexcel[24, 25]. However, studies have been performed [21] which determined the Young’s Modulus ofNOMEX R© paper to be on the order of 3.1 GPa. In the experiments presented here, the honeycombshad an approximate cell size and thickness of 3.175 and 0.08 mm respectively. Therefore, using Equa-tion 2.4, the in-plane Young’s Modulus of the honeycomb is approximately 0.6 MPa. Similarly, theplateau stress can be calculated using Equation 2.8. This gives a plateau stress of 57 kPa. Thus, it doesnot require much stress to significantly deform the honeycomb. This is corroborated by the experimen-tal results above. Although the experimental values are double the analytical results (111 kPa vs 57kPa), there is likely a significant friction component that affects the experimental results. Moreover,the analytical results assume uniaxial compression, whereas in reality, biaxial affects may influencedeformation.6400:00:00 00:03:00 00:06:00 00:09:00 00:12:00Time (hh:mm:ss)-800-600-400-2000200400600800Pressure (kPa)20.017.515.012.510.07.55.02.5Displacement (mm)Pressure Sensor 1 - WVacuumAutoclave PressureLVDT 1 - WDepressurizationOnset (111 kPa)Collapse (147 kPa)Pressure sensor failedElasticrecovery0.0Figure 5.21: In-situ results from dry core test. Cells deform upon loading and recover almostentirely upon unloading.651 23 4Figure 5.22: Snapshots taken during autoclave test for dry honeycomb. 1: Prior to pressure appli-cation. 2: Moment of collapse. 3: Maximum deformation. 4: Fully unloaded honeycombcore following depressurization.66A BLWFigure 5.23: Plastic deformation in W following autoclave pressurization. (A) Image of entirehoneycomb. (B) Magnified view of deformation in W.675.5 Core movement mechanicsThe following section details observations regarding the initiation and progression of core movement.These observations along with results from the previous sections provide the basis for an updated me-chanical model.5.5.1 Initiation of core movementDuring the bending phase of honeycomb compression, the entire structure exhibits this response. Thesame pattern presents itself during the initiation of core movement. As the cellular structure bends, theprepreg similarly deforms until eventually, plies are dragged inward with the core and wrinkles form.The initial deformation of the prepreg creates a rougher surface which is picked up by the video cameraand LVDT during onset. Figure 5.24 provides an example of the change in surface roughness of thepanel, seen through the breather cloth, prior to collapse. The images in the figure come from the in-situvideo footage.The change in surface roughness associated with bending of the honeycomb structure is sensitive tominute deformations. In the case of the full tie down test, core movement was not detected, howeversmall distortions in the cellular pattern were observed from in-plane section cuts as shown in Figure 5.13.The surface roughness of the tie down test is notably different to the touch from that of the roomtemperature test where no distortion in the cellular pattern was seen. This infers that slight alterationsin the cellular pattern similarly present themselves in the prepreg surface. However, minor bending isnot easily detectable in-situ. Haptic feedback may prove a possible means of detection, but is limited topost-processing assessment. It is also possible that rearrangement of the LVDTs sensors would allow forsuch small deformations to be picked up. That said, aside from very minute distortions, the experimentalsetup described here provides a reliable means to detect core movement during the early stages of theprocess.Figure 5.25 provides a comparison of in-plane section cuts showing different levels of core move-ment during initiation. The first two images are of the room temperature sample displaying no coremovement, and the full tie down test showing minor bending. The latter two images are of the half tiedown test displaying considerable bending and of the 180◦C test showing early stages of collapse. Re-call that the 180◦C test exhibited relatively little core movement in comparison with the other samplesdue to a rapid advancement of DOC. The level of densification is significantly more advanced in theL direction than in the W direction as mentioned in Section 5.1.3. The image provided for the half tiedown test is between the two restrained sides. Note that the scale of bending in this sample is greaterthan in the full tie down test, showing that movement along the unrestrained edges influences deforma-tion along the restrained edges. Ideally, images would come from the same sample at different pointsalong the processing window. However, due to the nature of the experiments this was not possible. Asa result, varying extents of deformation are observed through different experiments.681 - Initial state 2 - Onset3 - one min after onset 4 - CollapseFigure 5.24: Change in surface roughness from onset of core movement through to collapse. Tran-sitioning from 1-4, a notable change in surface roughness is seen within the green circle.The bagside plies conform to the honeycomb structure as it deforms, observable throughthe breather cloth.5.5.2 Progression of core movementCollapse patternData from CMM scans allowed for a gradient map displaying slope angle to be created for each sample.This data is presented in Figure 5.26. The maps gives an indication of where core movement occurs.Artifact noise, presenting itself as a sudden line across the sample, is seen in a few of the imagesand should be disregarded. The images in the figure are representative of the experiments performedin Section 5.3, as the variation in layup results in an altered mechanical response. The experimentsperformed in Section 5.1 and Section 5.2 show similar results to the standard layup experiment, with theprimary difference being the magnitude of core movement. The slope-gradient maps for each sampleare provided in Appendix B.Near the chamfer radius, the angle of the chamfer drastically increases. Further along the chamferedge, however, the slope angle is largely unchanged, indicating crush does not occur along the lowerportion of the chamfer edge. This was also observed in experiments by Alteneder et al [6] and mentionedas being a reoccurring theme by Renn et al [42]. This effect is particularly true for the CMS sample,where the angle of the chamfer remains at 20◦ except for right at the radius, where there is a steep69LW(a) No core movement - room temp. (b) Minor bending - full tie downs.(c) Bending - half tie downs. (d) Collapse - 180◦C.Figure 5.25: In-plane initiation of core movement showing the progression of cell bendingthroughout the honeycomb. Images are from different tests showing (a) no core movement,(b) minor bending, (c) widespread bending, and (d) early stages of collapse.70NoiseNoiseNoiseFigure 5.26: Slope angle of processed panels. Dark red indicates a vertical slope with an angle of90◦. Dark blue indicates a flat plane with an angle of 0◦. For most samples, a slope anglegreater than 20◦ indicates crush has occurred along the chamfer edge. For the 45◦ chamfer,a slope angle larger than 45◦ indicates crush.increase in slope. This shows that CMS restricts core movement from occurring within the stabilizedzone. The unstabilized zone remains susceptible to crush, however. The sample with a 45◦ chamferdisplays a different result. In this test a drastic increase in slope is seen immediately, indicating thatcrush occurs along the majority of the chamfer edge if the chamfer is steep enough.In most samples, notable movement initiates as the panel begins to crush at the chamfer radius. Thecrush front (i.e. the edge of the crush zone) propagates through the flat section of the panel and the coremoves inward like an accordion. As the chamfer edge pulls in, it too crushes near the radius and thecrush zone extends further down the chamfer edge. This creates a change in slope along the chamferedge. With CMS however, crush does not originate at the chamfer radius. Rather, it originates beyondthe stabilized zone. As the crush front progresses, the chamfer edge pulls inward resulting in crush atthe radius. However, the crush zone does not extend down the stabilized chamfer edge. Therefore, thechamfer angle remains relatively constant. The difference in movement pattern during core movementfor the standard layup versus CMS are displayed in Figure 5.27a and Figure 5.27b respectively. Theimages in the figures come from the in-situ video footage of the 95◦C and CMS samples.Figure 5.27a shows a gradual increase in the size of the crush zone as the crush front progressesinward from the chamfer radius toward the center of the panel. In Figure 5.27b the crush front originates16.5 mm from the chamfer radius. Stabilization extends 6.41 mm past the chamfer radius meaning thatthe crush front originates 10.1 mm past the stabilized area. The crush front is observed through the71LW1 23 404:52 (mm:ss)06:16 (mm:ss)05:45 (mm:ss)05:26 (mm:ss)(a) Standard crush pattern - 95◦C sampleWL16.5 mm1 23 404:07 (mm:ss) 05:24 (mm:ss)06:15 (mm:ss) 06:43 (mm:ss)(b) CMS crush pattern.Figure 5.27: (a) Standard and (b) CMS crush patterns, displaying growth of the crush zone fromimage 1-4. Red arrows show the crush front in the L direction while blue arrows show thecrush front in the W direction. Elapsed time from the application of pressure is displayed.Note the change in orientation between samples.72formation of a large bagside wrinkle (image 1). Initially, as the crush front progresses, many wrinklesform within the crush zone (image 2). However, eventually as the chamfer edge moves inward, theradius of the core crushes and the crush zone begins to smooth out as the wrinkles congregate past thecrush front (image 3-4).Contrast is poor in the W direction, however LVDT sensors indicate that, for all samples, onsetand collapse occur simultaneously in the L and W directions. The video footage shows that the crushfront progresses at a faster rate in the L direction as compared with the W direction, particularly inFigure 5.27a. That is not to say that the core necessarily moves inward quicker. Rather, in the Ldirection cells beyond the crush front seemingly undergo collapse before the current row fully densifies,thus extending the crush zone. However, in the W direction densification at the crush front seems toprecede collapse in neighbouring rows. In both cases, however, the chamfer edge moves inward withcell displacement. In fact, for the 95◦C sample, LVDT data indicates that movement in the L and Wdirection occur at a similar rate (Figure 5.7a). Unfortunately, this is difficult to distinguish in the videofootage due to poor contrast in the W direction. This observation makes sense considering the mannerof cell collapse in the two directions. The homogenous, buckling pattern in the L direction would allowfor an extended projection of area susceptible to crush. On the other hand, the row-by-row collapsepattern in the W direction would only allow for the crush front to progress once the current row hasdensified. Both increase the magnitude of core movement, however the latter option results in a crushzone that grows at a slower rate. A similar result was presented by Heimbs et al [23]. It is reasonableto assume that this manner of crush progression would result in regions of higher densification in theW direction. Through-thickness and in-plane section cuts reveal this exact revelation and are presentedbelow. Through-thickness sections of the 95◦C sample are provided in Appendix B.As observed in the 600 kPa sample, it is possible for the W crush front to progress at a faster rate thanthe L crush front. This is shown in Figure 5.28. Again, the rate of core movement is similar among thesamples, with the L direction actually showing a slightly quicker deformation response initially (recallFigure 5.12). This suggests that the crush front that progresses through the material quicker will result ina greater final magnitude of core movement in that direction. As shown, initially the rate of progressionof the crush zone does not necessarily correspond to the rate of core movement due to the differentcollapse modes. However, it is likely that the crush front which progresses quicker will dominate thelate stage response, allowing for greater core movement in that direction.It should be noted that a faster crush front can also result in a faster rate of core movement initially.This was the case for the 120◦origC and 180◦ tests.Deformation of chamfer edgeIn the 120◦C, standard layup and 45◦ chamfer experiments the LVDTs were placed along equivalentedges to track deformation of the chamfer edge itself. Moreover, a 12.7 x 12.7 mm grid pattern wasimplemented to measure the rate of displacement of the chamfer edge from the video footage. Onlytwo tests of this caliber were conducted to see the difference in the rate of core movement for a chamferangle of 20◦ versus 45◦. The results of these tests are shown in Figures 5.29 and 5.30. In both Figure73LW1 23 404:32 (mm:ss) 05:32 (mm:ss)07:32 (mm:ss)06:32 (mm:ss)Figure 5.28: Crush pattern for 600 kPa sample, displaying growth of the crush zone from image1-4. W crush front (blue arrow) progresses at a faster rate than the L crush front (red arrow).Elapsed time from the application of temperature is displayed.5.29b and Figure 5.30b the dark lines mark the initial position of the radius and chamfer bottom andserve as reference points for measuring displacement. The points corresponding to the camera snapshotsare shown along the LVDT curves.For the 20◦ chamfer, initially, both the the top and bottom of the chamfer edge pull in at similar ratesaccording to Figure 5.29b. The video footage confirms this observation as both the top (yellow arrow)and bottom of the chamfer (red arrow) move one square over an equivalent timeframe (photos 1 to 2in Figure 5.29b). As crush continues, the bottom of the chamfer edge displaces at a faster rate than thechamfer radius. This is evidenced by the steeper slope of the bottom (cyan) LVDT. It is also apparent inthe video footage, as the chamfer radius moves less than one square for each square that the bottom edgedisplaces (photos 2 to 3). However, this pattern is not constant. The top and bottom edge are again seendeforming at approximately the same rate of one square per unit time (photos 3 to 4). The same resultis also seen in Figure 5.29a as the two LVDT curves display a similar average slope between points 3and 4. Unfortunately, the top LVDT fails shortly after point 4, however video footage indicates that thechamfer edge continues to progress at a relatively constant rate of deformation along its length.A similar pattern of deformation is seen at a chamfer angle of 45◦, albeit at a much quicker rate. As7400:21:00 00:24:00 00:27:00 00:30:00 00:32:59Time (hh:mm:ss)-800-600-400-2000200400600800Pressure (kPa)151050Displacement (mm)Autoclave PressureLVDT 1 - 'L','R'LVDT 2 - 'L','C-bottom'LVDTFailed1 234Onset Collapse(a) In-situ LVDT dataLW1 23 4(b) In-situ video camera footageFigure 5.29: Rate of deformation of the top and bottom of a 20◦ chamfer edge witnessed through(a) in-situ sensors and (b) in-situ camera footage. Red arrows points to the bottom of thechamfer edge while yellow arrows points to the top. Each square in (b) is 12.7 x 12.7 mm,allowing for the crush rate to be determined.7500:20:00 00:25:00 00:30:00Time (hh:mm:ss)-800-600-400-2000200400600800Pressure (kPa)151050Displacement (mm)Autoclave PressureLVDT 1 - L,C-topLVDT 2 - L,C-bottomCollapseOnsetNoise2 3 41(a) In-situ LVDT dataLW1324(b) [In-situ video camera footageFigure 5.30: Rate of deformation of the top and bottom of a 45◦ chamfer edge witnessed through(a) in-situ sensors and (b) in-situ camera footage. Red arrows points to the bottom of thechamfer edge while orange arrows points to the top. Each square in (b) is 12.7 x 12.7 mm,allowing for the crush rate to be determined.76seen in Figure 5.29a, the initial rate of deformation is equivalent between the chamfer radius and bottomof the chamfer. Unfortunately, the positioning of the LVDTs is such that the LVDT located on thechamfer radius is at a steeper angle than than the LVDT on the bottom of the chamfer. Therefore, for anequivalent displacement, the upper LVDT measures significantly more movement. Moreover, the corequickly pulls away from the upper LVDT whereas the lower LVDT maintains contact for longer. Hence,other than the initial pattern of movement after collapse, the LVDT data can not be used to measure therate of deformation of the two edges for this sample. The video footage has poor contrast, but it can stillbe used to approximate the rate of deformation of the chamfer. The radius is marked by an orange arrowand is distinguishable as the point where the squares begin to lay flat. The same is true for the bottomof the chamfer edge, which is marked by a red arrow. Figure 5.30b shows that between photo 1 and 2,the bottom and top edge each displace approximately one square as evidenced by the number of squaresabove and below the dark lines. Between photos 2 and 3 the bottom edge displaces one square while thetop edge displaces less than one square. Between photos 3 and 4 both edges displace one square. Thepattern is such that both edges initially deform at the same rate; the bottom edge then begins to displaceat a faster rate than the top edge, before both then start moving at the same rate again. This matcheswith the pattern of collapse seen for the 20◦ chamfer.This is in line with the aforementioned observation that crush originates at the chamfer radius andspreads inward. Assuming crush originates at the chamfer radius and propagates toward the center ofthe panel, it follows that, initially, the bottom chamfer edge matches the displacement of the top edge;the entire chamfer will pull inward as a unit. However, as crush extends down the chamfer edge, onlythe bottom of the chamfer edge will displace. In other words, the bottom edge will deform at a fasterrate than the top. This results in a steeper chamfer angle within the crush zone. Figure 5.31 providesa schematic explaining this. The first two illustrations show the difference in displacement for crushoccuring just beyond the radius versus crush occuring along the chamfer edge. The third illustration isrepresentative of the crush pattern exhibited in the experiments. Originally, crush occurs at the radius,dragging both the bottom and top edge in. As the chamfer itself crushes, the bottom edge begins todisplace at a faster rate than the radius creating a concave surface. Further crush proceeds at a similarrate among the top and bottom edges as cells beyond the radius continue to collapse. For a 20◦ chamferangle, crush does not extend fully down the chamfer edge, preserving the original angle in the lowerportion of the chamfer. However, for a 45◦ chamfer, crush extends along the entire chamfer edge.Therefore, the difference in velocity between the bottom and top edge is more drastic than for a 20◦chamfer. The amount of deformation the bottom edge must undergo to entirely crush the 45◦ chamferedge is less than for a 20◦ chamfer, since the 45◦ chamfer is shorter (both cores are the same height).Hence, for an equivalent relative deformation between bottom and top edge, the change in slope angleis more severe for a steeper core.The magnitude of the velocity vectors is affected by the level of densification; therefore, as the crushfront progresses, the rate of crush decreases. In contrast, as pressure increases, deformation velocityincreases. As such, the rate of core movement changes throughout the cure cycle and depends whetherthe part failed on a pressure hold or a ramp. To reduce complications associated with acceleration due77C-bottomC-top (radius)ABCC0 R0 C0 R0C0 R0C0 R0C0 R0 C0 R0 C0 R0 C0 R0Figure 5.31: Schematic showing the manner of crush progression through the core. A: Crushinitiates at chamfer radius and progresses inward - bottom and top of chamfer move at samerate. B: Crush initiates at chamfer radius and progresses down chamfer edge - bottom ofchamfer displaces quicker than top creating a concave surface. C: Actual manner of crushprogression through core - combination of A-B-A. 1 - Crush initiates at chamfer radius andprogresses inward (similar to A). 2 - Chamfer edge crushes quicker than cells beyond radius(similar to B). 3 - Crush switches again to collapse cells further within core (similar to A).to densification and increasing pressure, an instantaneous rate of core movement can be approximatedaccording to the time it takes the chamfer to displace 12.7 mm (i.e. one square). While not a trueinstantaneous velocity, this measurement provides an estimate as to the relative rate of collapse for thetwo types of panels. For a chamfer angle of 20◦ the rate of movement was determined to be 2.8 mm/s,whereas for the 45◦ panel the rate of movement was 3.4 mm/s.Intuitively it makes sense that collapse originates just beyond the chamfer radius. At this locationthe core is at maximum height, hence the effective area over which the horizontal component of pressureacts is also maximum. So too then is the applied force. Renn et al [42] also noted this. It therefore,follows that this would be the first location of cell collapse. As the collapsed cells increase in density,the force required to collapse cells farther ahead of the crush front would increase. Thus, the location ofcollapse shifts to along the chamfer edge. As densification increases along the chamfer edge, collapseoccurs further within the core again and vice versa. The steeper the chamfer angle, the easier it is tocrush cells along the chamfer edge.78~20°Slope steepens due to crushLWChamfer Radius(a) StandardLSlope angle steepens immediatelyCrush extends to core edgeWChamfer RadiusCrush almost extends to core edge(b) 45◦ chamferFigure 5.32: Through-thickness section cuts of processed samples. Purple indicates the crushzone, blue the densified regions, and red the bagside wrinkles. Additional features areoutlined in black. (a) The 120◦C, standard layup and (b) 45 chamfer samples are providedhere. Other samples are shown in the following images.79~20°Sudden steep bump at Chamfer radiusLW(a) CMSWL~20°Steepens due to crushChamfer Radius(b) Half tie-downsFigure 5.33: (a) CMS and (b) half tie-down samples.80Constant 20°Large (smooth) chamfer radius LW Rough surface(a) Full tie-downsLarge (smooth) chamfer radiusLWRelatively smooth surfacePorosityConstant 20°(b) Room temperatureFigure 5.34: Full tie-down and room temperature samples.81Through-thickness section cutsThe specific patterns of crush between the L and W directions are highlighted in through-thicknesscross sections, as shown in Figures 5.32 to 5.34. Samples from Section 5.3 are provided to represent thepatterns observed during core movement, as changing the layup can alter the mechanics of the response.Equivalent images for the experiments outlined in Section 5.1 and Section 5.2 are given in Appendix B.Generally speaking, the crush pattern for these latter samples show similar results to that of the 120◦C,standard layup test, with the primary difference being the size of crush zone.The crush zone is defined as the area where beyond which densification decreases drastically to thepoint of not having collapsed. In the L direction, densification in the crush zone varies linearly, with thelevel of densification decreasing away from the chamfer radius. In the W direction, however, severalunique highly densified regions are seen within the crush zone. Typically those nearest the radius exhibitthe greatest densification. Between each densified region, the level of densification is reduced, indicatingthose rows have not fully collapsed. This pattern of crush fits with the propagation behaviour observedin the video footage and matches the expected collapse mode associated with the two directions. Thatis, homogeneous collapse in the L direction produces a uniform crush zone with a densification gradientthat decreases in magnitude away from the initiation point. Row-by-row collapse in the W directionproduces a crush zone with scattered regions of high densification throughout. It follows that if crushoriginates at the radius, densification should be highest at this location, as observed.The specific pattern of core movement associated with each of the samples in Figures 5.32 to 5.34,match with the results seen in Figure 5.26. Both the standard layup (Figure 5.32a) and the half tie downexperiment (Figure 5.33b) display a uniform chamfer angle away from the radius. Closer to the chamferradius the chamfer angle begins to steepen as you enter the crush zone. With a pre-processing chamferangle of 45◦ (Figure 5.32b), however, the crush zone extends to the core edge in the L direction andnear the core edge in the W direction. As a result, the chamfer angle immediately begins to increase atthe edgeband transition. Conversely, as previously mentioned, CMS (Figure 5.33a) virtually eliminatescrush from occurring along the chamfer edge. The crush zone does, however, extend to the radiusresulting in a steep bump at the chamfer radius. The full tie down experiment (Figure 5.34a) doesnot display visible core movement. However, as noted in Section 5.5.1, the surface is notably rougherthan the standard layup sample processed at room temperature (Figure 5.34b) due to minor bending ofthe cell walls. In contrast, the room temperature sample experienced no core movement whatsoever,but displayed a large amount of porosity. In both cases, the chamfer edge remains at a constant angleequivalent to the pre-processing angle of the panels. The chamfer radii are also much larger as comparedwith the crushed samples, providing a smooth transition to the flat, bagside section of the panels.Note that a smaller cut was made for the half tie down sample in the L direction. The cut does notdisplay the area of core beyond the crush zone; hence, why no bagside wrinkles are seen in 5.33b.Figure 5.35 is a through thickness cut of the 120◦C, standard layup experiment. The cut is madealong the entire length of the panel in the W direction such that the crush zone on either side of thepanel is evident. The figure provides a visual for the general pattern of core movement as the crush frontprogresses through the panel from opposite sides. The pattern of movement is similar in the L direction82Figure 5.35: General crush pattern through panel in the W direction. Image is taken from the120◦C, standard layup experiment. Purple indicates the crush zone, red the bagside wrin-kles, and blue the densified regions.Table 5.5: Average panel height within the crush zone compared with the degree of core movementin the L direction.Test Height in L (mm) Displacement in L (mm)Standard, room temp 27.0 0Standard, 95◦C 30.2 108Standard, 120◦C 29.1 100Standard, 180◦C 28.1* 41Standard, 325 kPa 29.3 78Standard, 600 kPa 29.2 8545◦ chamfer 29.0 128CMS 28.7 88Half tie-down 28.8 109Full tie-down 27.1 0without the intermittent, highly densified regions as shown in Figures 5.32 to 5.34a. The cut was notmade along the same line as that displayed in Figure 5.32a, although both sections are from the samesample.In each case, immediately beyond the crush zone, large bagside wrinkles are seen. The wrinkles area result of plies pulling in with the core and then buckling under compression to absorb the extra lengththat is generated. Within the crush zone, the plies nearest the core are compressed with the collapsedhoneycomb and buckle with the cells themselves. Figure 5.36 shows a magnified image of this for the45◦ sample. Similar buckling is seen on the toolside surface of the crush zone. This indicates that uponcollapse, instead of cellular compression occurring independent of plies, the immediate core-side pliesbuckle with the core. The plies buckle over the span of the underlying honeycomb cells. In other words,the section of ply over the honeycomb cell acts as an unsupported Euler column with a length equalto the honeycomb cell size. Given that plies nearest the core experience the greatest movement understandard processing conditions, the core-side plies likely move relative to other plies at first.As the crush front progresses, bagside wrinkles coalesce together, leaving the crush zone relativelysmooth. As a result, the panel height along this zone is markedly higher as compared with an unde-formed panel. Figure 5.37 shows the comparison in height between an undeformed and severely de-formed sample. The images were generated using data from CMM scans, allowing height at any givenpoint to be extracted. The white bars indicate the crush zone in the L and W directions. Interestingly, a83(a) Bagside ply buckling - taken from the 45◦-W sample(b) Toolside ply buckling - taken from the 45◦-L sampleFigure 5.36: Buckling of plies nearest the core (blue arrows) on the (a) bagside and (b) toolsidesurfaces within the crush zone.84LWNoise35302520151050Height (mm)(a) No core movement - Full tie downsLW35302520151050Height (mm)(b) Drastic core movement - Standard, 95◦CFigure 5.37: Map of panel height extracted from CMM data for (a) no core movement and (b)drastic core movement. White bars in (b) represent the length of the crush zone in L and W.There is a notable increase in height along the crush zone.85Table 5.6: Average panel height within the crush zone compared with the degree of core movementin the W direction.Test Height in W (mm) Displacement in W (mm)Standard, room temp 27.0 0Standard, 95◦C 29.9 104Standard, 120◦C 29.7 86Standard, 180◦C 28.1* 16Standard, 325 kPa 28.8 52Standard, 600 kPa 29.6 9445◦ chamfer 28.8 124CMS 28.6 82Half tie-down 28.7 99Full tie-down 27.1 0greater degree of core movement does not necessarily result in a higher average panel height throughoutthe crush zone. This is shown in Tables 5.5 and 5.6. A possible explanation for this is that it is not theextent of crush that matters, but rather the level of densification within the crush zone. Assuming thesection of ply over each honeycomb cell buckles with collapse of the cell, then the level of buckling, andtherefore height, will correspond to the degree of densification of that cell. As wrinkles coalesce, theaverage height within the crush zone will correspond to the average level of densification. At high levelsof densification, the average increase in height of the first bagside ply within the crush zone shouldapproach half the honeycomb cell size. As additional plies buckle, the total increase in height of thepanel may be even greater. In these experiments, the cell size was 3.175 mm, with the average heightof undeformed panels around 27 mm. This gives a minimum average height of approximately 28.6 mmwithin the crush zone for panels exhibiting core movement. Most of the crushed panels display a totalheight near or above this value. The 95◦C test displays the greatest total height within the crush zone.This may be attributed to a greater degree of densification or simply be the result of inconsistencies inthe amount of paint used during CMM scanning. Paint film thickness was not measured for the panels.Note that because the 180◦C test did not allow much time for progression of core movement, wrin-kles within the crush zone did not yet coalesce. This is particularly true in the L direction. Hence,the height varies considerably between points, thus skewing the average height. In the W direction,few wrinkles were picked up. Therefore, the height presented in Table 5.6 is the max amplitude of anobserved wrinkle. Values displayed in Tables 5.5 and 5.6 were taken along the center lines of the panels.In-plane section cutsThe progression of core movement can also be observed from in-plane section cuts, similar to thosepresented in Figure 5.25. Figure 5.38 shows two different samples displaying drastic core movement.In each case, the sample has undergone densification. Figure 5.38a is from the 120◦C, standard layupsample and is representative of the typical pattern of deformation seen amongst all samples exhibitingcore movement. Figure 5.38b is from the 600 kPa sample, and represents the only case where the W864 cm3.2 cm(a) 120◦C, standard layupWL2.2 cm3.2 cm(b) 600 kPa, standard layupFigure 5.38: In-plane section cuts of processed panels showing densification along the L and Wdirections. The orange line represents the L crush front whereas the red line representsthe W crush front. Yellow arrows point to regions of low densification in the non-ribbondirection87direction exhibited a higher degree of core movement than the L direction. In both images, the patternof collapse in the L and W directions is apparent. Just as observed in the through-thickness sections, theL direction cells show fairly uniform densification within the crush zone. A slight gradient is observed,with the highest level of densification occuring at the core edge and decreasing towards the center of thepanel. For the most part, however, the level of densification remains constant until a point, signifyingthe edge of the crush zone (i.e. the crush front), where after densification is significantly reduced. Inthe W direction, regions of high densification are intermittent with regions of less densification. Thiscreates low density pockets within the crush zone that have not fully collapsed yet. The crush front isidentified by the last region of high densification. Beyond this point, densification is reduced. Alongtheir respective edges, the unique collapse pattern in the L and W directions is apparent. However,towards the center of the panel, the patterns interfere with one another creating a mixed response. This iscommon in biaxial compression of naked honeycomb (i.e. without facesheets) [22]. With the sandwichpanels presented here, often one collapse pattern will dominate the other. Typically, this is the L collapsepattern.In Figure 5.38a, densification in the L direction extends considerably further into the panel than inthe W direction. This is true for all other panels except that shown in Figure 5.38b, where the crush zoneis larger in the W direction. This implies that the two crush fronts vie for progression through the core.If one crush front progresses faster than the other, the faster collapse mode will constitute the majority ofthe densification pattern. The slower crush front will eventually be arrested by the pre-collapsed cells ofthe faster crush front. Given that in every test but one, the L direction exhibited greater core movementthan the W direction, it can be concluded that, in general, homegenous buckling in L presents a fastercollapse mode than row-by-row plastic deformation in W. That said, any given row in W at its bendinglimit could collapse [23]. It is therefore possible that rows further in the core could collapse before rowsnearer the radius. As a result the crush zone could be extended beyond that of the L crush zone, thus“winning” the biaxial response. Any un-collapsed cells within the crush zone are isolated from failurein L and may collapse later, thus still allowing for significant core movement in the W direction. Thisa possible explanation as to the failure pattern exhibited in the 600 kPa sample, where W deformationwas more extensive. In fact, several low density regions are seen within the sample’s crush zone (Figure5.38b). Moreover, the densified regions that are present are less dense as compared with equivalentregions in Figure 5.38a. This suggests an extended crush projection of the nature described above.In the 120◦Corig and 180◦C tests core movement progressed quicker in the L direction; likely due toa faster rate of collapse. However, as shown for the 95◦C and 600 kPa samples, the rate of progressionof core movement does not necessarily correspond to the rate of progression of the crush front. Unfor-tunately, for many of the experiments one LVDT sensor failed. Thus, the rate of progression of coremovement in the two directions is indeterminable for the other tests. Video footage in general does notprovide clear enough contrast to accurately measure the rate of core movement (excluding the two testswhere a grid pattern was implemented).Additional in-plane section cuts are provided in Appendix B.88Table 5.7: In-core pressure at onset and collapse.Test Onset [sensor 1, sensor 2] (kPa) Collapse [sensor 1, sensor 2] (kPa)Standard, 95◦C 77.1 74.1Standard, 120◦C 114 113.5Standard, 180◦C 116.4, 66.3 *179.8, 86.6Standard, 325 kPa 102.4, 81.42 (at 55◦C) 113.3, 92.4 (at 82◦C)Standard, 600 kPa 102.5 (at 34◦C) 97.1 (at 43◦C)45◦ chamfer, 120◦C *24.9 *59.8CMS, 120◦C Failed FailedHalf tie-down, 120◦C Failed Failed5.5.3 In-core pressureDuring vacuum application within the autoclave, the in-core pressure sensors often showed slightlyhigher pressure than vacuum. Upon venting of the vacuum bag, the in-core pressure increased fairlyrapidly to a level similar to that of the bag. This is shown in Figure 5.39. The magnitude of in-corepressure at onset and collapse for each sample displaying core movement is provided in Table 5.7. Inmost cases, onset and collapse occurred after venting, with an in-core pressure of around 1 atm. The 45◦chamfer displays considerably lower in-core pressure since onset occurred prior to venting of the bag andcollapse occurred immediately after venting; before the core pressure was able to stabilize (Figure 5.15).Interestingly, there is no significant change in in-core pressure between onset and collapse, as wellas following collapse for most tests. It is expected that in-core pressure would increase as cells arecompressed and volume is reduced. The fact that this was not seen indicates that gas is vented as thehoneycomb compresses. It is possible that as cells fail, new air paths are created through which gas canflow easily. More likely, however, is that the pressure sensor wires provide a continual path for gas flow.In addition, a cut-line was made in the core around the wires in order to firmly seat the sensors. Thistoo would have promoted gas exchange between cells. With this setup, pressure sensors cannot be usedto track core movement on their own. Furthermore, in each case, the sensors failed as crush progressed.The pressure at collapse for the 180◦C sample demonstrates the expected result of a pressure increasethrough cell volume reduction.The results suggest that the contribution of in-core pressure to core movement is similar to the levelof vacuum application. This matches results presented by Alteneder et al and Renn at al [6, 42]. Gaspaths introduced by insertion of the sensor wires exaggerated this effect. Therefore, the measured in-core pressure is not necessarily indicative of the in-core pressure further within the core (away fromthe sensors). Prior to venting of the bag, the samples displayed an in-core pressure of 10-20 kPa onaverage. It is possible that the internal pressure further within the core remained closer to this range ofpressure. This would exacerbate the force acting on the core. However, as the panels were never keptunder active vacuum for extenuating periods of time, it is also possible that the center region of the corenever achieved depressurization in the first place; thus, remaining near 1 atm throughout processing.Under typical cure cycles, pressure and temperature are applied simultaneously. Once the autoclavepressure reaches 1 atm gauge, the vacuum bag is vented. If the consolidation pressure is less than 1258900:16:00 00:18:00 00:20:00 00:22:00Time (hh:mm:ss)0100200300400500600700800Pressure (kPa)302520151050Displacement (mm)Pressure SensorVacuumAutoclave PressureLVDT 1 - 'W',''LVDT 2 - 'L',''Onset CollapseSensorFailureBag vented, in-corepressure recoversFigure 5.39: In-core pressure sensor response to core movement taken from the standard, 95◦Csample. Pressure recovery is captured upon release of vacuum. No difference in in-corepressure is observed between onset and collapse. Following collapse, considerable noise isseen due to pulling of the sensor wires.kPa, the prepreg remains relatively porous allowing gas to flow in an out of the core according to thelevel of vacuum bag pressure [6]. As consolidation pressure is increased, gas becomes entrapped withinthe core. This gas then expands with temperature resulting in an increase of in-core pressure as the curecycle progresses [6, 8]. In this study, however, pressure build up in the core was not a reality for most ex-periments. For the constant temperature experiments, those in Section 5.1 and Section 5.3, vacuum wascontinually applied during the heating phase. Therefore any buildup of pressure was constantly beingvented while the prepreg remained porous. Eventually the set temperature was achieved and autoclavepressurization began. At a compaction pressure of 1 atm, the bag was vented to atmosphere, resultingin a rise of in-core pressure to a similar level. Since temperature was held constant for the remainderof the cycle, the in-core pressure would not have increased further. In the case of the constant pressureexperiments (Section 5.2), heating was applied after venting of the vacuum bag. However, core move-ment occurred only minutes later, therefore in-core pressure was not allowed to significantly increase.This matches results presented by Brayden and Darrow [8], where it was shown that core movementwill preferentially occur early in the cure cycle before in-core pressure is allowed to sufficiently buildup. In fact, a slight increase in in-core pressure is observed between onset and collapse for the 325 kPaexperiment. This may be due to an increase in temperature between the two points or a reduction in cellvolume due to crush. The same result is not seen for the 600 kPa sample. The change in temperature90between onset and collapse is much less for this sample however.91Chapter 6Summary and Improved ModelThis body of work aims to investigate a variety of conditions involved in the processing of sandwichpanels that may influence core movement. The benefit to this research over previous studies is theability to capture core movement as it happens in-situ. Video footage provides a direct look into howcore movement initiates and then progresses through the panel. Considering the results and observationsoutlined in the previous section, an improved mechanical model is suggested.6.1 Drawback of current modelsPrevious work mentioned in Section 2.3.1 list three major factors that are involved in core movement.They are friction, core and ply stiffness, and the geometry of the core. Core stiffness is a combination ofthe material stiffness as well as in-core pressure. Friction is broken down into friction between prepregplies (Fp−p), friction between prepeg and the tool (Fp−t), and friction between prepreg and the bag(Fp−b). Geometry constitutes the chamfer angle and the height of the core. As presented in this study,there are a number of considerations that should be considered in the core movement phenomena. Theyare outlined below.Temperature effects on frictionIn previous studies [28, 35, 42], the prepreg-bag friction term has always been negated. However, inSection 5.1, it is shown that the processing temperature affects the frictional interfaces. At high tem-perature, the slip interfaces shift from between prepreg plies to between the immediate bag and toolsidesurfaces. Therfore prepreg-bag friction becomes of concern. That said, as noted in Section 5.2, if pres-sure is increased immediately, core movement is likely to occur before high temperature conditions areachieved. Therefore, temperature-induced changes in the frictional resistance of core movement onlybecome of concern if panels are brought to high temperature prior to pressure application. For typicalcure cycles where high pressure is introduced early on, temperature affects on core movement can benegated.92Ply bucklingAn important factor to consider is wrinkle formation through ply buckling. As crush progresses, plieswithin, and ahead of, the crush front buckle out of plane along the bag and toolside surfaces. In otherwords, the compressive strength of the plies offer additional resistance to crush which has largely beenglanced over. Previously this has been combined with the material resistance of the core and in-corepressure under the overarching term Fsti f f ness [28, 42]. However, since the bending modulus of pliesevolves with cure to become a significant value, it is separated here. Assuming the length of ply acrosseach honeycomb cell acts as an unsupported Euler beam, then the resistance to crush offered by the pliescan be approximated according to the Euler buckling criterion shown below.Pcrit =pi2EplyI(Kc2)(6.1)Where c is the unsupported length of ply, equivalent to the honeycomb cell size, and K is the columneffective length factor determined by the end condition of the plies at the cell walls. Eply of an uncuredply is quite low. However, as cure progresses this factor becomes more significant. Therefore, theresistance offered by ply buckling increases with the DOC. In fact, the reason the 180◦C experimentdoes not continue crushing is because ply stiffness increases significantly.Use of a surfacing filmUnlike in previous models, slippage across the prepreg-tool interface is less likely than slippage betweenprepreg layers for the layup used. This is evidenced by the fact that ply 1 (the first ply on the toolside)remains stationary in all but two tests (i.e. the 600 kPa and 180◦ tests). Furthermore, in the 600 kPatest, ply 1 moves considerably less than the other plies. This suggests that surfacing film increases thefrictional resistance across the toolside interface. Recall also that vacuum was applied following place-ment of ply 1 during layup. This too, may have acted to increase the frictional resistance across the ply1-surfacing film interface. In any case, the prepreg-tool friction term, Fp−t , can be ignored in determin-ing the onset of core movement. Just as in the Hsiao model, slippage across the bagside surface seemsto preferentially occur between prepreg layers rather than across the prepreg-bag interface. Therefore,the prepreg-bag friction term, Fp−b can also be disregarded. These statements are no longer true at highprocessing temperatures, however.Technically, an additional frictional interface exists between the surfacing film and the tool, Fs f−t .However, in no case did the surfacing film move. Therefore, it can be said that for tacky surfacingfilms, such as the one used here, the friction between surfacing film-tool is much greater than betweenprepreg-surfacing film. Thus, Fs f−t can be neglected and only the interface between prepreg-surfacingfilm need be considered along the tool. For simplicity, the prepreg-surfacing film interface is referredto as the prepreg-tool interface, Fp−t . For boardy surfacing films (i.e. not tacky) this relation may nolonger apply and Fs f−t should be considered.93Core-ply frictionFriction between the core and prepreg is typically ignored, as the plies adjacent to the core are assumedto be bonded with the core. However, as seen for for the 45◦ chamfer and CMS samples in Figure 5.17,it is possible for slip to occur across this interface. That said, slippage across the core-prepreg interfacewas not seen in any of the other samples. Moreover, considering that the core did not exhibit notablemovement in the full tie-down case shows that interfacial slip between core and prepreg is unlikely; atleast for unstabilized chamfer angles of 20◦ and less. If friction between the core and prepreg was notsignificant, then despite restraining adjacent plies, the core should have collapsed in this experiment.Therefore, in considering the conditions necessary for the initiation of core movement, friction betweencore and prepreg (Fp−c) is unlikely to be a factor for standard panel layups. Even in the case of CMSand steeper chamfer angles, it is unlikely that initiation of core movement occurs along the core-prepreginterface. Rather, given the extent of relative ply movement, slippage likely occurs first across a prepreg-prepreg interface followed later by slippage of the core itself as the crush front progresses. Video footageconfirms this as the plies can be seen moving as soon as core movement initiates. That said, the core-plyinterface may need to be considered for certain stabilization techniques or steeper chamfer angles.Straining of pliesPly slippage during core movement has traditionally been treated as a series of rigid members slidingpast one another. However, the elastic modulus of uncured carbon-epoxy prepreg is relatively low.Therefore, the plies are compliant and may experience some level of strain as the core attempts to pullinward. This would result in a tension force that resists core movement. Moreover, if it assumed thatall plies are strained equally prior to core movement initiation, then resistance to core movement wouldincrease further from the core, where more plies must be considered. Hence, if the coefficient of frictionis equivalent among plies, then slip will occur across the nearest ply-ply interface to the core. At thislocation, only one ply contributes to core movement resistance through tension. The problem can belikened to springs resting on top of one another in parallel. The resistive force due to strain energyis equal to the combined force exerted on each engaged spring. This is modeled in Figure 6.1 for thebagside. A similar scenario occurs on the toolside, but is not shown as the problem can be assumedsymmetric. The core is pulled with force Fx. At each interface, there exists a friction force F12 - FBag.In addition, each ply experiences some level of strain resulting in tensile forces F1 - F4. The furtherup slip occurs, the more plies (springs) become engaged. The force required to cause core movementis therefore a function of the magnitude of friction across the slip interface and the number of pliesengaged at that interface. Considering only these two forces, the following general relation can be madein regards to core movement initiation:Fx > 2[(F1+F2+ ...+Fn)+Fn,n+1] (6.2)Where, Fn,n+1 represents the magnitude of friction across the slip plane. The ‘2’ comes from thefact that the problem is symmetric and a similar scenario develops in the toolside plies. Assuming the94Core FxF1F2F3F12F23F34F4Ply 4Ply 1Ply 2Ply 3FBagFigure 6.1: Schematic showing the tension (F1-F4) developed in bagside plies as core deforms.Friction (F12-FBag) at the interfaces is provided. The plies can be approximated as springsslipping relative to one another.tension in each ply is the same (FT ), then the expression can be simplified to:Fx > 2(nFT +Fn,n+1) (6.3)Where n represents the number of plies above or below the slip interface.Thus, the failure scenario requiring the least energy is slippage across the ply 1-2 interface, whereonly tension in ply 1 is of concern. Under this scenario, ignoring all other factors (material stiffness, in-core pressure, etc.), core movement initiates when Fx > 2(FT +F12). Once slippage occurs, the strainedplies will unload somewhat and friction will dominate.Under high temperature conditions, slippage occurs across the bag and tool interfaces. Therefore,according to Equation 6.3, all plies are engaged and Fx > 8FT +FBag +FTool . Since the force requiredto cause core movement does not change much at high temperature, it implies that either the frictionbetween prepreg-bag and prepreg-tool decreases drastically compared to prepreg-prepreg friction or thetensile contribution is minimal at high temperatures. It is likely that the latter case is the predominantfactor. In the 180◦C test, the DOC was quite advanced at the moment of pressure application. Therefore,plies are no longer compliant and will preferentially slip rather than experience much (if any) strain.95Bag SideCoreTool SideLeθcorexyLcPPhcorePxPy PN(a) Location of initial cellular collapse (shaded area).ORSlippingBuckling(b) Condition for initial cellular collapse.Figure 6.2: Schematic showing the first zone of resistance against core movement (i.e. the chamferradius). In order for the cells to collapse, they must either slide relative to the under andoverlying plies or the plies must buckle with the cells.966.2 Core movement process6.2.1 Conditions for initiationDuring bending, cell distortion is exhibited throughout the core. The prepreg responds in kind by de-forming and creating a rough surface. Once collapse occurs the core and plies begin dragging inward.The same forces are involved in both cases, yet initiation of collapse is a localized response. To con-ceptualize the forces involved in core movement, it is best to look at the problem at the moment of cellcollapse. It was shown that collapse of the core seems to initiate just beyond the radius. This representsthe first zone of influence for resistive forces. At this location, the effective collapse area is constant andat a maximum. Looking at the problem from a 2D perspective, pressure is distributed per unit lengthrather than per unit area. Therefore, the width of the core is not important. The chamfer angle of thecore only serves to increase the horizontal component of pressure. This is presented in 6.2a with thelocation of initial cell collapse shaded.Focusing on this single cell, it can be said that in order for collapse to initiate the cell must sliderelative to the adjacent ply or the ply must buckle with the cell ( 6.2b). In each experiment performed,buckling seems to be the preferred mode of initial collapse. Relative sliding only becomes of concern forsteep chamfer angles or CMS during progression of core movement. Therefore, the forces of concern inthis zone are those generated from the core, Fcore, and buckling resistance of the plies. The core resistscrush by means of material stiffness and in-core pressure.During buckling, plies along the laminate edge are pulled inward. This gives rise to resistancethrough frictional interactions and ply tension, as mentioned previously. Either the plies slip in a purelyfrictional manner or the plies are strained. In the later case, slippage must still occur along the adjacentsurfaces to allow the plies to deform. A combination of the two modes is also possible. The laminateedge can therefore be taken as the second zone of influence for resistive forces. This is highlighted inFigure 6.3.Therefore, the condition for core movement initiation can be summarized as follows:Fapplied > (Fresist)Initiation (6.4)Where,Fapplied = f (Pnet ,hcore,θcore) (6.5)And,(Fresist)Initiation = f (Friction,FT ,Fcore,Pcrit) (6.6)If the laminate edge is taken to represent the zone of primary influence for frictional interactions,then the length (Le) of the laminate edge becomes important. Ignoring gravitational effects, the normalforce acting on this zone (Nlam) is equal to the net pressure multiplied by the laminate edge length.97Bag SideCoreTool SideLeθcorexyLcPPhcorePxPy P(a) Resistance develops along the laminate edge (purple circle) as plies are pulled inward.AND/ORStrainingPure slipTensionFrictionFrictionApplied forceApplied force(b) Condition for ply movement.Figure 6.3: Schematic showing the second zone of resistance against core movement (i.e. thelaminate edge). In order for the plies to move, they must either stretch and/or slip across aninterface.98Nlam = PnetLe (6.7)Coloumb fricton is proportional to the normal force according to Equation 2.16. Therefore Equa-tion 6.7 is important in understanding the magnitude of frictional resistance between prepreg layers aswell as across the prepreg-bag interface and across the prepreg-tool interface. It also highlights theimportance of geometry of the panel itself. A larger laminate edge allows for an increase in frictionalresistance. Each one of these friction regimes can therefore be said to be a function of the net pressure,laminate edge length, and their respective friction coefficient. In considering only the conditions neces-sary for the initiation of core movement, hydrodynamic shear stress can be assumed negligible. Hydro-dynamic effects are proportional to the relative velocity between layers. Therefore, prior to movement,τhyd does not influence frictional resistance. That is not to say that viscous effects do not contribute tofrictional resistance pre-movement, however. Rather, an adapted friction coefficient is used that evolveswith cure. It is itself a function of viscosity and dry friction and must be experimentally determined forthe given processing conditions. Once movement initiates, hydrodynamic effects should be considered.The following relations can, therefore, be made concerning the frictional interfaces during initiation ofcore movement:Fp−b ≈ PnetLeµp−b (6.8)Fp−p ≈ PnetLeµp−p (6.9)Fp−t ≈ PnetLeµp−t (6.10)Equations 6.8-6.10 show that while pressure is the driving force for core movement, it also con-tributes to core movement resistance by increasing the normal force along the laminate edge. However,increasing compaction pressure has also been shown to reduce the coefficient of friction [34, 35]. Re-gardless, a longer laminate edge results in an increase in friction force. An infinitely long laminate edgetherefore represents a system with infinite friction. Tie-down plies can be represented in this way. Thatis, using tie-downs is mathematically similar to drastically increasing the laminate edge length. Theequations are listed as approximate, as frictional interactions across the length of the chamfer edge, Lc,are ignored.Core movement initiation can, therefore, be simplified in an easy-to-understand way. Since move-ment originates at the radius, the core can be represented as a single cell of height hcore. The plies canbe thought of as a single entity sliding against two fixed interfaces, wherein the interfaces depends onthe processing conditions. In order for the cell to crush it must overcome the resistive forces outlined inEquation 6.6. That is, frictional resistance, ply tension, resistance of the core, and buckling resistanceacross the bag and toolside surfaces. Such a model is presented in Figure 6.4. The magnitude of thefriction forces are dependent on the pressure, laminate edge length, and friction coefficient according to99nFTLePnetNFp-pFp-pxyFcorePcritPcritPxhcoreLePnetNFp-bFp-tFcorePcritPcritPxhcoreStandard processingHigh temp processingFTnFTFTPnet+θcorePxPyPnetZones of resistance1. Chamfer radius 2. Laminate edgeCombined resistive force responsehcoreFigure 6.4: Core movement initiation for standard and high temperature processing. Each modelincorporates the use of bagside PTFE release film and a tacky surfacing film across the tool.100Equations 6.8-6.10. The driving force comes from the horizontal component of pressure, Px, and can bedenoted as follows:Fx = Pnet sinθcore(hcore) (6.11)Using this model, the condition for core movement initiation becomes:Standard processingPnet sinθcore(hcore)> 2FT +2Fp−p+2Pcrit +Fcore (6.12)*Pcrit is likely negligible.High temperature processingPnet sinθcore(hcore)> 2nFT +Fp−b+Fp−t +2Pcrit +Fcore (6.13)*FT is likely negligible.Fcore is the combination of resistive forces due to material stiffness of the honeycomb and entrappedgas within the honeycomb. Pcrit is the force required to cause buckling of the plies, described in Equa-tion 6.1. Figure 6.4 assume the use of bagside PTFE release film and a tacky surfacing film along thetool. If a boardy surfacing film is implemented instead, the models may need to be updated to includean additional friction term, Fs f−t , immediately adjacent to the tool. In this scenario, Fp−t becomesFp−s f (i.e. friction between prepreg-surfacing film). Similarly, if the core angle surpasses 45◦ or ifstabilization techniques are used, friction between the core and prepreg may need to be considered (seeSection 6.2.2).It may be useful to define core movement as a strain requirement. Equations 6.12 and 6.13 are setupbased on collapse of a single cell (i.e. core movement initiation). However, if tolerances allow forgreater deformation, additional forces may need to be considered as mentioned in Section 6.2.2.6.2.2 Conditions for progressionThe factors concerning core movement progression are more complicated than for initiation. Sincevideo footage does not capture individual ply movement as it happens, patterns can only be inferredpost-processing. The focus of this paper is not to delve into the physics of core movement progression.As such, this section is kept brief with key points highlighted.Following initiation, as the crush front progresses through the core, hydrodynamic effects need to beconsidered. This is particularly true for low temperature processing, where resin viscosity contributessignificantly to core movement resistance. At high temperature, hydrodynamic lubrication does notseem to play as much of a role, and can likely be negated.101θcoreLcNcxyFp-cFp-cPxPyPFigure 6.5: Schematic showing friction across the core-ply interface.As described in Figure 5.31, cell collapse originates at the radius and moves inward. The crush frontthen shifts to along the chamfer edge before shifting back to beyond the radius. This continual shiftingof collapse location can likely be attributed to densification increasing the resistance to crush. Relativedensity of cells then becomes an important attribute in considering core movement progression, as iteffectively changes the honeycomb’s stiffness.Therefore, in general, the resistive forces are not only a function of friction, ply tension, core materialstiffness/in-core pressure, and bending strength of plies, as is the case for initiation; but also hydrody-namic shear force and relative density. This is outlined in Equation 6.14. The factors controlling theapplied force don’t change.(Fresist)Progression = f (Friction,FT ,Fcore,Pcrit ,τhyd ,ρρs) (6.14)Ply tension likely plays less of a role in progression, as plies may unload once slippage occurs. Itis still considered in Equation 6.14, however. The “friction” term includes the same three frictionalinterfaces as for initiation. However, for steep chamfer angles (>20◦) or CMS, core-ply friction (Fp−c)must also be considered, as the core can slide relative to the plies. In the case of CMS, this is only truealong the toolside interface. Friction between the core and adjacent plies is of concern over the chamferedge. The toolside effective length in this zone is represented as Lc in Figure 6.2a. Bagside friction overthe chamfer edge differs from toolside friction as the interface between core and prepreg is not underpure shear as is the case on the toolside. The normal force (Nc) acts across the horizontal length of thechamfer edge for the toolside plies, and across the chamfer itself for the bagside plies. A schematic isprovided in Figure 6.5.(Nc)toolside = Pnet cos(θcore)Lc (6.15)102(Nc)bagside =PnetLccos(θcore)(6.16)Therefore,(Fp−c)toolside ≈ Pnet cos(θcore)Lcµp−c (6.17)(Fp−c)bagside ≈ PnetLcµp−ccos(θcore) (6.18)Just as with initiation, only the weakest frictional interfaces above and below the core are respon-sible for slippage. Therefore, for standard processing, either “friction” refers to two prepreg-prepreginterfaces, a single prepreg-prepreg interface and single core-prepreg interface, or both core-prepreginterfaces. As crush progresses down the chamfer edge, the effective length of the chamfer edge isreduced.Finally, directionality should also be considered in dealing with core movement progression. Themode of collapse is different between the ribbon and non-ribbon directions, such that ribbon deformationtypically progresses faster. As a result, the level of densification and corresponding stiffness vary alongthe two directions. Ribbon and non-ribbon effects can be negated in considering initiation of collapseas the effects are small compared to other terms. Moreover, in-situ data confirms that collapse occurssimultaneously in both directions.6.3 Processing windowIt is beneficial for manufacturers to understand the processing conditions that lead to core movement.While only one of each type of experiment was performed in this study, the results are still usefulin understanding the approximate range of pressure and temperature at which core movement mayoccur. The model outlined in Figure 6.4 allows for speculation into the failure conditions outside of theexperimental results. Figure 6.6 is an attempt to capture the processing conditions necessary for coremovement for a chamfer angle of 20◦. The curves were generated using results from Section 5.1 andSection 5.2 even though the experiments were conducted slightly differently (i.e. holding pressure vsholding temperature). Vertices represent actual data points along the curves. The bottom and top linesrepresent onset and collapse respectively. Between these two lines is the critical zone, wherein coremovement initiates. Additional experimental results from Section 5.3 are provided at 120◦C.Referencing the empirical failure envelope for a 20◦ chamfer, an expected failure envelope can begenerated for other chamfer angles using Equation 6.11. Figure 6.7 shows the expected failure envelopefor a chamfer angle of 45◦. The black dots represent the experimental data points at 120◦C for the 45◦chamfer test. The fact that the expected curves for the 45◦ chamfer do not fit with the experimentalresults at 120◦C may infer that the resistive forces are dependent on the chamfer angle. It is possiblethat the strain developed in the plies is greater for the 45◦ sample due to the increase in horizontalpressure. This would increase the tensile force component in Equation 6.12, resulting in a vertical10301002003004005006007000 25 50 75 100 125 150 175 200Net pressure (kPa)Temperature (°C)Core movementNo movement140160180200220240260Net pressure (kPa)onsetonsetonsetonsetcollapsecollapsecollapsecollapseHalf tie-downsCMSStandard45° chamferExperimental data points20°Figure 6.6: Empirical failure envelope outlining the processing conditions necessary for coremovement. The curves are based on experimental results for a chamfer angle of 20◦. Dottedlines are assumed values. Additional experimental results concerning a change in layup areprovided at 120◦C.shift in the curve of the manner seen. It is also possible that Fp−c does actually play a role in coremovement initiation for steep chamfer angles and is the source of this discrepancy (i.e. the frictionalfailure mode changes at steeper angles). However, more data needs to be collected in order to confirmthat it isn’t experimental variation. In any case, the theoretical failure envelope seems to be conservativein predicting core movement.In order to build processing maps with predictive capabilities, it is necessary to understand how theforces involved in core movement evolve during processing. The factors relating to core movementinitiation are outlined in Equations 6.12 and 6.13. Many of the resistive forces depend on temperaturedirectly or indirectly through advancement of the DOC. Figure 6.8 outlines this dependence. The curvesdemonstrate the general relation that friction, core resistance, ply tension, and ply buckling have withtemperature.FrictionIn these experiments prepreg-prepreg friction represents the weakest frictional interface at low tempera-tures. However, at higher temperatures (>120◦C) the interfaces of least friction shift to the prepreg-bagand prepreg-tool interfaces. This is similar to results presented by Martin et al [35], where they showedthat both prepreg-prepreg friction and prepreg-tool friction decrease with temperature before leveling10401002003004005006007000 25 50 75 100 125 150 175 200Net pressure (kPa)Temperature (°C)Core movementNo movementExperimental collapse at 120°CExperimental onset at 120°C45°Figure 6.7: Expected failure envelope for a chamfer angle of 45◦. The experimental results at120◦C are also provided.off or slightly increasing due to fiber intermingling. However, the leveling of the curve occurs earlierfor prepreg-prepreg friction than for prepreg-tool friction. As a result, at high temperature, prepreg-toolfriction is considerably less than prepreg-prepreg friction. They did not measure friction between theprepreg and bag however. Combining this understanding with the results presented in this body of workgives a friction curve as presented in Figure 6.8a.Core resistanceAs mentioned in Section 5.4, in-plane mechanical properties of NOMEX R© honeycomb are low. Hence,the majority of core resistance offered during core movement comes from the in-core pressure. Fig-ure 6.8b outlines the shape of the resistance curve. The material resistance of the core is denoted asFhoneycomb. As temperature is increased, gas pressure within the core (Fin−corepressure) increases linearlyaccording to the ideal gas law. In reality, the slope of the curve does vary slightly due to evaporation ofmoisture [6, 42].105TFf Fp-t & Fp-bFp-p>120°C(a) Influence of temperature on frictional resistanceTFcoreFhoneycomb + Fin-core pressure - initialFin-core pressure(b) Influence of temperature on core resistanceDOCPcritDOC GelationFTEplyGelationPcrit, FT(c) Influence of DOC on tensile and buckling resistance offered by pliesFigure 6.8: (a-b) General evolution of resistive forces with temperature and (c) DOC.106Ply tension and bucklingTemperature does not affect the modulus of the ply itself, but rather through advancing the DOC. It isgenerally accepted that the bending and tensile modulus of an uncured ply is low. Upon gelation, themoduli increase rapidly to approach that of the tensile modulus of the fibers. This can be shown usingsimulation software [3]. This has opposing effects for resistance due to strain buildup and resistance dueto ply buckling. Initially, when the modulus is low, the plies experience strain which contributes to coremovement resistance by means of a tensile force. However, this force is proportional to the modulus ofthe plies. Therefore, while tensile force contributes to core movement resistance, its contribution per plyremains relatively low. As the DOC advances, plies become less compliant and will preferentially sliprather than experience strain during core movement. However, with stiffer fibers, the resistance offeredby ply buckling increases drastically. This is reflected in Figure 6.8c.In the experiments performed, failure of the panels typically occurs before gelation. As such, theresistance offered by ply buckling is minimal and can likely be negated in most cases. However, inthe 180◦C test the DOC is around 0.5-0.6, indicating gelation has already occured. This shows that itis possible for core movement to occur after gelation and that in high temperature cases, ply bucklingcannot be negated; however, ply tension likely can. In-core pressure also becomes more significant withtemperature. That said, it should not be negated at low temperatures. The initial internal core pres-sure may still be relatively high depending on the level of vacuum pressure applied during layup andprocessing. Regardless, at low temperatures friction dominates the resistive response. As temperaturerises, friction decreases but remains a considerable factor. The value of ply tension is constant for lowDOCs and likely lower in magnitude than that of the friction response. In fact, the failure envelopeoutlined in Figure 6.6 follows a similar profile to the friction model shown in Figure 6.8a, implying coremovement is largely dictated by the response of the frictional interfaces regardless of the processingconditions. However, as explained in Section 5.5.3 the manner of vacuum application prevented buildupof in-core pressure for constant temperature tests. Therefore, the heightened in-core pressure at highertemperatures did not manifest itself. In general, under typical processing cycles where pressure is setand temperature is increased, core movement occurs early enough in the cure cycle that in-core pressuredoes not become significant. This was shown here in the 325 and 600 kPa tests and has been previouslyreported in literature [6, 8]. The significance of this is that, unless in-core pressure is specifically manip-ulated or the DOC is allowed to advance significantly before pressure application, friction will controlcore movement resistance; particularly prepreg-prepreg friction if a tacky surfacing film is implemented.The resistance offered by in-core pressure will, more or less, be equal to the vacuum pressure. Since plytension does not evolve with temperature, it will only serve to shift the processing window up or down.The shape of the failure curve will still be dictated by the friction response, however.Assuming this statement holds true, then by knowing the level of strain in the plies, the coefficientof friction can be approximated at each point on Figure 6.6. Of particular interest is the coefficient offriction at 120◦C, where the failure pressure is at an apparent minimum. An approximate upper boundis determined below by setting the strain in the plies to be zero (i.e. FT = 0). The values from Table 5.7107are used for in-core pressure.Pnet sinθcore(hcore)>*02FT +2Fp−p+>02Pcrit +Fcore ≈ 2Fp−p+Fin−core pressure (6.19)Ignoring ply stiffness, material stiffness of the honeycomb, and substituting Equation 6.9 for Fp−pyields,(Pnet −Pin−core)sinθcore(hcore)> 2(PLeµp−p) (6.20)Subbing in the appropriate values at onset and collapse gives,(µp−p)120◦C onset = 0.028 (6.21)And,(µp−p)120◦C collapse = 0.037 (6.22)Taking the coefficient of friction at 120◦C as the average of these two values yields,(µp−p)120◦C ≈ 0.032 (6.23)This represents the upper bound of prepreg-prepreg friction at 120◦C if ply tension is assumed negli-gible. If initiation of core movement is taken as collapse of the first cell then this seems a fair approx-imation, as the maximum deformation experienced by the plies is small (≈ 3.2 mm) relative to theirlength. However, this wrongly assumes no strain builds up in the plies during the bending phase ofhoneycomb compression.The above friction value is equivalent to the minimum prepreg-prepreg friction value reported byMartin et al [35] for another material system (Hexcel F-593) qualified to BMS8-256. In their study,the coefficient of friction was measured under a compaction pressure of 394 kPa with a minimum valueoccurring around 71◦C, versus 170-200 kPa compaction pressure at 120◦C in this work.6.4 Design improvementsThe implication of friction dominating the resistive response under standard processing is that relativelysimple design changes can be made to mitigate the risk of core movement. Table 6.1 labels the factorsinvolved in core movement that influence the applied force as well as the resistive forces. The threeterms, θcore, hcore, and Le are geometrical in nature, which would mean the processing cycle need notbe altered. Of these terms, hcore and θcore may be restricted by dimensional tolerances. However,the laminate edge can be trimmed after processing to meet design specifications. It was previouslyshown that prepreg-prepreg, prepreg-bag, and prepreg-tool friction are proportional to the laminate edgelength. Therefore, increasing the length of the laminate edge seems a viable option for improving coremovement resistance without altering the processing cycle. Larger tools would need to be implemented108Table 6.1: Main parameters involved in the initiation of core movement.Fapplied FresistPnet Pnetθcore Lehcore µPin−coreEhoneycombEply01000200030004000500060000 100 200 300 400 500 600 700Applied force (N)Net pressure (kPa)Fp-p: Le = 50mmFp-p: Le = 100mmFp-p: Le = 115mmFp-p: Le = 150mmApplied ForceFigure 6.9: Effect of altering the laminate edge length on prepreg-prepreg friction at 120◦C. Reddots represent core movement initiation. In-core pressure is assumed to be equal to 1 atm.to account for the size increase; however, the idea is that grit strips and tie-downs no longer becomenecessary.Figure 6.9 shows the effect increasing the laminate edge has on prepreg-prepreg friction at 120◦C.The curves were generated assuming a coefficient of friction equal to 0.032 and an in-core pressure valueof 1 atm. Buckling resistance of plies, ply tension, and material resistance of the core were assumednegligible. The point at which the applied force crosses the friction line(s) represents a conservativeestimate of core movement initiation. In order to prevent core movement from occurring up to netpressures of 600 kPa, the laminate edge length would need to be at least 115 mm; more than double thelaminate edge length used in this study.Previous attempts [6] have been made to understand the scale-up effect of a longer laminate edge onsandwich panel quality. However, the focus of these studies was in determining the viability of in-corepressurization techniques for lab-scale and large-scale parts. An understanding of how an increase in109the laminate edge length affects core movement resistance on its own has not yet been reported.Unlike increasing the laminate edge length, increasing the panel height and chamfer angle wouldact to promote core movement. The latter effect was shown in the 45◦ chamfer test, albeit not to theextent predicted. Figures 6.10a-6.10b show the expected effect of increasing these two parameters onthe applied force. This is contrasted with the effect of increasing the laminate edge length. Again, thecurves were generated assuming a coefficient of friction equal to 0.032 and an in-core pressure value of1 atm. The point(s) where the applied force lines (dotted) cross the friction lines (solid) represent theinitiation of core movement for the given geometry. Maps such as these are useful in making designdecisions as they allow manufacturers to construct panels to meet the processing conditions.Increasing the core height is less severe than increasing the chamfer angle in terms of the appliedforce generated. That said, as the chamfer angle approaches 90◦ the change in applied force petersout. In both cases, a laminate edge length of 150 mm is insufficient to prevent core movement in themost extreme cases. If both an increase in height and an increase in chamfer angle is implemented, thepromotion of core movement would be even more significant. Similarly, decreasing both the core heightand chamfer angle would significantly demote core movement initiation. Curves showing the effect ofchanging both parameters together are not provided.6.4.1 RecommendationsIn order to build accurate predictive maps such as those provided in Figures 6.9 and 6.10, it is necessaryto understand the exact coefficient of friction under different processing conditions. Frictional character-ization tests need to be carried out at a variety of temperatures to develop a friction model for the givenmaterial system. This includes prepreg-prepreg friction, prepreg-bag friction, and prepreg-tool friction.From this, more sandwich panel tests at specific temperatures should be conducted wherein the panelis brought to failure and the failure pressure measured. Strain gauges (or alternate methods) shouldbe implemented to measure the strain developed in the plies at initiation of core movement. This willdetermine the contribution ply tension plays in regards to core movement resistance. If it is insignificantin comparison to friction then it can be disregarded. At the moment, the extent of strain in the plies priorto movement is relatively unknown. The process map provided in Figure 6.6 is a good starting point, butonly represents a few tests, some of which implemented a different cure cycle. If the process map canbe shown to correlate well with the friction model, then it can be said that friction dominates the coremovement response unless in-core pressure (or other properties) are specifically manipulated. In otherwords, the mechanical model would boil down to applied pressure, vacuum pressure, panel geometry,and frictional resistance. If this is the case, then accurate predictive failure maps can easily be madefor a wide range of processing temperatures and pressures under standard cure cycles. This will provea useful tool for manufactures in allowing them to design for the processing cycle, rather than makeempirical based choices.With a baseline friction model and understanding of the primary forces involved in core movement,further experiments should be conducted for specialized panels to understand their exact effect on themechanical response at different temperatures. That is, implementation of CMS, tie-downs, steeper11001000200030004000500060000 100 200 300 400 500 600 700Applied force (N)Net pressure (kPa)Fp-p: Le = 50mmFp-p: Le = 100mmFp-p: Le = 115mmFp-p: Le = 150mmFapplied: h = 20mmFapplied: h = 25mmFapplied: h = 30mmFapplied: h = 35mm(a) Increasing hcore01000200030004000500060000 100 200 300 400 500 600 700Applied force (N)Net pressure (kPa)Fp-p: Le = 50mmFp-p: Le = 100mmFp-p: Le = 115mmFp-p: Le = 150mmFapplied: Θ = 10°Fapplied: Θ = 20°Fappplied: Θ = 30°Fapplied: Θ = 40°(b) Increasing θcoreFigure 6.10: Effect of increasing the (a) core height and (b) chamfer angle for various laminateedge lengths at 120◦C. In-core pressure is assumed to be 1 atm.111chamfer angles, combinations of these, etc. The results shown in this study provide a first look at howthe mechanics are altered at 120◦C. However, it is relatively unknown whether the same relations wouldhold true at other temperatures. Moreover, it would interesting to see the results for panels with differentsurfacing films and film adhesives. It is suspected that changing from a tacky surfacing film to a boardyone will alter the slip interfaces. Other studies of interest would be understanding exactly when slippageoccurs between core and ply for steeper chamfer angles. If it turns out that this is the first layer to slide,then altering the chamfer angle has a more significant effect than simply changing the applied force; itwould also change the slip interface. The same goes for CMS.Finally, scale-up studies investigating the effect of the laminate edge length on core movement re-sistance should be conducted. This may prove a relatively simple and inexpensive technique to combatcore movement.112Chapter 7Conclusions7.1 Summary of resultsCore movement is a multi-faceted phenomenon that plagues sandwich panel production during the au-toclave processing stage. The initiation, progression, and extent of core movement are highly dependenton the manner of panel construction and the processing conditions. This body of work investigated threemain parameters influencing core movement. Namely, processing temperature, pressure, and alterationsin the panel layup. Novel techniques were used to track core movement evolution in-situ, includingfilming the process in real time. Experimental results and observations were analyzed and a simplemechanical model was proposed.Resistance to core movement seemed to match previous friction models, wherein at lower temper-atures (95◦-120◦C) resistance to core movement is dependent on viscosity and decreases with tem-perature. At high temperature (180◦C), resistance to core movement changes irrespective of viscosityimplying a shift from a hydrodynamic friction regime to one exhibiting boundary lubrication. In the lat-ter scenario, the slip interfaces shift from between prepreg plies to between the tool and bag interfaces.This results in bulk movement of plies as opposed to relative movement as exhibited in the lower tem-perature samples. However, unless the set temperature is first achieved prior to increasing pressure, coremovement will occur before high temperature conditions are reached. In fact, at high pressure (500 kPa)core movement will occur if temperature rises to just 34◦C. That said, if temperature is kept below 30◦C,core movement is unlikely to occur even for pressures around 600 kPa. This demonstrates the need forstrict tolerances during processing. Similarly, pressure fluctuations of just 30 kPa can result in drasticcore movement if the processing conditions approach those necessary for core movement initiation.Tie-downs proved effective at mitigating core movement. Single ply tie-downs were enough to elim-inate the vast majority of core movement for a core angle of 20◦, with only minor bending seen. Itis suspected that utilizing double ply tie-downs will prevent all deformation. Interestingly, restrainingonly two edges increased the resistance to movement along the unrestrained edges. This implies thatdeformation along the ribbon and non-ribbon direction are somewhat coupled. Implementing Core Ma-chining Stabilization (CMS) slightly increased the panel resistance; however, it also allowed for slippage113across the toolside core-ply interface in the non-ribbon direction. It is suspected that further stabilizationmay result in increased slippage and possibly even decrease the panel’s resistance to core movement.That said, collapse of the cellular structure was prevented along the stabilized zone. Finally, a steeperchamfer angle (45◦) increased the applied force, resulting in earlier onset of core movement and con-siderable relative movement of plies. Notable slip between core and prepreg was also observed on bothbagside and toolside interfaces.Core movement initiates near the radius of the chamfer edge where the core height is maximum.From there collapse progresses inward. As densification increases beyond the radius, crush extendsdown the chamfer edge. Similarly as densification increases along the chamfer, cells further in the corebegin collapsing at a faster rate again. This results in a concave-like chamfer edge following processing.Typically the original angle is persevered further down the chamfer. In the case of CMS, the angle ispreserved along the entirety of the chamfer edge. For steeper angles, crush extends along the entirelength of the chamfer edge.Generally speaking, deformation in the ribbon direction proved greater than in the non-ribbon direc-tion, with the exception of one sample. Ribbon deformation allows for quick progression of the crushfront through the core by means of homogenous collapse. Non-ribbon deformation, however, exhibitsrow-by-row collapse resulting in high localized densification but typically slower progression throughthe core. Surprisingly, initiation of core movement occurs simultaneously in the two directions.A simple mechanical model was created based on the observations and results attained. It includes theinfluence of temperature on frictional resistance and the addition of ply tension which was not previouslyincluded in core movement models. Moreover, the importance of geometry is highlighted. The sequenceof events leading to core movement are also captured. First, cell collapse occurs at the radius, bucklingthe adjacent plies. In doing, so plies are dragged inward across the laminate edge. Resistive forces canbe grouped into two categories, those occurring in the crush zone and those arising from interactionsacross the laminate edge. They are broken down as follows:1. Crush zone (chamfer radius)• Material strength of core• In-core pressure• Buckling strength of plies2. Laminate edge• Friction across the slip interface• Tension in pliesConsidering that the driving force is a function of applied pressure and panel geometry, that givessix major factors that contribute to core movement. Namely, net pressure across the vacuum bag, corestrength (including in-core pressure), buckling strength of the plies, friction, ply tension, and geometryof the panel. At low DOCs, prepreg-prepreg friction seems to be the dominant mechanism controlling114core movement resistance. If true, friction models can be used to predict the initiation of core movement.That said, ply tension has largely been overlooked and may prove an important parameter to consider.In order to truly predict the initiation of core movement for any DOC, the magnitude of each of thesesix parameters need to be known throughout the cure cycle for the given material system. In any case,altering geometrical features such as the length of the laminate edge in order to increase frictionalresistance seems a promising method to combat core movement.7.2 Future workIn order to affirm whether friction is the dominant mechanism in the core movement process, frictionmodels need to be developed for a variety of material systems. Sandwich panels of the same materialsystems should be built and processed at a variety of temperatures. If core movement is shown to matchthe friction behavior then it can be said that friction dictates core movement unless other factors, such asin-core pressure, are artificially manipulated. Ply strain should also be measured in the sandwich paneltests to understand the level of resistance offered by ply tension. If significant, the processing windowwill shift upward from a purely friction-based response. Understanding ply strain prior to movement willallow for friction and tension effects to be decoupled. It is suspected that ply strain contributes to coremovement resistance, however, less so than friction. As our understanding of pre-gelation behaviourimproves, an accurate representation of the resistance offered by the plies can be realized. Currently, itis assumed that the plies themselves offer little resistance to core movement at low DOCs.With regards to layup effects, cores of different geometries should be investigated. This body ofwork only dealt with cores of a single height. Furthermore only two chamfer angles were investigated,20◦ and 45◦. It would be beneficial to evaluate the core movement model against panels with a widerange of geometries. That is, combinations of cores with varying heights and chamfer angles, as well aspanels with differing laminate edge lengths. In addition, it would be useful to understand how effectivetie-downs are for steeper chamfer angles such as 45◦. Moreover, it would be interesting to note howthe use of 45◦ doubler plies influence core movement. Finally, differing surfacing films should beinvestigated. It is suspected that the use of boardy surfacing films will give drastically different resultsinto where slippage initiates and how ply movement progresses. Similarly, adjusting the extent of CMSis suspected to yield different results in regards to core-ply slippage. On that same note, changing thefilm adhesive system will likely alter the frictional resistance between core and ply as well. Investigatingeach of these parameters will likely involve several studies with multiple experiments pertaining to eachtopic.It is recommended that future experiments should be run under constant vacuum pressure to eliminateany effects associated with venting. This may include not applying any vacuum pressure at all or runningthe tests at full vacuum. Furthermore, it is suggested that ply 1 should not be debulked during layup, asthis likely alters the frictional resistance across the ply 1-surfacing film interface. Although cumbersomeand time-consuming, including filler plies in the layup is useful as they allow for easy tracking ofrelative movement between the core and prepreg. Finally, in order to eliminate gas pathways duringcore movement, high gauge pressure sensors wires should be used. In order to get more accurate results,115multiple sensors should be placed throughout the core as gas pressure can vary from cell to cell. Theyshould be placed on the underside of the core, to prevent intrusion of the film adhesive. Slicing thehoneycomb to seat the pressure sensors should be avoided if possible. The duration of time which thepanel is held under vacuum prior to processing should also be measured as this will influence the levelof in-core pressure.116Bibliography[1] Daytronic, retrieved from: http://www.daytronic.com. 2019. → page 121[2] Candylabs, retrieved from: http://www.candylabs.com. 2019. → page 123[3] Convergent manufacturing technologies, retrieved from: http://www.convergent.ca. 2019. →pages 28, 107, 123[4] L. Aktay, A. F. Johnson, and B.-H. Kro¨plin. Numerical modelling of honeycomb core crushbehaviour. Engineering Fracture Mechanics, 75(9):2616–2630, 2008. → page 2[5] H. G. Allen. Analysis and design of structural sandwich panels: the commonwealth andinternational library: structures and solid body mechanics division. Elsevier, 2013. → page 2[6] A. Alteneder, D. Renn, J. Seferis, and R. Curran. Processing and characterization studies ofhoneycomb composite structures. In INTERNATIONAL SAMPE SYMPOSIUM ANDEXHIBITION, pages 1034–1034. SAMPE SOCIETY FOR THE ADVANCEMENT OFMATERIAL, 1993. → pages 14, 17, 19, 69, 89, 90, 105, 107, 109[7] M. F. Ashby and R. M. Medalist. The mechanical properties of cellular solids. MetallurgicalTransactions A, 14(9):1755–1769, 1983. → page 7[8] T. Brayden and D. Darrow. Effect of cure cycle parameters on 350◦F cocured epoxy honeycombcore panels. In Proceedings of the Society for the Advancement of Material and ProcessEngineering - Thirty-fourth Technical Conference, 1989. → pages 3, 14, 15, 17, 18, 90, 107[9] J. Brown and B. Ennis. Thermal analysis of nomex R© and kevlar R© fibers. Textile ResearchJournal, 47(1):62–66, 1977. → page 10[10] F. C. Campbell Jr. Manufacturing processes for advanced composites. elsevier, 2003. → pages5, 17, 59[11] Q. Chen and N. M. Pugno. In-plane elastic buckling of hierarchical honeycomb materials.European Journal of Mechanics-A/Solids, 34:120–129, 2012. → pages 7, 10[12] D. H. Corbett and S. A. Smith. Tiedown ply for reducing core crush in composite honeycombsandwich structure, Apr. 20 1999. US Patent 5,895,699. → page 17[13] D. A. Crump, J. M. Dulieu-Barton, and J. Savage. The manufacturing procedure for aerospacesecondary sandwich structure panels. Journal of Sandwich Structures & Materials, 12(4):421–447, 2010. → page 17[14] Daytronic. 5M30 AC LVDT Conditioner Module product manual, . → pages xv, 121, 122117[15] Daytronic. LVDT AC, short stroke, unguided, [SUH Series] product datasheet, . → page 121[16] C. Duffner, N. Zobeiry, and A. Poursartip. Examination of pre-gelation behavior in as4/8552prepreg composites. In Proceedings of the American Society for Composites—Thirty-thirdTechnical Conference, 2018. → page 14[17] J. Dulieu-Barton, D. Crump, and D. Griffith. Effect of core crush on honeycomb sandwich panels.In Proceedings of the 9th International Conference on Sandwich Structures, 2010. → page 17[18] DuPontTM. Nomex R© history, retrieved from:http://www.dupont.com/public sector er/en gb/nomex the fibre/nomex history.html. 2011. →page 2[19] S. Erland, T. Dodwell, and R. Butler. Characterisation of inter-ply shear in uncured carbon fibreprepreg. Composites Part A: Applied Science and Manufacturing, 77:210–218, 2015. → pages13, 14[20] N. Ersoy, K. Potter, M. R. Wisnom, and M. J. Clegg. An experimental method to study thefrictional processes during composites manufacturing. Composites Part A: Applied science andmanufacturing, 36(11):1536–1544, 2005. → page 13[21] C. C. Foo, G. B. Chai, and L. K. Seah. Mechanical properties of nomex material and nomexhoneycomb structure. Composite structures, 80(4):588–594, 2007. → page 64[22] L. Gibson and M. Ashby. Cellular Solids: Structure and Properties. Pergamon Press, 1988. →pages x, 5, 7, 8, 9, 10, 11, 88[23] S. Heimbs, S. Schmeer, P. Middendorf, and M. Maier. Strain rate effects in phenolic compositesand phenolic-impregnated honeycomb structures. Composites Science and Technology, 67(13):2827–2837, 2007. → pages 9, 73, 88[24] C. Hexcel. Hexweb honeycomb attributes and properties. 1999. → page 64[25] HEXCEL R©. HexWeb R© HRH-10, Aramid Fiber/Phenolic Resin Honeycomb, Product Data Sheet,2017. → pages 5, 24, 64[26] W. B. Hopkins and D. E. Hartz. Adhering tiedown plies in composite construction, Nov. 11 1997.US Patent 5,685,940. → page 17[27] H.-M. Hsiao, S. M. Lee, R. A. Buyny, and C. J. Martin. Core-crush resistant fabric and prepregfor fiber reinforced composite sandwich structures, Dec. 16 2003. US Patent 6,663,737. → page18[28] H.-M. Hsiao, S. Lee, and R. Buyny. Core crush problem in the manufacturing of compositesandwich structures: Mechanisms and solutions. AIAA journal, 44(4):901–907, 2006. → pagesx, 2, 3, 14, 15, 16, 18, 92, 93[29] T. R. Kain Jr. Composite honeycomb sandwich panel for fixed leading edges, Jan. 30 2001. USPatent 6,180,206. → page 17[30] N. A. Kalebek and O. Babaarslan. Effect of weight and apllied force on the friction coefficient ofthe spunlace nonwoven fabrics. Fibers and Polymers, 11(2):277–284, 2010. → page 13118[31] J. Klintworth and W. Stronge. Elasto-plastic yield limits and deformation laws for transverselycrushed honeycombs. International Journal of Mechanical Sciences, 30(3-4):273–292, 1988. →pages 9, 11[32] J. Kratz and P. Hubert. Processing out-of-autoclave honeycomb structures: Internal core pressuremeasurements. Composites Part A: Applied Science and Manufacturing, 42(8):1060–1065, 2011.→ page 19[33] J. Kratz and P. Hubert. Anisotropic air permeability in out-of-autoclave prepregs: Effect onhoneycomb panel evacuation prior to cure. Composites Part A: Applied Science andManufacturing, 49:179–191, 2013. → pages 18, 19[34] Y. R. Larberg and M. A˚kermo. On the interply friction of different generations of carbon/epoxyprepreg systems. Composites Part A: Applied Science and Manufacturing, 42(9):1067–1074,2011. → pages x, 12, 13, 14, 16, 99[35] C. Martin, J. Seferis, and M. Wilhelm. Frictional resistance of thermoset prepregs and itsinfluence on honeycomb composite processing. Composites Part A: Applied Science andManufacturing, 27(10):943–951, 1996. → pages 13, 15, 16, 17, 18, 38, 92, 99, 104, 108[36] C. Martin, J. Putnam, B. Hayes, J. Seferis, M. Turner, and G. Green. Effect of impregnationconditions on prepreg properties and honeycomb core crush. Polymer composites, 18(1):90–99,1997. → page 15[37] C. J. Martin and J. C. Seferis. Effect of prepreg resin composition on honeycomb core crush.Materials and process affordability- Keys to the future, pages 366–375, 1998. → pages 15, 17, 18[38] S. D. Papka and S. Kyriakides. In-plane compressive response and crushing of honeycomb.Journal of the Mechanics and Physics of Solids, 42(10):1499–1532, 1994. → pages 7, 9[39] S. D. Papka and S. Kyriakides. Experiments and full-scale numerical simulations of in-planecrushing of a honeycomb. Acta materialia, 46(8):2765–2776, 1998. → pages 7, 9[40] T. Pelton, T. Schneider, R. Martin, and B. C. Airplane. Material factors influencing compositepart producibility in relation to prepreg frictional measurement. In INTERNATIONAL SAMPETECHNICAL CONFERENCE, volume 31, pages 463–477, 1999. → pages 15, 17, 18[41] D. D. Pham, M. W. Tollan, R. M. Outzen, J. C. Lockleer, and C. G. Harris. Composite plystabilizing method, July 23 2013. US Patent 8,491,743. → page 17[42] D. Renn, T. Tulleau, J. Seferis, R. Curran, and K. Ahn. Composite honeycomb core crush inrelation to internal pressure measurement. Journal of advanced materials, 27(1):31–40, 1995. →pages 14, 17, 69, 78, 89, 92, 93, 105[43] G. S. Springer. Resin flow during the cure of fiber reinforced composites. Journal of CompositeMaterials, 16(5):400–410, 1982. → page 14[44] G. Stachowiak and A. W. Batchelor. Engineering tribology. Butterworth-Heinemann, 2013. →page 12[45] J. Sun, M. Li, Y. Gu, D. Zhang, Y. Li, and Z. Zhang. Interply friction of carbon fiber/epoxyprepreg stacks under different processing conditions. Journal of Composite Materials, 48(5):515–526, 2014. → page 12119[46] S. S. Tavares, V. Michaud, and J.-A. Ma˚nson. Through thickness air permeability of prepregsduring cure. Composites Part A: Applied Science and Manufacturing, 40(10):1587–1596, 2009.→ pages 18, 19[47] S. S. Tavares, N. Caillet-Bois, V. Michaud, and J.-A. Ma˚nson. Non-autoclave processing ofhoneycomb sandwich structures: Skin through thickness air permeability during cure.Composites Part A: Applied Science and Manufacturing, 41(5):646–652, 2010. → page 19[48] S. S. Tavares, N. Caillet-Bois, V. Michaud, and J.-A. Ma˚nson. Vacuum-bag processing ofsandwich structures: Role of honeycomb pressure level on skin–core adhesion and skin quality.Composites Science and Technology, 70(5):797–803, 2010. → pages 18, 19[49] J. Zhang and M. Ashby. The out-of-plane properties of honeycombs. International journal ofmechanical sciences, 34(6):475–489, 1992. → pages 5, 7[50] N. Zobeiry and C. Duffner. Measuring the negative pressure during processing of advancedcomposites. Composite Structures, 203:11–17, 2018. → page 14120Appendix AAppendix A Data AcquisitionA.1 LVDTsThe LVDTs were AC-Operated Short-Stroke Unguided Armature models. They are part of Daytronic’sSUH LVDT series. The specific models were DS1000SUH. “Unguided” refers to the fact that the corerod was not spring loaded within the solenoid housing. Rather, it relied on gravity for displacement.LVDTs included a two meter cable containing four separate wire types. Namely, a wire pair for exci-tation voltage, a ground wire pair, a signal wire pair, and a sensor wire pair. The wires were connectedto ports within the autoclave which fed to the outside of the autoclave to connect to the AC conditionermodule.A.1.1 BackgroundThe LVDTs consist of a small rod that slides within a cylindrical housing. The rod itself is nonmagneticexcept for the tip which consists of a nickel-iron core. The housing has one primary solenoid and twosecondary solenoids. Excitation voltage is applied to the primary coils which induces a voltage in thesecondary coils (connected in series opposition). The secondary coils provide the AC signal that istransmitted to the conditioner module. When the core is in the “null” position, the induced voltagescancel and the output voltage is zero (when calibrated with the conditioner module). However, as thecore displaces relative to the coils, the induced voltage will increase in one of the secondary coils anddecrease in the other. The magnitude of the net voltage is proportional to the displacement of the rod.Phase polarity is referenced with the excitation voltage which determines the displacement direction.This is transmitted to the conditioner module by sensor wires connected to the positive and negativeterminals of the primary solenoid. The AC conditioner module is necessary to distinguish betweenpositive and negative displacements and filter out residual null voltage effects [1, 14, 15].A.1.2 AC conditioner calibrationThe LVDT conditioner was calibrated prior to experimentation according to the calibration step outlinedin the 5M30 AC LVDT Conditioner product manual [14]. Calibration was performed such that a 6 V121Figure A.1: LVDT wiring to conditioner module. Taken from [14].output signal corresponded to 1 in (25.4 mm) of displacement. The switch states on the conditionermodule were as follows:• Switch 1 - Mode: Voltage• Switch 2 - Volts: 5.0 Vdc• Switch 3 - Current: Set to 4 mA but NA since mode is voltage• Switch 4 - Filter: 100 Hz• Switch 5 - Filter: 100 Hz• Switch 6 - Sync: Master for LVDT 1 and Slave for LVDT 2• Switch 7 - Zero Adj.: Normal• Switch 8 - Range: Low (160-1600 mV/V)• Switch 9 - Phase: AutomaticThe schematic used for wiring is shown in Figure A.1.122A.2 In-core pressure sensorsThe in-core pressure sensors are developed by Convergent Manufacturing Technologies [3]. They arepiezoresistive MEMS pressure sensors utilizing a Wheatstone bridge setup. Specs are provided below.Owing to proprietary reasons, limited information is available.• Bridge resistance: 3.4 kΩ (± 0.4 kΩ)• Operating pressure: 7 bar• Operating temperature: -40◦C to 180◦C• Repeatability: ± 0.2% F.S.The sensors comprise the pressure transducer with four wires corresponding to excitation voltage,ground, positive output voltage, and negative output voltage. Excitation voltage is 5 volts. Outputvoltage corresponds to the level of pressure application and is measured across the Wheatstone bridge.The output voltage is temperature dependent and is correlated to pressure through a specific transferfunction. Each sensor is individually calibrated by Convergent Manufacturing Technologies.A.3 Autoclavable cameraThe in-autoclave camera was developed by Convergent Manufacturing Technologies [3]. Owing toproprietary reasons, limited information is available. The camera was mounted on an 80/20 aluminumframe to capture in-situ video footage of the core movement process. Video data was passed through aport within the autoclave directly to an external Laptop. The free software, VideoVelocity, developedby CandyLabs [2] was used to record and export video footage and real time images. The images werelater compiled on MATLAB and correlated with the sensor data to produce synchronized, real timevideo footage with sensory display. An example of the synchronized footage is provided in Figure A.2.Figure A.2: Video footage synchronized with in-situ sensor data. 95◦C experiment.123Appendix BAppendix B Additional Post-processingImagesB.1 Post-processed panelsImages of each of the panels post processing are provided the below. They are organized according totheir experimental group. That is, the top row represents the “Effect of Temperature” tests, the middlerow “Effect of Pressure”, and the bottom row “Effect of Layup”.124Room temp 95°C 120°Corig 120°C180°C 325 kPa 600 kPaFull tie-down Half tie-down CMS 45° chamferWLFigure B.1: Top-down view of processed panels.B.2 Slope gradient mapsSlope gradient maps are provided below for each sample. Unfortunately, the room temperature samplewas not scanned on the CMM. As such, no slope gradient map exists for this sample.125Figure B.2: Slope gradient map for 95◦C sample. Although not shown, the colour bar applies toeach of the following images. Noise is seen in some of the samples which can be disregarded.Figure B.3: Slope gradient map for 120◦Corig sample.126Figure B.4: Slope gradient map for 120◦C sample.Figure B.5: Slope gradient map for 180◦C sample.127Figure B.6: Slope gradient map for 325 kPa sample.Figure B.7: Slope gradient map for 600 kPa sample.128Figure B.8: Slope gradient map for full tie-down sample.Figure B.9: Slope gradient map for half tie-down sample.129Figure B.10: Slope gradient map for CMS sample.Figure B.11: Slope gradient map for 45◦ chamfer sample.130B.3 Through-thickness section cutsImages of through-thickness section cuts in the L and W direction for each sample are provided below,including those shown previously. Crush features are not identified.131WLFigure B.12: Through-thickness section cut of room temperature sample.LWFigure B.13: Through-thickness section cut of 95◦C sample.132LWFigure B.14: Through-thickness section cut of 120◦Corig sample.LWFigure B.15: Through-thickness section cut of 120◦C sample.133LWFigure B.16: Through-thickness section cut of 180◦C sample.LWFigure B.17: Through-thickness section cut of 325 kPa sample.134WLFigure B.18: Through-thickness section cut of 600 kPa sample.WLFigure B.19: Through-thickness section cut of full tie-down sample.135WLFigure B.20: Through-thickness section cut of half tie-down sample.WLFigure B.21: Through-thickness section cut of CMS sample.136LWFigure B.22: Through-thickness section cut of 45◦ chamfer sample.137B.4 In-plane section cutsThe following section provides additional images of the in-plane section cuts made. Note that cuts werenot made for each sample. All cuts made are displayed below including those previously shown.LWFigure B.23: In-plane section cut of room temperature sample.138LWFigure B.24: In-plane section cut of 120◦C sample.139LWFigure B.25: In-plane section cut of 180◦C sample.140LWFigure B.26: In-plane section cut of 600 kPa sample.141LWFigure B.27: In-plane section cut of full tie-down sample.142LWFigure B.28: In-plane section cut of half tie-down sample between two restrained edges.143LWFigure B.29: In-plane section cut of half tie-down sample between two unrestrained edges.144LWFigure B.30: In-plane section cut of CMS sample.145LWFigure B.31: In-plane section cut of 45◦ sample.146Appendix CAppendix C Additional Data SetsC.1 120◦Corig resultsThe following dataset represents the results from the 120◦Corig experiment (refer to Section 4.5.2). Un-fortunately, LVDT 2 failed during crush resulting in premature flattening out of the W curve. However,the rate of deformation in W can be inferred from the original slope of LVDT 2 following collapse. Thisfigure shows a quicker rate of deformation in L compared to W.00:20:00 00:25:00 00:30:00Time (hh:mm:ss)-1000100200300400500600700Pressure (kPa)302520151050Displacement (mm)Autoclave PressureLVDT 1 - LLVDT 2 - WSensor failedCollapseFigure C.1: Core movement in ribbon and non-ribbon direction for 120◦Corig sample. Autoclavepressure is absolute external pressure. Vacuum pressure is not shown.147

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:
https://iiif.library.ubc.ca/presentation/dsp.24.1-0380364/manifest

Comment

Related Items