Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

An experimental study of fluid flow in a low consistency refiner Mithrush, Troy Lindsay 2013

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

Item Metadata


24-ubc_2014_spring_mithrush_troy.pdf [ 2.34MB ]
JSON: 24-1.0165714.json
JSON-LD: 24-1.0165714-ld.json
RDF/XML (Pretty): 24-1.0165714-rdf.xml
RDF/JSON: 24-1.0165714-rdf.json
Turtle: 24-1.0165714-turtle.txt
N-Triples: 24-1.0165714-rdf-ntriples.txt
Original Record: 24-1.0165714-source.json
Full Text

Full Text

AN EXPERIMENTAL STUDY OF FLUID FLOW IN A LOWCONSISTENCY REFINERbyTroy Lindsay MithrushB.Eng. Aerospace Engineering, Carleton University, 2011A THESIS SUBMITTED IN PARTIAL FULFILLMENTOF THE REQUIREMENTS FOR THE DEGREE OFMaster of Applied ScienceinTHE FACULTY OF GRADUATE AND POSTDOCTORAL STUDIES(Mechanical Engineering)The University Of British Columbia(Vancouver)November 2013c? Troy Lindsay Mithrush, 2013AbstractTransport phenomena inside a low consistency disc refiner were experimentally investigated. Atransparent refiner door was designed and fabricated with four acrylic viewports enabling plate-scale and groove-scale visual observation. High speed video, ultra-violet fluorescent tracer parti-cles and a MATLAB program were used to perform particle tracking velocimetry to gain furtherunderstanding of the flow field. The experimental working fluid under study was water. The effectsof refiner operating parameters on the flow field were of particular interest. Refiner flow rates werevaried from 300 to 700 litres per minute. Refiner rotational speeds were varied from 400 to 1200RPM. Plate gap values under study included 7.5, 2.5, 1.5, and 0.75 mm. Two plate configurationswere studied, including a smooth rotor and grooved rotor with a machined acrylic stator plate. Theplate geometry under test was designed for softwood pulp having a bar edge length equal to 0.99km/rev.A set of phenomenological characterizations of observed particle behaviour was identified.Qualitative results were provided for the effect of gap, refiner speed, and flow rate on the flowfield. Lagrangian pathlines were shown to reveal tortuous flow for grooved rotor experiments.Quantitative results were presented for grooved rotor experiments for gaps of 0.75 mm. Eulerianmeasurements of groove axial velocity indicated fluid transport into and out of the stator grooves,while net transport occurred out of the grooves. The presence of backflow in the stator grooves wasobserved at all operating points for the grooved rotor under test. The relationship between statorbackflow velocity and operating parameters was reported showing an increase with refiner speedand a minimal decrease with refiner flow rate. It has been shown that there is a linear relationshipbetween stator backflow velocities and the pressure differential across the refiner. Rotational mo-tion in the stator grooves was quantified by angular velocity and turnover rate of the fluid. Turnoverrate was defined as the number of rotations of the fluid as it travels the length of the groove. Angu-lar velocity increased proportionally with refiner speed and turnover rate did not vary significantlywith refiner operating parameters.iiPrefaceThis thesis is original, unpublished work by Troy Mithrush. The research program was proposedby Dr. James Olson and Dr. Mark Martinez. The research, including experimental design, experi-mental procedures and data analysis was performed by the author.The tracking algorithm implemented in the PPC MATLAB Image Processing and ParticleTracking Velocimetry Program was originally developed by J. Crocker, D. Grier and E. Weeksand made available in MATLAB by D. Blair and E. Dufresne. Segments of code used to condi-tion input and output data for the tracking algorithm were developed by a research colleague, J.Mackenzie. The data flow architecture, image processing, and data analysis was sole work of theauthor.Journal SubmittalsBelow is a list of co-authored journal submittals. Contributions included performing compu-tational fluid dynamics (CFD) simulations, results analysis, and contributions to written content.This work is referenced in Chapter 1 and Chapter 4.1. Rajabi Nasab, N., Mithrush, T., Olson, J.A. & Martinez, D.M. (2013), Turbulent Couetteflow between two parallel corrugated walls: The case with motion of one wall perpendicularto the corrugation cavities, accepted for publication in the Canadian Journal of ChemicalEngineering.This publication presents a CFD study on the flow field in the cross-section of a low consis-tency refiner.2. Rajabi Nasab, N., Mithrush, T., Olson, J.A. & Martinez, D.M. (2013), On the relationshipbetween plate pattern and the flow field in LC refiners: Insight into the groove depth effectand no-Load power, submitted for publication.In this publication the flow field of a low consistency refiner was investigated using CFDmodelling to study the effect of groove depth on no-load power.iiiTable of ContentsAbstract . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iiPreface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . iiiTable of Contents . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ivList of Tables . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viiList of Figures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . viiiNomenclature . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiiAcknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiv1 Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11.1 Working Fluid in a Low Consistency Refiner . . . . . . . . . . . . . . . . . . . . . 11.2 Geometry and Operation of an LC Refiner . . . . . . . . . . . . . . . . . . . . . . 21.3 Fibre Treatment and The Refining Action . . . . . . . . . . . . . . . . . . . . . . 41.4 Fibre and Pulp Transport . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61.4.1 Gross Refiner Flow . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61.4.2 Groove Cross-sectional Flow . . . . . . . . . . . . . . . . . . . . . . . . . 71.4.3 Fibre Capture . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 91.5 Background Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101.6 Research Objectives and Thesis Organization . . . . . . . . . . . . . . . . . . . . 112 Research Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122.1 Experimental Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12iv2.1.1 Refining Facility . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122.1.2 Modified Refiner Door . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132.1.3 Experimental Setup . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182.1.4 Experimental Test Matrix . . . . . . . . . . . . . . . . . . . . . . . . . . 192.1.5 Particle Tracking Velocimetry . . . . . . . . . . . . . . . . . . . . . . . . 222.2 Analysis Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 252.2.1 Approach . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 252.2.2 Image Processing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 262.2.3 Particle Tracking Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . 302.2.4 Instantaneous Velocity Measurements . . . . . . . . . . . . . . . . . . . . 302.2.5 Lagrangian Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 312.2.6 Eulerian Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 312.2.7 Coordinate Transformations . . . . . . . . . . . . . . . . . . . . . . . . . 322.2.8 Calibration Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 332.2.9 Uncertainty Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 343 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 353.1 Qualitative Observations of Particle Behaviour . . . . . . . . . . . . . . . . . . . . 353.1.1 Phenomenological Characterizations and Observations . . . . . . . . . . . 353.1.2 Qualitative Observations . . . . . . . . . . . . . . . . . . . . . . . . . . . 393.2 Effect of Operating Parameters on Stator Groove Velocities . . . . . . . . . . . . . 453.2.1 Velocity Distributions at Cross-sections of Interest . . . . . . . . . . . . . 453.2.2 Axial Velocity Profile Along Stator Grooves . . . . . . . . . . . . . . . . . 463.2.3 Bulk Axial Velocity Groove Estimates versus Refiner Operating Parameters 483.2.4 Discussion of Stator Groove Velocity Results . . . . . . . . . . . . . . . . 493.3 Effect of Refiner Operating Parameters on Groove Rotational Flow . . . . . . . . . 533.3.1 Results and Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 534 Conclusions and Recommendations . . . . . . . . . . . . . . . . . . . . . . . . . . . 564.1 Conclusions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 564.2 Strengths and Limitations of Research . . . . . . . . . . . . . . . . . . . . . . . . 574.3 Recommendations for Future Work . . . . . . . . . . . . . . . . . . . . . . . . . . 57Bibliography . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 59vA Preliminary Mechanical Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63A.1 Preliminary Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63A.2 Design Considerations for Acrylic as an Engineering Material . . . . . . . . . . . 64A.2.1 Material Selection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64A.2.2 Planar Disc Windows . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65A.2.3 Crazing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65A.2.4 Water Exposure Effects . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65B Mechanical Design Drawings . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66C Structural Finite Element Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76D Industrial Operating Region Data . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77E Groove Axial Velocity Plots . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79F Fluid Transport Justification . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87viList of TablesTable 2.1 Experimental test plan for flow rate, refiner speed and plate gap. . . . . . . . . . 20Table 2.2 Summary of UV fluorescent tracer particle properties provided by CosphericInnovations in Microtechnology. . . . . . . . . . . . . . . . . . . . . . . . . . 24Table B.1 Mechanical Design Drawing List . . . . . . . . . . . . . . . . . . . . . . . . . 66viiList of FiguresFigure 1.1 LC refiner model showing important operational features including rotor andstator plates, refiner inlet and outlet, gap actuator and electric motor. . . . . . 2Figure 1.2 Single disc refiner plate geometry. (a) is a frontal view of a simplified plategeometry showing the sector angle ? and the bar angle ? . (b) is the cross-sectional area of the rotor and stator plates depicting important dimensions. Uis the translational velocity of the rotor over the stator at radius R in the 2Drepresentation. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3Figure 1.3 (a) Un-refined chemical fibres showing intact cell walls. (b) Fibrillated andcollapsed chemical fibres following the refining process, reproduced with per-mission. (M Polan (1993)) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4Figure 1.4 Gross refiner flow transport as depicted by Fox et al. (1979). In (a) the exitregion defines a zone where stator flow is outward and the circulation regiondefines a zone where stator flow is inward. In (b) the circulation pattern as aresult of groove flow behaviour is shown. . . . . . . . . . . . . . . . . . . . . 7Figure 1.5 Flow transport in a two-dimensional cross-section of a disc refiner as depictedby Fox et al. (1979). ?? and ?+? represent stator backflow and outward radialflow in the direction of the groove, respectively. . . . . . . . . . . . . . . . . . 8Figure 2.1 UBC variable speed 16 inch LC refiner. . . . . . . . . . . . . . . . . . . . . . 13Figure 2.2 An external view of the MRD (left) and an internal view of the MRD (right)installed in monitoring operation mode. . . . . . . . . . . . . . . . . . . . . . 14Figure 2.3 Monitoring operation: A double-pane acrylic window design enables observa-tion via four viewports revealing 75% of the refining zone. . . . . . . . . . . . 16Figure 2.4 The acrylic window machined flat shown on a Haas milling machine. . . . . . 16Figure 2.5 The machined acrylic stator plate shown on a Haas milling machine. . . . . . . 17viiiFigure 2.6 Experimental setup diagram showing interactions between the test environ-ment, DACS and data processor. . . . . . . . . . . . . . . . . . . . . . . . . . 19Figure 2.7 Plate geometries under investigation. (a) is a frontal view of the grooved rotor,(b) is the plate cross-section of the smooth rotor configuration and (c) is theplate cross-section of the grooved rotor configuration. . . . . . . . . . . . . . . 20Figure 2.8 A frontal view of the modified refiner door showing acrylic viewports and thestainless steel housing (left). The viewport layout is labelled: U (upper win-dow), R (right window), L (lower window) and L (left window). A photo ofthe mounted camera and the installed MRD (right). . . . . . . . . . . . . . . . 21Figure 2.9 Industrial operating region for refiner flow rates, the 16 inch refiner valueswere extrapolated from the resulting industrial flow rate curves from differentmanufacturers. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22Figure 2.10 Industrial operating region for refiner speeds, and the experimental operatingregion investigated for the 16 inch refiner. . . . . . . . . . . . . . . . . . . . . 23Figure 2.11 Primary and flow groove regions of interest and reference frame definition. . . 26Figure 2.16 Cross-sections of interest equally spaced along the length of the grooves wereused to calculate Eulerian velocity measurements. . . . . . . . . . . . . . . . . 32Figure 2.17 Image frame of reference (x,y) and the groove reference frame in terms of theaxial A and tangential T coordinates. . . . . . . . . . . . . . . . . . . . . . . . 33Figure 3.1 Lagrangian trajectories of the particles circled in blue following their pathlinesthrough the refining zone. A rotor bolt hole is circled in black. The raw image(left) and the spatial Laplacian high pass filtered image (right) are shown withthe stator bar configuration superimposed. . . . . . . . . . . . . . . . . . . . . 36Figure 3.2 Illustration of smooth rotor characterizations for helical stator groove flow (1),groove departure (2), gap flow (3), groove entry (4) and bolt hole entrainment(5). The observed camera view from the underside of the stator plate is shownin (a) and the predicted particle behaviours in the 2D groove cross-section aredepicted in (b). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37ixFigure 3.3 Illustration of grooved rotor observations for helical stator groove flow (1),groove departure (2), gap flow (3), groove entry (4), bolt hole entrainment (5)and helical rotor groove flow (6). The observed camera view from the under-side of the stator plate is shown in (a) and the predicted particle behaviours inthe 2D groove cross-section are depicted in (b). Note that helical rotor grooveflow is portrayed in the rotor reference frame. . . . . . . . . . . . . . . . . . 38Figure 3.4 Smooth rotor particle trajectories at constant flow rate (Q= 400LPM), constantgap (G = 0.75mm) and refiner speeds of 400 RPM (a), 800 RPM (b) and 1200RPM (c). The refiner inlet is located on the left of each image and the rotormotion is upwards. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40Figure 3.5 Smooth rotor particle trajectories at constant flow rate (Q= 400LPM), constantrefiner speed (N = 800) and plate gaps of 7.5 mm (a), 2.5 mm (b) and 0.75 mm(c). The refiner inlet is located on the left of each image and the rotor motionis upwards. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 41Figure 3.6 Spatial (above) and temporal (below) Lagrangian trajectories showing the fre-quency shift along a primary groove at G=0.75 mm, N = 800 RPM, and Q =500 LPM. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42Figure 3.7 Spatial (above) and temporal (below) Lagrangian trajectories showing the fre-quency shift along a flow groove at G=0.75 mm, N = 800 RPM, and Q = 500LPM. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43Figure 3.8 Spatial (above) and temporal (below) Lagrangian trajectories showing the pres-ence of corner eddies at G=0.75 mm, N = 800 RPM, and Q = 500 LPM. . . . . 44Figure 3.9 Histogram of groove velocities at mid-length of a primary groove for the oper-ating point: G=0.75 mm, N=800 RPM and Q=300 LPM. . . . . . . . . . . . . 45Figure 3.10 Scatter plot of tangential velocity VT versus axial velocity VA at mid-length fora primary groove at the operating point: G=0.75 mm, N=800 RPM and Q=300LPM. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 47Figure 3.11 Bulk axial velocity versus groove position for primary groove at the operatingpoint: G=0.75 mm and Q=300 LPM. . . . . . . . . . . . . . . . . . . . . . . . 48Figure 3.12 Bulk axial velocity versus groove position for flow groove at the operatingpoint: G=0.75 mm and Q=300 LPM. . . . . . . . . . . . . . . . . . . . . . . . 49Figure 3.13 Primary groove axial velocity versus flow rate for lines of constant refiner speed. 50Figure 3.14 Primary groove axial velocity versus refiner differential pressure. Linear trend-line R2 = 0.99. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 51xFigure 3.15 Temporal Lagrangian trajectory showing an example period measurement atG=0.75 mm, N = 800 RPM, and Q = 500 LPM. . . . . . . . . . . . . . . . . . 53Figure 3.16 Groove angular velocity versus refiner speed. . . . . . . . . . . . . . . . . . . 54Figure 3.17 Turnover rate versus refiner speed with lines of constant flow rate. . . . . . . . 55Figure E.1 Histogram of groove velocities at mid-length of a flow groove at the operatingpoint: G=0.75 mm, N=800 RPM and Q=300 LPM. . . . . . . . . . . . . . . . 79Figure E.2 Scatter plot of tangential velocity vt versus axial velocity va at mid-length of aflow groove at the operating point: G=0.75 mm, N=800 RPM and Q=300 LPM. 80Figure E.3 Bulk axial velocity versus groove position for primary groove at the operatingpoint: G=0.75 mm, and Q=500 LPM. . . . . . . . . . . . . . . . . . . . . . . 81Figure E.4 Bulk axial velocity versus groove position for primary groove at the operatingpoint: G=0.75 mm, and Q=700 LPM. . . . . . . . . . . . . . . . . . . . . . . 82Figure E.5 Bulk axial velocity versus groove position for flow groove at the operatingpoint: G=0.75 mm, and Q=500 LPM. . . . . . . . . . . . . . . . . . . . . . . 83Figure E.6 Bulk axial velocity versus groove position for flow groove at the operatingpoint: G=0.75 mm, and Q=700 LPM. . . . . . . . . . . . . . . . . . . . . . . 84Figure E.7 Flow groove axial velocity versus flow rate for lines of constant refiner speed. . 85Figure E.8 Flow groove axial velocity versus refiner differential pressure with linear trend-line. Linear trend-line R2 = 0.96. . . . . . . . . . . . . . . . . . . . . . . . . 86Figure F.1 Simplified groove coordinate system. . . . . . . . . . . . . . . . . . . . . . . 87xiNomenclatureA Groove axis coordinate ?P Differential pressure? Average intersecting bar angle ? Plate bar anglebx? Systematic standard uncertainty rc Critical radiusB Refiner plate bar width re Escape radiusC Consistency rx? Random standard uncertaintydp Particle diameter R,r Plate radiusD Refiner plate groove depth Re Reynolds numberDi Inner plate diameter Ri Inner plate radiusDo Outer plate diameter Ro Outer plate radiusG Refiner plate gap R2 Linear coefficient of determinationI Fibre impact intensity ? f Fluid densityQ Refiner flow rate ?p Particle densityLG Groove length St Stokes numberm? Mass flow rate sx? Standard deviation of the meansmf Dry fibre mass t Timemw Water solution mass t? ,P Student t-valueMc Circularity metric ??t Mean signal periodn Number of plate bars T Groove tangential coordinatenr Number of rotor bars Ti Pixel threshold valuens Number of stator bars ? f Fluid characteristic time scaleN Refiner speed ?p Particle relaxation timeNc Crowding number ? Turnover rateNI Number of impacts ? Plate sector anglePi Mean pixel value ux? Combined uncertainty of the meanPnet Net power U Rotor velocityxii? Dynamic viscosityVA Groove axial velocityVT Groove tangential velocityW Refiner plate groove width? Angular velocityx? Sample set meanxiiiAcknowledgementsI would like to express my utmost gratitude to Dr. James Olson and Dr. Mark Martinez as myresearch supervisors for providing valuable guidance and support throughout my research. Fur-thermore, I would like to acknowledge my appreciation to everyone who contributed to the work:The engineers and machinists from Aikawa Iron Works, in particular Keiji Yasuda, for theircontributions to the mechanical design and fabrication of the modified refiner door.Dr. Jens Heymer from Aikawa Fibre Technologies for providing important discussions andadvice that was integral to experimental design and test procedures.Dr. Richard Kerekes for providing advice on the practicality and applicability of experimentalmethods and results.Colleagues and staff at The Pulp and Paper Centre, who have been of great assistance; namely,Frank Saville, Jordan Mackenzie, Fatehjit Singh, Nici Darychuk and George Soong.Support staff in the Mechanical Engineering Department, such as Markus Fengler.Alain Mercier at All Points Machining for the excellent workmanship and cooperation in ma-chining the acrylic components required for this project.Finally, I would like to thank Canfor Pulp Limited Partnership and the Natural Sciences andEngineering Research Council of Canada for their financial contributions and interest in this re-search.xivChapter 1IntroductionLow consistency refining is a process in which a low consistency fibre suspension is modified by abeating mechanism with means to favorably alter paper properties. In its history, papermaking hasalways incorporated some means of refining fibres: from its genesis when plant stock was manuallybeaten, to the first proficient Hollander beater used by the Dutch in the 1600?s. Refining is wellestablished as a process that is beneficial to resultant fibre properties and the end paper product.Today pulp refining is commonplace in papermaking and has become a large focus of research inmany institutions and businesses worldwide.Pertinent background information is provided in the following sections to outline the prob-lem with respect to low consistency refining and fluid mechanics. The background precedes theresearch objectives and an overview of the organization of this thesis.1.1 Working Fluid in a Low Consistency RefinerLow consistency (LC) refining is performed on wood fibres suspended in water, typically at 3-6%consistency. Consistency is a measure of dry fibre mass in relation to the total suspension massgiven as:C =Dry f ibre massTotal suspension mass?100% =mfmf +mw?100% (1.1)At these consistencies a fibre suspension, termed pulp, naturally flocculates to create a fibrenetwork. Much research has been carried out on fibre networks to characterize the degree of floc-culation and network development. A common parameter used to define network flocculation isthe crowding factor, Nc, as defined by Kerekes and Schell (1992). In early work by Mason (1950)fibre suspensions with Nc ? 1 were identified as a critical concentration where collisions first oc-1cur between fibres in shear flow. Martinez et al. (2001) identified a gel crowding number wheresuspensions with Nc < 16 essentially act dilute. Furthermore, Kerekes and Schell showed that forNc > 60, fibres have 3 or more contacts per fibre. Once the fibre network becomes sufficiently en-tangled, it becomes a load bearing network exhibiting a yield stress which must be exceeded beforethe fluid flows. Once the yield stress is surpassed, shear-thinning behaviour is exhibited and in highshear, typically present in LC refining, fibre suspensions become turbulent or fluidized. That is, thesuspension acts as a continious medium, Newtonian in nature, with an apparent viscosity (Kerekes(1983), Bennington and Kerekes (1996)). For a thorough review of fibre suspension rheology referto the review by Derakhshandeh et al. (2011).In recent numerical studies of fluid flow in a refiner by Wittberg et al. (2012) and Kondora andAsendrych (2013) fibre suspensions were simulated as Newtonian fluids. Wittberg et al. confirmedthat the numerical results were similar to experimental results as obtained by visually observingthe refining zone in an LC refiner.1.2 Geometry and Operation of an LC RefinerThe mechanical devices which perform the beating action inside LC refiners are typically groovedfillings in either disc, conical or cylindrical form. Of specific focus in this work is the single discrefiner as shown in Figure 1.1. The fibre suspension flows into the refiner at the pulp inlet, betweenthe rotor and stator plates, and out via the pulp outlet. The flow is pump driven, in addition to therefiner hydraulically functioning as a pump.Electric MotorRemotely ActuatedGap Controller Rotor StatorPulp OutletPulp InletFigure 1.1: LC refiner model showing important operational features including rotor and sta-tor plates, refiner inlet and outlet, gap actuator and electric motor.2For a simple single disc refiner, the governing process variables that define machine operationare: flow rate Q, plate gap G and refiner rotational speed N. Secondary variables are parametersthat play a large role in refining but are not directly controllable by the LC refiner, those beingpower, fibre properties, consistency, plate design parameters and temperature. Note that for thepurpose of research, motor power is indirectly controlled via plate gap and refiner speed, whereasin industry, motor power is generally the controlled variable. Dependent variables are defined asparameters that are dependent on either process or secondary variables, including no-load power,specific refining energy, and differential pressure ?P. Output variables are the measured fibre andpaper properties following the refining process. Examples of output variables are fibre length,freeness, and tensile strength.The refining action entails a rotor disc spun in close proximity to the stator in order to achievegaps on the order of 100 ?m, or 2 to 5 fibre diameters. The desired morphological changes tofibres are achieved by repeated bar crossing impacts resulting in cyclic compression, shear andtensile loading.U = R?(a) (b)DiDoFigure 1.2: Single disc refiner plate geometry. (a) is a frontal view of a simplified plate geom-etry showing the sector angle ? and the bar angle ? . (b) is the cross-sectional area of therotor and stator plates depicting important dimensions. U is the translational velocity ofthe rotor over the stator at radius R in the 2D representation.A typical disc refiner may have a plate geometry as depicted in Figure 1.2. Refiner plates havean opposing groove pattern on the rotor and stator with repeating sectors around the plate. Of3geometric interest are the sector angle ? , bar angle ? , plate outer diameter Do, plate inner diameterDi, groove width W , groove depth D, bar width B, and plate gap G.1.3 Fibre Treatment and The Refining ActionThe main purpose of LC refining is to alter fibre morphology to achieve increased strength andsmoothness for the resultant paper. Morpholical changes to fibres occur through external fibrillationand fibre collapse as can be seen from the unrefined and refined fibres in Figures 1.3a and 1.3b,respectively. These changes lead to increased bonding area during papermaking to form higherbond strength, and thus paper strength.Figure 1.3: (a) Un-refined chemical fibres showing intact cell walls. (b) Fibrillated and col-lapsed chemical fibres following the refining process, reproduced with permission. (MPolan (1993))Numerous refining models have been developed in order to model the refining action and pre-dict refining outcomes. Refining theories are introduced next in order to show their dependencyon the process, secondary and dependent variables introduced in Section 1.2. The following theo-ries provide parameters that may be related to fibre output properties and are helpful in achievingconsistent results.The refining action is related to output variables most commonly by the specific refining energy(SRE), a value used to define the amount of energy expended upon a specific mass of fibres during4refining and is given by:SRE =Pnetm?(1.2)Typical SRE values for chemical pulp refining range from 80 to 250 kWh/t depending on desiredfibre and paper properties. The following theories define a machine intensity value with no speci-fication of exactly how the energy is expended on the fibres.One of the first successful attempts to characterize refining was the specific edge load (SEL)theory as introduced by Wultsch and Flucher (1958). The more renowned variation of SEL wasexpanded upon by Brecht and Siewert (1966) by demonstrating the domininant impact of the barcrossing. SEL is given by:SEL =Pnet?BEL=PnetCEL(1.3)where BEL is the bar edge length in km/rev, a physical parameter to describe a plate geometrycalculated by mutiplying the number of rotor bars nr by the number of stator bars ns as a functionof radius r and integrating over the refining zone.BEL =? RoRinr(r)ns(r)cos?dr ?n?1nrns?rcos?(1.4)The specific surface load (SSL) theory is an extension of SEL theory developed by Lumiainen(1990). Lumiainen studied the effect of bar width while maintaining the cutting edge length (CEL)and determined that bar width played a role in the refining process during an edge-to-surface phasepresent during a bar crossing.Meltzer and Rautenbach (1994) developed the modified edge load (MEL) theory which in-troduced bar width B, groove width W and average intersecting angle of stator and rotor bars ?as:MEL = SEL12tan?B+WB(1.5)Theories have also been proposed based on a two parameter model to characterize a fibreintensity; one parameter to describe the intensity of fibre impacts I and the other to quantify thenumber of impacts NI on the fibre during refining. Leider and Nissan (1977) came up with an indepth derivation for the number of impacts per fibre based on speed of rotation, residence time fora given fibre, and plate geometry such as groove and bar width and depth, plate inner and outerdiameter and flow rate. In order to address some of the shortcomings of this theory Kerekes (1990)developed the C-factor, which estimated values for both intensity I and number of impacts NI .The C-factor approach took into account bar and groove geometry, refiner speed, consistency, fibre5length, coarseness and probability of fibre capture.Other theories have been developed to describe the refining intensity based on refining forces,see Kerekes and Senger (2006) and Kerekes (2011). In addition, refining has been modelled as ahydrodynamic process by Radoslavova et al. (1997) and Roux and Joris (2005), as a lubricationprocess by Rance (1951), Steenberg (1951), and Frazier (1988) and as turbomachinery from amechanical disk friction process by Herbert and Marsh (1968).1.4 Fibre and Pulp TransportWhile the fibre treatment and the refining action have been of great interest in many previousstudies, the efficiency of the refining action is also dependent on fibre and pulp transport in therefiner.1.4.1 Gross Refiner FlowGross refiner flow refers to macro-scale flows including plate-scale circulation patterns and axialgroove flow. The rotational motion of the rotor creates high velocity outward flow in the rotorgrooves due to an added centrifugal body force. Often LC refiners operate as a pump, creating apositive pressure differential across the refiner, however, at other times there may be pressure losses.Halme (1962) made visual observations for water in a conical refiner and was one of the earliestcontributors to refiner transport phenomena, reporting radially inward stator flow. Inward statorflow is caused by an increased housing pressure at the plate periphery. This induces an adversepressure gradient across the stator grooves which results in inward flow (Figure 1.4). Inward statorflow will be referred to as stator backflow in this work. Later contributors also observed statorbackflow (Banks (1967), Herbert and Marsh (1968), Fox et al. (1979), Lumiainen (1994), andWittberg et al. (2012)).Fox et al. (1979) observed stator backflow in what was arbitrarily defined as the circulationregion. An exit region was also defined in proximity to the outlet zone where outward radial flowwas observed in both the rotor and stator (Figure 1.4).Halme (1962) was the first to report the presence of a critical radius rc at each refiner operatingpoint beyond which flow reversed and the stator backflow condition existed. Lumiainen later notedthat in a conical refiner, at low flow rates, increased backflow was observed yet at high flow rates,backflow was nearly non-existent. In recent work by Wittberg et al. (2012) it was also observedthat backflow velocity increased with refiner speed and decreased with flow rate.6(a) (b)ExitRegionOutletCirculationRegionNInletFigure 1.4: Gross refiner flow transport as depicted by Fox et al. (1979). In (a) the exit regiondefines a zone where stator flow is outward and the circulation region defines a zonewhere stator flow is inward. In (b) the circulation pattern as a result of groove flowbehaviour is shown.1.4.2 Groove Cross-sectional FlowFox et al. (1979) presented a thorough description of the rotational flow present in rotor and statorgrooves as depicted in Figure 1.5. Motion of the rotor over the stator in cylindrical, conical ordisc refiners induces secondary vortical flow in both rotor and stator grooves by means of viscousmomentum transport. The vortex that forms, coupled with axial groove flow, creates spiral flow inthe form of a helix. Corner eddies form as a result of the 90 degree corners present in the refinergrooves, shown as tertiary vortices in Figure 1.5. In addition, Fox et al. noted the existence ofa tertiary wiping flow caused by a pressure gradient between the rotor and stator grooves. Thiswiping flow was said to aid in stapling fibres to the rotor bar leading edges.The 2D representation of cross-sectional flow observed in LC refiner plates closely resemblesthat of lid-driven cavity flows. Lid-driven cavity flow is an area of study in fluid mechanics thatis often used as a benchmark for computational fluid dynamic studies. For this reason, it has beenthoroughly studied and Ghia et al. (1982) is a well known contributor to this topic. For a review oflid-driven cavity flows, refer to a thorough survey given by Shankar and Deshpande (2000).7UTertiary Flow Secondary Flow Primary FlowFigure 1.5: Flow transport in a two-dimensional cross-section of a disc refiner as depicted byFox et al. (1979). ?? and ?+? represent stator backflow and outward radial flow in thedirection of the groove, respectively.Numerous studies have been performed on the effects of Reynolds number on lid-driven cavityflows. Pan and Acrivos (1967) found that for finite sized cavities and steady flow at Reynoldsnumber equal to 4000, an inviscid core of uniform vorticity existed. Viscous effects were confinedto the thin shear boundary layer at the wall boundaries. For cavities of larger aspect ratio it wasnoted that viscous and inertial effects remained of significant magnitude throughout the domain.Pan and Acrivos also showed that when the Reynolds number approached infinity, the primaryvortex extended throughout the entire cavity with finite aspect ratio. That is to say, theoretically,corner eddies will become infinitessimally small and thin boundary layers will develop along thewall. At higher Reynolds numbers between 7,000 and 11,000 regions of unsteady periodic flowexist and at Reynolds greater than 11,000, a turbulent chaotic regime sets in, as reported by Penget al. (2003).Nasab et al. (2013b) studied the effect of groove depth on the vortex flow using numerical2D simulations of a refiner cross-section. In this work it was found that as the groove aspectratio increased, multiple vortices formed in the groove under refining operating conditions. Giventhe presence of multiple primary vortices, Moffat (1964) showed mathematically that the relative8intensity of adjacent vortices is large, and for corner eddies this intensity falls rapidly as the corneris approached.The first to quantify the vortical motion in a conical refiner was Lumiainen (1994), who as-sumed water studies accurately depicted pulp flow behaviour. Using laser doppler anemometry(LDA) the tangential groove velocity was determined at 9 groove depths revealing the magnitudeof rotational flow in the stator groove. Wittberg et al. (2012) showed multiple vortex formationin both experimental and numerical work. Experimentally a port hole was used to view from thebottom of the stator grooves in which high speed videos were captured of a 3.5% consistency fi-bre suspension at operational refining gaps. Wittberg et al. (2012) also observed highly rotationalhelical flow for hardwood pulp in a coarse plate pattern, whereas softwood pulp exhibited a largeflow velocity in the direction of the groove with minimal rotation. Note that this was a coarse platepattern designed for softwood pulp with longer fibres.Wittberg et al. also performed 3D simulations with a geometry resembling a lid-driven cavitywith a small gap between the driving wall and top of the cavity. A pressure differential was appliedalong the length of the groove in order to induce groove axial flow. It was shown that whenmultiple vortices form in the groove, fibres may get trapped in the lower vortex and thus would notbe exposed to the upper region of the grooves where the refining potential exists.1.4.3 Fibre CaptureIt was proposed by Smith (1922) that fibres collect on bar edges in a definite formation and presentuniform fibrage. Increased fibre length and consistency resulted in an increased amount of fibragedevelopment on bar leading edges. Fibrage was defined as the fibre mat that forms on a bar movingin a suspension. Since the work of Smith there have been both confirmations of fibrage growthon refiner bars and in contradiction, only the presence of flocs on refiner bars. Page et al. (1962)showed no sign of fibrage as some bar crossing instances revealed the beating and breaking downof fibre flocs.The influence of flow field in the refiner grooves has been insinuated in several works (Foxet al. (1979), Wittberg et al. (2012)). It has been proposed by Wittberg et al. that the rotational flowin the groove imposes a downward shear force near the bar leading edges aiding in fibre transportand capture. In fact, this is where both the fibrage and flocs introduced above are captured in therefiner. The exact impact of the flow field on fibre capture is still relatively unknown and remainsa topic of interest.The existance of an escape radius re similar to the critical radius rc as observed by Halme9(1962) was also identified by Senger et al. (1998) in HC refiners for a 20% consistency suspensionwith no steam. It was observed that flocs remained stapled to a stator bar and moved outwardin short, random steps until a certain radius re was reached. Beyond re the flocs would unstaple,acquire significant tangential velocity and quickly depart from the refining zone. re was found tobe constant for a given refiner speed and decreased when rotational speed was increased, similar tothe stator backflow trend.1.5 Background ReviewIt has been shown that fibre suspension properties, operating parameters and plate geometry allplay a role in the refining process, whether that be contributing to the fibre treatment mechanismor fibre transport in the refiner.The refining action has been shown to be dependent on plate geometry and operating condi-tions, as these variables are present in most of the refining models that are described in the previoussections. Models relating expended energy, plate geometry, operating parameters, and intensity andfrequency of treatment have been attempted as per models for SRE, SEL, SSL, and MEL. Thesemodels reveal the drive to increase the number of bars (or decrease groove size), while decreasingintensity to achieve increased fibrillation and decrease fibre cutting. Modifying plate geometry alsoimpacts the obtainable operating regions of LC refiners. Hydraulic capacity is highly dependent onproduction requirements and plate geometry. The grooves need to be sufficiently narrow to allowadequate refining action, but wide enough to allow adequate hydraulic capacity. The optimizationproblem of plate geometry is further impacted by the need to maintain a favorable mobile fibrenetwork in the grooves.Motion of the rotor over the stator during bar crossings induces rotational flow in the fillingor plate grooves; recall this is the same motion that contributes to the beating mechanism. Thisrotational motion in the grooves acts to circulate fibres ensuring consistent fibre network turnoverand adequate fibre suspension mobility. Without adequate turnover in the grooves, fibres maynever be exposed to bar crossings and thus the refining action. It is speculated that the formationof multiple primary vortices in the grooves may trap fibres in the lower portion of the grooves andwould thus be detrimental to uniform refining action.In addition, stator and rotor groove flow impacts gross refiner circulation and fibre residencetime, ultimately influencing homogeneity of the refining process. The refining process is inherentlycomplex in nature, and is clearly dependent on both the physical refining action and fibre transport.It is clear that fibre transport within the refiner plays a role in achieving homogenous refining, that10is, to transport unrefined fibres to the gap or near the bar leading edges in a uniform manner wherethe beating mechanism occurs.From this review it remains obvious that there are optimizations yet to be made based on im-proving our understanding of the affects of plate geometry and refiner operating parameters on theflow field in LC refiners. Fibre transport in the refiner remains a key focus for future research inorder to improve fibre mobility and thus uniform exposure to the refining action.1.6 Research Objectives and Thesis OrganizationFibre and pulp transport in LC refiners remains an area of research that offers many questions withregards to refiner operation. The previous review of pertinent literature has shown the many facetsof the refining process and has exposed areas that are lacking in understanding. The objective ofthis research was to visualize and quantify the flow field inside an LC refiner. This study furtherlooked at the effects of refiner operating parameters on the flow field with means to characterize theeffects of the flow field on refining potential. Where refining potential encompasses fibre networkturnover and homogeneity of the refining treatment in disc refiners.To facilitate this research project it was necessary to develop an experimental methodology tostudy the flow field inside an LC refiner. The first phase of this project included designing andfabricating a transparent refiner door that would allow visual inspection in the laboratory scaleLC refiner. Secondly, two-dimensional particle tracking velocimetry was implemented to allowquantitative analysis of fluid transport.This thesis presents a description of the experimental and analysis methods in Chapter 2, resultsand discussion in Chapter 3 and a conclusion with recommendations for future work in Chapter 4.11Chapter 2Research MethodsExperiments were carried out using the LC refining facility at the UBC Pulp and Paper Centre. Toperform flow visualization in the LC refiner, the first phase in the research program was to design,fabricate and install a transparent refiner door. Following this, experiments were carried out obtain-ing high speed video of tracer particles for two plate configurations. For all experiments a groovedacrylic stator plate was investigated with one of two rotor plates: (1) a smooth rotor with no barsand (2) a grooved rotor. A MATLAB Particle Tracking Velocimetry Program, hereafter referredto as the MATLAB PTV Program, was developed to perform image processing, particle trackingand flow field analysis. This chapter presents details on the approach taken in the experimental anddata analysis methods in the flow visualization study performed in this work.2.1 Experimental MethodsThe experimental methods used to obtain high speed video of tracer particles in the refiner flow fieldare detailed in this section. A description of the facility, equipment, test setup and the experimentaltest matrix are included.2.1.1 Refining FacilityThe UBC Pulp and Paper Centre has a state-of-the-art low consistency refining facility ideal forindustrial focused university research. The refining facility consists of a variable speed 16 inchsingle disc refiner equipped with a set of experimental refining plates capable of achieving lowintensity refining with BEL ranging from 0.99 to 12.9 km/rev (Figure 2.1).12Figure 2.1: UBC variable speed 16 inch LC refiner.2.1.2 Modified Refiner DoorTo accommodate the research objectives the refiner door required large-scale modifications to allowvisual access to the plate refining zones. This endeavor was a large task which included: a prelim-inary design phase, a detailed design phase, fabrication of components at external machine shops,and final assembly and testing in the refining facility. For preliminary design details includingdesign requirements, concept design and material considerations refer to Appendix A. This sec-tion provides a design overview, component descriptions, assembly and limitations of the ModifiedRefiner Door (MRD).Design OverviewThe MRD accommodates two modes of operation:1. Normal operation mode is capable of operating the refiner with no noticeable changes withrespect to the original equipment manufacturer (OEM) refiner door. It allows the use of13standard 16 inch refiner plates and is capable of performing identically to the standard refinerdoor.2. Monitoring operation mode allows plate-scale viewing of the refining zone from the under-side of the stator plate. The modifed refiner door has four viewports that allow observationof 75% of the refining zone around the refiner plates. The MRD installed in monitoringoperation mode may be seen in Figure 2.2.Figure 2.2: An external view of the MRD (left) and an internal view of the MRD (right)installed in monitoring operation mode.The MRD components were classified as either refiner door housing components or acryliccomponents. The housing components were made up of steel housing components that compliedwith internal pressure and loading requirements. The transparent acrylic components allowed vi-sual observation inside the refiner and were made up of an acrylic safety window and a machinedacrylic stator plate.Modified Refiner Door Housing ComponentsThis section presents the design of the refiner door steel housing that was fabricated and installedon the LC refiner during experimentation.14The refiner door steel housing was made of stainless steel 304. Modifications were designed bythe author of this work and fabrication was performed in Japan by the refiner original equipmentmanufacturer, Aikawa Iron Works Co., Ltd.. The housing backplate acts as the bearing surfacefor the mounting plate (normal operation) and the acrylic window (monitoring operation) as canbe seen in the mechanical design drawings UBC-001 and UBC-002 in Appendix B. The refinerdoor steel housing had to be modified in order to accommodate thicker internal acrylic componentsand the desired viewports. In the new design, the bearing surface had to be repositioned 24 mm(0.95 in) further from the refiner-door interface. A new refiner feed pipe had to be redesigned tointegrate with the refiners fixed pipe fitting attachments. See drawing UBC-003 in Appendix B forthe modified feed pipe design.The largest modification to the refiner door was the spoke pattern cutout from the housingbackplate. This four-spoke design offered maximum viewing area of the refining plates meanwhilemaintaining a flat, stable bearing surface for the acrylic components to lie on. The design of themounting surface shape and thickness was decided upon following a parametric static structuralfinite element analysis, see Appendix C.In order to allow a standard refiner plate to be mounted in the MRD, a stainless steel mountingplate may be installed. The mounting plate acts as a flat mounting surface for the refiner plate asshown in drawing UBC-002 in Appendix B. The thickness of the mounting plate was determinedfrom the monitoring operation design as that was the driving constraint on the dimensions. Themounting plate/modified refiner door interface was sealed with a 1.6 mm (1/16 in) general purposenon-asbestos sheet gasket designed for pressures up to 3.3 MPa (480 psi).Modified Refiner Door Acrylic ComponentsThe acrylic window and plate assembly shown in Figure 2.3 shows a two piece acrylic design. Theacrylic window exposed to the low-pressure side, or ambient air, is a flat 22 mm thick disk. Themachined acrylic stator plate is 22 mm thick from its base to the top surface of the stator bars.The acrylic window is placed inside the door and gasket sealed by 8 bolts fastening the acrylicwindow and plate to the MRD steel housing. The gasket thickness was chosen to be 1.6 mm (1/16in) as thicker gaskets require less sealing stress. For fastening acrylic it is desirable to keep thecompressive stresses due to bolting at a minimum to decrease stress concentrations in the boltedregions.The acrylic plate was machined to the dimensions shown in Appendix B, drawing UBC-005.The flatness tolerance achieved when mounted flat was 0.1 mm (0.004 in).15Figure 2.3: Monitoring operation: A double-pane acrylic window design enables observationvia four viewports revealing 75% of the refining zone.Figure 2.4: The acrylic window machined flat shown on a Haas milling machine.The acrylic stator plate is mounted alongside the acrylic window onto the modified refiner doorsteel housing. The stator pattern mimics an experimental plate with 4.8 mm by 4.8 mm grooves16and 3.2 mm bar widths. There are 48 repeating segments with a bar angle ? of 15 degrees and asegment angle ? of 7.5 degrees. Figure 2.5 shows a picture of the acrylic stator plate on the millingmachine. Dimensions of the plate may be seen in drawing UBC-009 in Appendix B. A Haas VF4Milling machine was programmed to run a 3 mm (1/8 in) carbide bit at 6800 RPM with a feed rateof 0.3048 m/min (12 in/min). Undulations of the machined surface were hand polished, and oncesubmerged under water, the surface finish was indistinguishable from the cast surfaces.Figure 2.5: The machined acrylic stator plate shown on a Haas milling machine.To space the acrylic window and acrylic plate, a 3.2 mm (1/8 in) neoprene gasket was used.Spacing between the two plates was required to allow a liquid layer between the two surfaces.The fluid used for the liquid layer was de-ionized water with a refractive index within 10% of therefractive index of acrylic. This ensured that when light passed the interface, refraction effectswere negligible. In addition, surface imperfections in acrylic may cause a degradation of opticaltransmission characeristics and water acts to fill the minute undulations caused by the machiningoperations performed on the acrylic surfaces to achieve the required flatness tolerances.The other significant motivation for implementing the double pane design was for safety. Adouble pane design allowed the majority of the machining operations to be carried out on the innerstator plate. This component will be subject to high residual stresses due to machining and may besubject to crazing or crack initiation at sharp edges or stress concentration zones. The 22 mm thickacrylic window is a fail-safe precaution in case of stator plate failure.17AssemblyExploded assembly section views of the MRD can be seen in Appendix B (UBC-012 and UBC-013). The acrylic window is centered on the steel bearing surface of the MRD by a centering lipon the inner diameter of the spoke hub. This centering method ensured the window was seatedproperly on the MRD. The entire assembly was bolted together with 8 stainless steel socket headcap shoulder screws. Buna-N washers were used on the acrylic surface to evenly distribute boltloading and avoid stress concentrations at a metal-acrylic surface under the bolt heads.Once assembled and installed on the refiner, the liquid layer was filled with de-ionized watervia two injection holes near the top of the acrylic window.LimitationsIn comparison to the standard refiner door the transparent MRD does have several limitations thatshould be noted. Acrylic as an engineering material is strong under compressive stress but is limitedwhen exposed to tensile stress and pulsating loads, especially when residual machining stresses arepresent. The machined acrylic bars remain susceptible to damage as they come in close proximityto the rotor plate. Contaminants in the working fluid may cause scratching or worst case destructionof one or more stator bars leading to a large scale failure. That being said, actual fibre refining is notachievable as compounded tolerances will not accommodate operational gaps nor friction loadingdue to refining; this may shear the stator bars leading to failure of the stator plate. Achievable gapsfor the transparent MRD are as low as 0.75 mm and internal pressure has been tested up to 30 psi.2.1.3 Experimental SetupThe experimental setup consisted of an LC refiner with the MRD installed, a data acquisition andcontrol system (DACS), and a data processor programmed in MATLAB (Figure 2.6).The LC refiner was outfitted with a modified refiner door (MRD) with transparent viewports.With four acrylic viewports it provided visual access to approximately 75% of the refining zone onthe stator and rotor plates. The MRD was designed to offer both plate and groove scale observation.Pressure and temperature measurements for the inlet and outlet were provided by pressuretransducers and thermocouples. A Rosemount magnetic flowmeter measured flow rate. All sensormeasurement data and control signals were transmitted by National Instruments DAQ I/O hardwareto the LabVIEW user interface, where signals were logged and displayed to the operator.The image acquisition hardware consisted of a Vision Research Phantom V611 high speedcamera equipped with a Nikon Nikkor 50mm F/1.4D lens. This high speed camera is capable of18LC Refiner&Modified Refiner DoorTest EnvironmentMeasurement DevicesData Acquisition and Control System (DACS)Control DevicesFlow MeterPressure TransducersThermocouplesImage Acquisition HardwareFlow Control ValveVariable Speed PumpDACS I/O HardwareLabVIEW User InterfacePhantom Camera Control Software MATLAB  PTV Program Data ProcessorFigure 2.6: Experimental setup diagram showing interactions between the test environment,DACS and data processor.recording images at resolutions of up to 1280x720 HD, due to its wide CMOS sensor. A maximumspeed of 6242 frames-per-second is configurable at full resolution and at reduced resolution, framerates up to 680,000 frames-per-second are achievable. The Nikkor lens was outfitted with a HOYAUV(0) filter able to filter UV transmission to 5% at 365nm and produce an average transmissionof 99.7% abve 450nm. An AC powered 200W UV lamp illuminated the measurement zones. Thecamera was controlled by Phantom Camera Control Software.Data processing was then performed by a custom MATLAB PTV Program detailed further inChapter Experimental Test MatrixThe experimental test plan was derived from a survey of industrial operating points for LC refiners.Experiments were performed with both smooth rotor and grooved rotor configurations (Figure 2.7)for a range of flow rates, refiner speeds and plate gaps as summarized in Table 2.1. Smoothrotor experiments were performed initially at flow rates Q = 300 to 500 litres per minute (LPM)and refiner speeds N = 400 to 1200 rotations per minute (RPM) to capture a range of flow ratesand a large range of refiner speeds. When the grooved rotor was installed on the refiner, higher19Flow Rate Q (LPM) Refiner Speed N (RPM) Plate Gap G (mm)Smooth Rotor 300, 400, 500 400, 800, 1000, 1200 7.5, 2.5, 1.5, 0.75Grooved Rotor 300, 500, 700 600, 800, 1000 7.5, 2.5, 1.5, 0.75Table 2.1: Experimental test plan for flow rate, refiner speed and plate gap.Figure 2.7: Plate geometries under investigation. (a) is a frontal view of the grooved rotor,(b) is the plate cross-section of the smooth rotor configuration and (c) is the plate cross-section of the grooved rotor configuration.internal pressures were experienced for a given refiner speed. As a result of the MRD structurallimitations defined in Section 2.1.2, the maximum refiner speed had to be reduced to 1000 RPM. Italso was found that maximum flow rates could be elevated from 500 to 700 LPM without exceedingpressure limitations. In the performed experiments plate gap G was limited to 0.75 mm by theacrylic component tolerances. A range of gaps were chosen to represent large gaps (G=7.5mm),an industry standard no-load power gap (G=2.5 mm) and near operational plate gaps (G=1.5 mm,0.75mm).20URLBFigure 2.8: A frontal view of the modified refiner door showing acrylic viewports and thestainless steel housing (left). The viewport layout is labelled: U (upper window), R(right window), L (lower window) and L (left window). A photo of the mounted cameraand the installed MRD (right).All presented data has been arbitrarily gathered from the right window of the LC refiner, aslabelled R in Figure 2.8. A single video of approximately 2 second duration captured 7000 framesfor each of the smooth rotor cases above. This data allowed for adequate qualitative analysis. Tomake quantitative results statistically relevant, data from 10 videos were compiled for groovedrotor cases at low gaps of G=0.75 mm.Industrial Operating RangesRecall that important machine operating parameters include flow rate Q, refiner speed N, and plategap G. As per recommended operating practices by Aikawa, J&L, and Voith, reasonable ranges forboth flow rate and refiner speed were extrapolated from industrial sized refiners to the laboratoryscale 16 inch refiner. See Appendix D for tabulated data on industrial operating parameters.In Figure 2.9, manufacturer recommended flow rates in ranges of low to high are plotted againstrefiner diameter. Flow rates of interest were chosen to be 300, 400, 500 and 700 LPM as depictedby the green experimental operating region plotted on Figure 2.9.Figure 2.10 shows industrial ranges of refiner speeds for a set of Aikawa and Voith LC refiners.The experimental operating region for refiner speed in this work is marked on the plot.2110 20 30 40 50 6005001000150020002500300035004000Industrial Flow Rate Operating RegionsRefiner Diameter, D (inches)Flow Rate, Q (LPM)Voith ? LowAikawa/J&L ? LowAikawa/J&L ? MedAikawa/J&L ? HighVoith ? High16" LC RefinerExp. Op. RegionFigure 2.9: Industrial operating region for refiner flow rates, the 16 inch refiner values wereextrapolated from the resulting industrial flow rate curves from different manufacturers.2.1.5 Particle Tracking VelocimetryThere exist many velocity measurement techniques with applications in measuring internal flows inducts and rotating machinery. Non-intrusive flow visualization techniques, including laser doppleranemometry (LDA), particle imaging velocimetry (PIV), and particle tracking velocimetry (PTV),among others, require that the flow be visually accessible. In PIV the flow field is visualized byintroducing small tracers into the flow and correlating a pattern of tracers in image pairs. At hightracer concentration this technique is very powerful and allows a dense vector field to be generated.PTV is a variant of PIV in which a low particle image density is used. In this method the particleimage density is low so that a simple interrogation algorithm may be used; that is, particles aredirectly identified and tracked in image pairs. This method has been chosen because it offers asimple method to visualize the flow in the form of both velocity vectors and particle trajectories forgeometries that are difficult to illuminate and for complex and high speed flows.Tracer ParticlesNon-intrusive techniques often require tracer particles in the fluid to obtain velocity measurementsin a continuous medium. Each technique has varying desirable tracer qualities, those being light2210 15 20 25 30 35 40 45 50 55 600200400600800100012001400Industrial Refiner Speed Operating RegionRefiner Diameter, D (inches)Refiner Speed, N (RPM)AikawaVoith16" RefinerExp. Op. RegionFigure 2.10: Industrial operating region for refiner speeds, and the experimental operatingregion investigated for the 16 inch refiner.scatter, light emittance, material, and size. LDA requires tracer particles that scatter light withminimal size requirements, often suspended contaminants in unfiltered water achieve adequatesignal response (Drain (1980)). PIV techniques often make use of intense laser light sources andcorrespondingly, require tracer particles that scatter light well. Particle size ranges depend on thefluid. Gas flows typically implement smaller tracer particles than liquid flows. PTV requires alow particle concentration in the measurement zone and a high signal to noise ratio is preferredthus larger particles may be used. To reduce the signal to noise ratio obtained from intensity basedimages, uniform seeding size is desirable (Melling (1997)).Fluorescent tracer particles absorb photons from incident electromagnetic radiation and emitlight at longer wavelengths. This experiment utilized ultra-violet (UV) fluorescent polyethylenemicrospheres that absorb UV light with a wavelength of 365 nm. Light emission was most intenseat 510 nm which resulted in green illumination when excited with UV. This intense light emissionin the green spectrum offered adequate contrast between the rotating steel rotor plate in the back-ground of the images. In addition, fluorescent tracers have isotropic emission characteristics thuslight scattering properties were of little importance (Tropea et al. (2007)). Polyethylene UV fluo-rescent tracer particles were used in this work in order to accommodate light intensity requirements23and flow tracking fidelity in high speed flows. Table 2.2 lists the tracer particle properties of theUV fluorescent microspheres.Property Description/Value UnitsMaterial Polyethylene ?Shape Spherical (> 90%) ?Density 0.99 to 1.01 g/ccSize 500-600 (> 90%) ?mDaylight Color Fluorescent yellow ?Illumination Color Fluroescent yellow-green ?Excitation Wavelength 365 (Peak) nmEmission Wavelength 510 (Peak) nmTable 2.2: Summary of UV fluorescent tracer particle properties provided by Cospheric In-novations in Microtechnology.Tracer Flow Tracking FidelityThe accuracy of the velocity measurements obtained by PTV is inherently dependent on the flowtracking fidelity of the tracer particles. PTV and any technique involving tracer particles rely onproper tracer particle selection based on shape, size and density, in addition to lighting require-ments.The particle Stokes number is a well known indicator of flow tracking fidelity and it has beenshown that St < 0.1 gives tracing errors below 1% for flows in the Stokes regime. The Stokesnumber is given as:St =?p? f(2.1)where ? f is the fluid characteristic time scale l/v and ?p is the particle relaxation time given as:?p =| ?p ?? f |d2p18?. (2.2)With a maximum | ?p ?? f | of 0.01 g/cc the Stokes number was calculated to be 0.035 whichsatisfies the desirable limit of St < 0.1. Although the Stokes number is not directly applicable tohigh Reynolds flows it remains an indicator for flow tracking fidelity (Raffel et al. (2007)).242.2 Analysis MethodsThe MATLAB PTV Program performs image processing, particle tracking, data analysis, and im-age calibration. This section details the approach and the underlying numerical methods used toprocess images, track particles, calculate instantaneous velocities, plot particle trajectories basedon Lagrangian and Eulerian techniques, and report uncertainty values. The objectives of the anal-ysis were to produce particle trajectories and velocity fields which accurately represent the flowfield inside the LC refiner.2.2.1 ApproachRaw images from the high speed camera are image processed initially by the Phantom CameraControl Software and further processed by the MATLAB PTV Program. Particles are located inthe image sequences and particle positions in the form of x,y coordinates with respect to time twere input to the tracking algorithm. Instantaneous velocity measurements were then calculatedfrom position and time data for each labelled trajectory. Results were processed for groove data,using masks to perform analyses only on the regions of interest. Two types of grooves were ofinterest as labelled in Figure 2.11, those being the primary groove with constant width and the flowgroove with diverging width from inlet to the plate periphery. The plate under study was a coarsesoftwood refiner plate with BEL = 0.99 km/rev having 48 repeating sectors each containing oneprimary and one flow groove. Reference frames for each groove are defined in the A,T coordinatesystem as shown.Assumptions and LimitationsGroove velocity data presented in the results of this work for a given groove, primary or flow, arecombined results for 3 adjacent plate sectors. Velocity measurement discrepancies in adjacent platesegments were found to exhibit less than 5% difference, thus results and statistical uncertaintieswere calculated for a combined sample set of 3 sectors.Inherently, the experimental setup described previously in Section 2.1.3 is only capable ofobtaining 2D velocity data (VA,VT ) in a 3D flow field. This limits the information that can beobtained from the helical flow patterns. The tangential velocity defined above is perpendicular tothe groove direction, but is not the tangential velocity of the helical flow.25Figure 2.11: Primary and flow groove regions of interest and reference frame definition.2.2.2 Image ProcessingImage processing is a form of digital signal processing; it can be defined as the process of au-tomatically applying computer algorithms to digital images. A digital image is a representationof a two-dimensional image in a distinct number of data points known commonly as pixels. Forgrayscale images each pixel contains information about the intensity level of that region in the im-age. Grayscale pixels are generally 8-bit values ranging from 0 (black) to 255 (white). The numberof gray levels distinguishable by the human eye is approximately 64 thus 255 is an adequate com-promise for image visual quality and compact representation and storage (Marques (2011)).An overview of the digital image processing method performed on the .cine files output fromthe Phantom Camera Control (PCC) Software is presented in Figure 2.12. The Phantom V611 highspeed camera records video at 4000 frames-per-second to its high speed internal RAM. The raw.cine file is then transferred to a PC running the PCC software and converted to 8-bit digital images.A raw grayscale image output from the PCC software and the pixel values of an 8x8 neighbourhoodare shown in Figure 2.13. A linear neighbourhood-oriented image processing algorithm is used onthe raw images in order to emphasize an images high frequency features. This spacial Laplacianhigh pass filter effectively highlights intensity transitions within the image in order to enhance26edges or lines and is implemented via a convolution mask.			 					  ! 	"#  $$    "#	Figure 2.12: Digital image processing method showing the operations performed by PhantomCamera Control Software and the PTV MATLAB Program.Figure 2.13: A grayscale raw image (left), and the pixel values in an 8x8 neighbourhood(right).The Laplacian of an image f (x,y) is defined as:?2(x,y) =?2(x,y)?x2+?2(x,y)?y2(2.3)where a simple approximation of the second derivatives is given by Marques (2011) as:27?2(x,y)?x2= f (x+1,y)+ f (x?1,y)?2 f (x,y) (2.4)and?2(x,y)?y2= f (x,y+1)+ f (x,y?1)?2 f (x,y) (2.5)This is an example where the expressions may be combined and implemented by the convolu-tion mask:???0 ?1 0?1 4 ?10 ?1 0???(2.6)The high pass Laplacian filter used in this work is a 5x5 highpass filter of the same type; forbrevity, the mask is provided below omitting the derivative approximations:?????????1 ?1 ?1 ?1 ?1?1 ?1 ?1 ?1 ?1?1 ?1 24 ?1 ?1?1 ?1 ?1 ?1 ?1?1 ?1 ?1 ?1 ?1????????(2.7)An example image filtered with this mask is shown in Figure 2.14, note the enhanced edges ofthe tracer particle in comparison to the grayscale pixel values seen in Figure 2.13.Gray-level thresholding is then performed on the neighbourhood filtered image in order toeliminate the low intensity noise floor. The AC light source intensity pulsates at a rate of 120 Hz.To accommodate these fluctuations a dynamic threshold algorithm is used. The mean pixel valuePi of a given image is interpolated between the mean pixel values(Plow,Phigh)of the darkest andlightest frames in order to obtain Ti in the range(Tlow,Thigh). The gray-level thresholding is thenperformed for the image f (x,y) by:g(x,y) ={1 if f (x,y) > Ti0 otherwise(2.8)Figure 2.15 shows the binarization of the digital image, the tracer particle is now clearly iden-tified by the 1?s in the 8x8 neighbourhood.28Figure 2.14: A Laplacian high pass filtered image (left) and the pixel values in an 8x8 neigh-bourhood (right).Figure 2.15: Thresholded binarized image (left) and the pixel values in an 8x8 neighbourhood(right).Once the image has been binarized a pixel area bandpass filter was applied to each particlesignature. This used the bwareaopen function in MATLAB. A lower and an upper particle size,Alow and Ahigh, respectively, are defined based on the image acquisition settings and the particle29size in pixels. bwareaopen then filters out particles less than Alow and greater than Ahigh.Finally, to eliminate noise present in the images following the above filtering a circularity filterwas applied to each particle signature. The circularity metric was defined as:Mc =4? (Area)(Perimeter)2(2.9)where Mc = 1 for a perfect circle and Mc = 0 for a line. Particle signatures were thresholded basedon a circularity threshold value Tc which was manually determined for each camera calibration.2.2.3 Particle Tracking AlgorithmFollowing image processing the remaining particle signatures must be tracked in space and time.The position of each tracer particle is determined via the regionprops function in MATLAB in-voking the centroid option. Particle tracking from frame to frame is then performed by particletracking code developed originally in Interactive Data Language (IDL) by John C. Crocker, DavidG, Grier and Eric R. Weeks. This algorithm has seen applications in astronomy, medical imagingand colloidal particle tracking. A MATLAB variant written by Daniel Blair and Eric Dufresne hasbeen used in the PTV MATLAB Program.The tracking algorithm locates particles in image sequences by matching locations in eachimage with corresponding locations in subsequent images to produce a trajectory. Tracking a singleparticle from frame to frame is relatively simple as there is only one matching particle in thefollowing time step, thus the match is trivial. Introducing multiple particles introduces complexityas each particle can be matched with only one particle in each successive frame. The trackingalgorithm minimizes the total squared displacement for all possibilities of matching n particles attime ti with m possible new positions at time ti+1.The output of the tracking algorithm is particle position (xi,yi) with respect to time (ti) for eachaccepted trajectory. The raw trajectory data is then output to the data analysis algorithms detailedin the next section.2.2.4 Instantaneous Velocity MeasurementsIn order to extract quantitative results for characterizing the flow field velocity, vectors may begenerated from the tracked particle positions. Instantaneous velocity components may then beextracted for the nth trajectory in the timestep from i to i+1 by:30un,i =?xn,i?tn,i=xn,i+1 ? xn,itn,i+1 ? tn,i(2.10)andvn,i =?yn,i?tn,i=yn,i+1 ? yn,itn,i+1 ? tn,i(2.11)where un,i is the velocity component in the x-direction and vn,i is the component in the y-direction.2.2.5 Lagrangian AnalysisThe particle tracking velocimetry method is inherently a Lagrangian technique, that is, tracking agiven fluid element or particle in space and time along a path. Lagrangian trajectories are presentedin the results section to depict flow behaviour between the refiner plates. The trajectories representparticle, and thus fluid pathlines.Ouellette et al. (2008) showed that Lagrangian methods of particle tracking are more sensitiveto inertial effects. Even small effects of particle inertia may manifest itself in the form of pathlinemisalignment. Even neutrally bouyant particles are subject to pathline error as flow stresses areaveraged over the particle surface, especially in regions of high shear. Particle trajectories may besubject to propogated error when depicting fluid pathlines. For this reason, caution must be takenwhen relating trajectories to fluid behaviours. For the purpose of this research, the propogatederror remains adequately small as the Stokes number calculated in Section 2.1.5 was found to besufficiently low. For this reason, limited quantitative results are presented for Lagrangian particletrajectories.2.2.6 Eulerian AnalysisThe Eulerian specification of the flow field involves focusing on a particular location in space andobserving fluid flow as time passes. Eulerian data analysis methods were used extensively in thiswork to quantify velocity measurements at radial positions along the length of the groove. Ouelletteet al. (2008) showed that in the Eulerian specification single point statistics of large particles arenearly indistinguishable from their small tracer counterparts. In this manner, Eulerian velocitymeasurements are not subject to the same degree of error as the Lagrangian specification.In order to compute velocity profiles along the length of the stator grooves, cross-sections weredefined that were aligned perpendicular to the groove axes. All velocity measurements presentedfor a given groove position are computed as vectors that intersect the cross-section. This in mind,31it is important to note that reported averages of the axial velocity represent bulk fluid velocity inthe groove. The cross-sections are defined as lines in the 2D image reference frame (Figure 2.16).A method of intersecting lines was programmed in MATLAB in order to determine which vectorswere of interest for each cross-section.Figure 2.16: Cross-sections of interest equally spaced along the length of the grooves wereused to calculate Eulerian velocity measurements.2.2.7 Coordinate TransformationsIn order to present results from both the Eulerian and Lagrangian analysis in a relevant manner,it was required to transform the positions and instantaneous velocities from Section 2.2.4 to thegroove reference frames shown in Figure 2.17.Firstly, the coordinate transformation from image reference frame to groove reference frameincluded translation and rotation. The coordinates (x,y) were translated by:(A,?T ) = (x? tx,y? ty) . (2.12)Next, a coordinate transformation matrix was applied in the form of Equation 2.13 to rotate thecoordinate system by ? .[A?T]=[cos? sin??sin? cos?][xy](2.13)32Figure 2.17: Image frame of reference (x,y) and the groove reference frame in terms of theaxial A and tangential T coordinates.To better present the results of this work tangential components of position and velocity, T andVT respectively, were reflected in the axial axis so that positive values would be in the direction ofthe rotor translating over the stator.2.2.8 Calibration MethodsImage calibration was performed by an interactive program coded in MATLAB. Calibration had tobe performed for every camera configuration change. For this reason, whenever data was compiledfrom multiple videos, camera configuration was unchanged to limit calibration error.The pixel to meter conversion was computed as the mean of two values. Each value wascalculated by interactively sizing a circle on screen to match a plate feature. The feature size wasthen provided in the image reference frame in units of pixels. This value was then equated to theknown physical dimension of the feature. The two features used were the outer plate diameter andthe rotor bolt hole. The maximum percent difference between these two feature calibrations for alltests was found to be 3%.332.2.9 Uncertainty AnalysisReported values in this work are accompanied by uncertainty intervals that represent the 95%confidence limits on a given measurement or calculated value. The true value x? is given as:x? = x??ux? (P%) (2.14)where x? is the sample set mean and ux? is the combined uncertainty for all known errors with P%confidence. Contributing error in this experiment may be classified as either systematic bx?, orrandom error rx?. The combined uncertainty is then given as:ux? =[b2x? + r2x?]1/2(2.15)Systematic error bx? incorporates error that may cause either a high or low offset in the truevalue estimate x?. In this experiment, sources of systematic error included calibration error (Sec-tion 2.2.8) and lens abberations such as tilt and image distortion. For the purposes of uncertaintyreporting, lens abberations were deemed negligible and were not included in provided uncertaintiesfor velocity.Random error rx? for reported values was calculated using statistics for finite data sets. Finitedata sets were sampled and t-distributions were implemented to perform the statistical uncertaintyanalysis. Random error rx? was calculated with:rx? = t? ,Psx? (P%) (2.16)where t? ,P is the student t-value for ? degrees of freedom at P% confidence, and sx? is the standarddeviation of the means.When reported values are the result of a calculation from independent variables, propogationof error must be considered. The propagation of uncertainty to the result may be defined for afunction y = f (x1,x2, ? ? ? ,xn), where (x1,x2, ? ? ? ,xn) are independent variables, as:uy? =?{n?i=1(?y?xiux?i)2}12(2.17)34Chapter 3Results and DiscussionThe results of this work are based on high speed video footage capturing tracer particle movementin the fluid flow between LC refiner plates. Particles were tracked to develop Lagrangian trajecto-ries and quantify the velocity field in the stator grooves. The results of the collected and analyzeddata are presened in this Chapter. Firstly, the obtained particle trajectories are introduced, followedby qualitative characterizations of observed particle behaviour. Next, Eulerian velocity measure-ments and bulk estimates of groove axial velocity are presented and related to refiner operatingparameters. Lastly, estimates of groove turnover rates are provided to quantify fluid rotation in thestator grooves. Turnover rate is a term used in this work which defines a measure of fluid turnover,accounting for the number of turnovers that a fibre is exposed to in a single groove pass.3.1 Qualitative Observations of Particle BehaviourThis section provides qualitative observations of the behaviour of particles and characterizes themphenomenologically. Qualitative observations are provided for the effects of refiner speed, flowrate, and plate gap. Furthermore, Lagrangian trajectories reveal two other additional characteristicsof the flow.3.1.1 Phenomenological Characterizations and ObservationsCharacterizations were made based on common observations of pathline shape in stator grooves,the plate gap, and rotor grooves. First, typical trajectories are introduced followed by the charac-teristic behaviours.Example trajectories for particles travelling through the refining zone are shown in Figure 3.1.35In this example, the rotor is travelling to the left of the image, the bars of which may be seen in theraw image on the left. The red pathline begins at the blue circle near the plate periphery and travelsinwards in the stator groove. The particle is seen to exit the stator primary groove and exhibittangential motion in the gap or rotor groove until it re-enters the stator flow groove. It exhibits thisdeparture and entry once more until it is swept out of the view of the camera by the rotor. Thegreen particle is entrained by the bolt hole as it enters the refiner plate from the inlet. Figure 3.1depicts common behaviours which are further defined by the characterizations introduced next.Figure 3.1: Lagrangian trajectories of the particles circled in blue following their pathlinesthrough the refining zone. A rotor bolt hole is circled in black. The raw image (left)and the spatial Laplacian high pass filtered image (right) are shown with the stator barconfiguration superimposed.Six characterizations were developed based on qualitative observations. They have been cate-36gorized by the behaviours described below and are illustrated in Figures 3.2 and 3.3.NUSmooth Rotor(b)(a)(1)(2)(3)(4)(5)BLFigure 3.2: Illustration of smooth rotor characterizations for helical stator groove flow (1),groove departure (2), gap flow (3), groove entry (4) and bolt hole entrainment (5). Theobserved camera view from the underside of the stator plate is shown in (a) and thepredicted particle behaviours in the 2D groove cross-section are depicted in (b).1. Helical stator groove flow is the rotational flow pattern observed in stator grooves. Themotion of the rotor over the stator induces rotation of the fluid as depicted by the vortex ofblue particles labelled (1) in Figures 3.2b and 3.3b. The presence of a pressure differentialalong the length of the stator grooves induces axial flow; this results in the helical motionrepresented by (1) in Figures 3.2a and 3.3a.Figure 3.2a shows the camera view from the underside of the stator plate where the helix isviewed as a wave in the direction of the groove. The high speed videos and computed particletrajectories generally revealed outward stator flow for the smooth rotor. Particle trajectoriesobserved were sinusoidal-like for the smooth rotor.37NGrooved RotorU(1)(2)(3)(4)(5)(6)(a)(b)BLFigure 3.3: Illustration of grooved rotor observations for helical stator groove flow (1), groovedeparture (2), gap flow (3), groove entry (4), bolt hole entrainment (5) and helical rotorgroove flow (6). The observed camera view from the underside of the stator plate isshown in (a) and the predicted particle behaviours in the 2D groove cross-section aredepicted in (b). Note that helical rotor groove flow is portrayed in the rotor referenceframe.The grooved rotor experiments revealed stator backflow at all operating points as shown inFigure 3.3a. Unlike the simple sinusoidal flow observed in the smooth rotor, the grooved38rotor experiments showed more chaotic particle motion. Particles transferred among stream-lines with various radii with respect to the groove. This was observed as a shift in amplitudeof the observed wave. As a result, trajectories were observed to be sinusoidal-like with vary-ing amplitude along the groove.2. Groove departure is the ejection of particles out of the grooves into the gap. The particlesgenerally exhibited helical groove flow (1) with a large amplitude proceeded by groove de-parture (2) and gap flow (3). Particles with large amplitude travel nearest the bar surfacesthus have a higher probability to exit the groove flow and depart to the gap.3. Gap flow is represented by red particles (3) travelling over top of the stator and rotor grooves.These particles are directed radially outward with a large tangential velocity component.The tangential velocity approaches the rotor speed which is dependent on radial position andrefiner speed.4. Groove entry occurs when particles are travelling in the gap (3) and they enter the groovehelical flow pattern in the stator or the rotor, (4) or (6), respectively.5. Bolt hole entrainment occurs when particles are entrained and captured by the bolt hole inthe rotor plate; these particles originate from the eye of the refiner or the stator grooves whenbackflow exists. Particles were observed to follow the bolt hole for a portion of the platerevolution, often exceeding the width of the given viewport.6. Helical rotor groove flow was indistinguishable from gap flow in grooved rotor experimentsfor the obtained high speed videos because observations were made in a stationary referenceframe. It has been shown previously in Section 1.4.2 that vortical motion exists in rotorgrooves with high groove axial velocity as a result of centrifugal body forces. Figure 3.3bdepicts helical rotor groove flow (6) in the rotor reference frame.3.1.2 Qualitative ObservationsSeveral qualitative observations are described, including the effects of refiner speed, flow rate andplate gap on the flow field for smooth rotor experiments. In addition, the existence of a frequencyshift along the groove and the presence of corner eddies are identified.The smooth rotor experiments provided a simple test case where steady flow could be expected,as opposed to the unsteady flow regime induced by rotor bars passing over the stator. This offered a39simplified geometry to study the effects of operating parameters on the groove flow. The trajectorieswere sinusoidal-like in shape and were non-chaotic in comparison to grooved rotor results.Effect of Refiner SpeedA comparison of particle trajectories for a range of refiner speeds Q= 400, 800 and 1200 RPM withconstant flow rate and gap is shown in Figure 3.4. The waveform trajectories seen in the statorgrooves have a restricted amplitude as restrained by the stator bars. Meanwhile, the frequency ofrotation increases with RPM, as can be seen by the increased number of turnovers along the groove.It can be concluded that the refiner speed has a direct impact on the angular frequency, and thus,the turnover rate of particles in a single groove passing. The turnover rate, that is, the number ofFigure 3.4: Smooth rotor particle trajectories at constant flow rate (Q = 400LPM), constantgap (G = 0.75mm) and refiner speeds of 400 RPM (a), 800 RPM (b) and 1200 RPM (c).The refiner inlet is located on the left of each image and the rotor motion is upwards.turnovers the fluid is subject to in one groove passing, increased considerably with increased refinerspeed at constant flow rate. This result was expected for the smooth rotor and is revisited for thegrooved rotor in Section 3.3.Effect of Flow RateQualitative observations of flow rate variation from 300 to 700 LPM offered no clear evidence offlow pattern changes. The effect of flow rate on stator axial flow becomes apparent when quantita-tive results are studied in Section of Plate GapFigure 3.5 shows the particle trajectories visualized for G= 7.5, 2.5 and 0.75 mm. Figure 3.5ashows particles exhibiting primarily gap flow for large gaps and minimal rotation when in thegrooves. The proportion of particles in the stator grooves can be seen to increase as gap decreases,as expected. Figure 3.5c reveals trajectories in the stator grooves showing primarily helical grooveflow behaviour for G=0.75 mm.Figure 3.5: Smooth rotor particle trajectories at constant flow rate (Q = 400LPM), constantrefiner speed (N = 800) and plate gaps of 7.5 mm (a), 2.5 mm (b) and 0.75 mm (c). Therefiner inlet is located on the left of each image and the rotor motion is upwards.As the plate gap decreased, flow was constrained to the stator grooves and increased rotationalvelocities were observed in the grooves by the presence of helical flow. As a result of pathlinesbeing constrained to the grooves, the average tangential velocity was decreased as fewer particleswere present in the gap.Frequency Shift Along the GrooveIn the following grooved rotor results, Lagrangian particle trajectories are presented showing path-lines of tracked particles as they flow from the periphery inwards along the groove.Consistent with expectations, along the groove the frequency of rotation was observed tochange from periphery to inlet (Figure 3.6). At the periphery, the rotor bars cross the stator bars ata higher velocity imparting increased momentum transfer. By observing the temporal Lagrangiantrajectory shown in the lower plot in Figure 3.6, the frequency of the signal is seen to decrease withtime.410 2 4 6 8 10 12 14?202Primary Groove Tangential Position versus TimeTime (ms)Tangential Position, T (mm)15 20 25 30 35 40 45 50 55?202Primary Groove Tangential Position versus Axial PositionAxial Position, A (mm)Tangential Position, T (mm)Trajectory StartFigure 3.6: Spatial (above) and temporal (below) Lagrangian trajectories showing the fre-quency shift along a primary groove at G=0.75 mm, N = 800 RPM, and Q = 500 LPM.The frequency shift suggests that the fluid has an increased turnover rate at the outer limitsof the plate. In refining, when stator groove flow has increased angular velocity it increases thenumber of turnovers a fibre experiences along the length of the groove. This exposes more fibresto the refining action. By exposing a larger proportion of fibres to the refining action the processbecomes more uniform, which is a desirable outcome of LC refining.Spatial and temporal Lagrangian trajectories for the flow groove (Figure 3.7) show larger am-plitudes as a result of the increased width of the flow groove. As expected, the signal frequencydoes not change considerable amounts along the groove because the convergence of the groovewidth counteracts the rotor tangential speed. In addition, the signal amplitude converges concur-rently with the groove width.420 2 4 6 8 10 12 14?505Flow Groove Tangential Position versus TimeTime (ms)Tangential Position, T (mm)0 10 20 30 40 50 60 70?505Flow Groove Tangential Position versus Axial PositionAxial Position, A (mm)Tangential Position, T (mm)Trajectory StartFigure 3.7: Spatial (above) and temporal (below) Lagrangian trajectories showing the fre-quency shift along a flow groove at G=0.75 mm, N = 800 RPM, and Q = 500 LPM.Presence of Corner EddiesCorner eddies were observed at most operating points being observed more often at lower refinerspeeds. For example, a Lagrangian trajectory of a particle trapped in a corner eddy is shownin Figure 3.8. At positions A = 30 to 40 mm notice the trajectory oscillating near the grooveextremity for several turnovers. The presence of corner eddies of the size capable of capturing 500?m particles is indicative of large eddies with diameter on the order of 1 mm.In the review of lid-driven cavity flows it was shown that corner eddies exist in the bottomcorners of a cavity as a result of the conservation of angular momentum. Recall as the Reynoldsnumber increases primary vortices tend to extend throughout the entire cavity and the corner ed-dies become infinitesimally small at Re = ?. Also, decreasing Reynolds number has the effectto increase the size and intensity ratio of the corner eddy to that of the primary vortex. If corner430 2 4 6 8 10 12 14?202Primary Groove Tangential Position versus TimeTime (ms)Tangential Position, T (mm)10 20 30 40 50 60 70?202Primary Groove Tangential Position versus Axial PositionAxial Position, A (mm)Tangential Position, T (mm)Trajectory StartFigure 3.8: Spatial (above) and temporal (below) Lagrangian trajectories showing the pres-ence of corner eddies at G=0.75 mm, N = 800 RPM, and Q = 500 LPM.eddies become large enough they will join to form a second primary vortex resulting in multiplevortices aligned vertically in the groove. The formation of large corner eddies or a second pri-mary vortex is detrimental to the refining process as fibres may get trapped in lower vortices andthus remain removed from where the refining action occurs; at the bar leading edges and surfaces.Work by Nasab et al. (2013b) showed that above groove aspect ratio D/W ? 3 multiple primaryvortices will exist in the groove during LC refining. This simulation was 2D so it did not considerthe effects of the groove length in a 3D geometry. Wittberg et al. (2012) showed particle pathlinesthat revealed multiple vortices in 3D simulations of refiner plate grooves; this was also confirmedexperimentally. In this research the existence of a single primary vortex was observed for all cases.This is due to the groove aspect ratio of the plate under study being W/D = 1.443.2 Effect of Operating Parameters on Stator Groove VelocitiesThis section presents a study of groove axial velocities VA, for both the primary and flow grooveas defined in Section 2.2.1 on page 25. Velocity measurements were made possible by the particletracking velocimetry method detailed in the Research Methods chapter. Groove velocities in thestator grooves were calculated for a set of 9 operating points at G = 0.75 mm. Multiple videosfor each operating point were required to obtain a large number of particles travelling through thegrooves under observation. The operating points were parametrically varied with Q = 300, 500 and700 LPM and N = 600, 800 and 1000 RPM for the grooved rotor case.3.2.1 Velocity Distributions at Cross-sections of InterestA Eulerian measurement technique was implemented as described in Section 2.2.6 to determineaxial velocities along the length of the groove. For each position along the groove, a similar dis-tribution to that seen in Figure 3.9 can be shown for each radial position at all operating points.Figure 3.9 shows a histogram for all measurements of primary groove axial and tangential veloci-?15 ?10 ?5 0 5 10 15 20050100Primary Groove Axial Velocity HistogramAxial Velocity, VA (m/s)Frequency?15 ?10 ?5 0 5 10 15 20050100Primary Groove Tangential Velocity HistogramTangential Velocity, VT (m/s)FrequencyFigure 3.9: Histogram of groove velocities at mid-length of a primary groove for the operat-ing point: G=0.75 mm, N=800 RPM and Q=300 LPM.ties mid-length along the groove. The velocity distributions shown are due to data collection overa planar cross-section perpendicular to the groove axis with length equal to groove width. Data45is also time-averaged for the unsteady flow caused by passing bars and the turbulent flow regime.Axial velocities for both the primary and flow grooves can be seen to exhibit a clear bimodal dis-tribution. This same distinction cannot be seen in tangential velocities as the tangential velocitydistribution is skewed to the positive side; recall positive tangential velocities represent flow in thedirection of rotor rotation. The bimodal distribution is indicative of two distinct types of flow forvelocity measurements at a given groove cross-section: stator flow and gap/rotor flow. This be-haviour is an artifact of 2D velocity data being obtained in a 3D flow field. Furthermore, in Figure3.10, a scatter plot of all velocity measurements passing through the cross-section is shown withstator and gap/rotor flow distinguished by:(VA,VT ) ={Stator flow if VA < 0Gap & rotor flow otherwise(3.1)In the subsequent study, velocity measurements are presented as stator flow and all measure-ments have been thresholded as in Equation 3.1. A histogram and scatter plot showing the sameflow distinction for the flow groove may be seen in Appendix E Figures E.1 and E. Axial Velocity Profile Along Stator GroovesAxial velocity variations along the stator grooves were calculated as average values of the Eule-rian stator groove velocity measurements introduced in the previous section. Recall that velocitymeasurements reported here represent bulk fluid velocity in the stator grooves. The velocity profilealong the length of the primary groove for G = 0.75 mm and Q = 300 LPM is shown in Figure 3.11for three different refiner speeds. In each of the 9 operating points studied, the axial velocity wasobserved to decrease in magnitude as it travelled from the periphery of the plate A = 0.07 mm tothe inlet side of the groove A = 0.0 mm. Continuity of an incompressible fluid dictates that netfluid transport must be occurring out of the groove. Refer to Appendix F for the application of theequations of continuity and momentum for a simplified flow representing this case.It is clear that with increased refiner speed the magnitude of axial velocity increases along theentire groove length. Similar trends were observed for all flow rates, with axial velocity magnitudesshowing a stronger dependency on refiner speed than flow rate. For plots of the axial velocity profilefor Q = 500 LPM and Q = 700 LPM refer to Appendix E Figures E.3 and E.4. The existence ofaxial velocity fluctuations along the length of the groove is discussed later.Uncertainty in the axial velocity values is presented with errorbars representing 95% confidencelevels. The uncertainty of the axial velocities were in the range of 5% for most measurement points,46?15 ?10 ?5 0 5 10 15 20?10?5051015Groove Tangential Velocity vs Groove Axial VelocityAxial Velocity, VA (m/s)Tangential Velocity, VT (m/s)Stator FlowGap\Rotor FlowFigure 3.10: Scatter plot of tangential velocity VT versus axial velocity VA at mid-length for aprimary groove at the operating point: G=0.75 mm, N=800 RPM and Q=300 40% for the extreme cases shown with large errorbars. At the outer limits of the groove, increaseduncertainty occurred because illumination quality of the measurement zone was decreased, thusfewer particles were detected.The same dependency of bulk axial velocity on refiner speed exists for the stator flow groovesas Figure 3.12 depicts; increased refiner speed again increased the negative axial velocity. Contraryto the primary groove, the axial velocity trend is observed to be relatively constant along the lengthof the flow groove. This is again indicative of fluid departing the groove as the width of the grooveis decreasing from periphery to inlet.See Appendix E Figures E.5 and E.6 for flow groove axial velocity profiles for Q = 500 LPMand Q = 700 LPM. Similar trends are revealed.4780 10 20 30 40 50 60 70?10?9?8?7?6?5?4?3?2?10Primary Groove Axial Velocity vs Axial Position with Constant Flow RateGroove Axial Position, A (mm)Groove Axial Velocity, VA (m/s)600 RPM800 RPM1000 RPMFigure 3.11: Bulk axial velocity versus groove position for primary groove at the operatingpoint: G=0.75 mm and Q=300 LPM.3.2.3 Bulk Axial Velocity Groove Estimates versus Refiner Operating ParametersThe bulk axial velocity values presented are mean values of all cross-sections of interest that liein the range A = 0.01 to 0.06 mm. This is to eliminate data on the extremities of the groove todecrease entry effects on the bulk velocity estimate. Figure 3.13 shows the dependency of bulkaxial velocity on flow rate and refiner speed. With an increase in flow rate, stator axial velocitydecreased, but minimally. From the lines of constant refiner speed it is apparent that there is astrong positive correlation for increased stator flow with an increase in refiner speed. Equivalenttrends were observed in the flow grooves; refer to Appendix E Figure E.7 for flow groove data.When bulk axial velocity is plotted against refiner differential pressure ?P it is clear that axialvelocity is a linear function of ?P (Figure 3.14). The linear trend line offers an R2 value of 0.99for primary groove flow and R2 value of 0.96 for the flow groove data as seen in Figure E.8. Itis also apparent that even when the refiner is experiencing a negative pressure differential acrossthe refiner there remains inward stator flow. This implies there remains a high pressure zone at the480 10 20 30 40 50 60 70?8?7?6?5?4?3?2?10Flow Groove Axial Velocity vs Axial Position with Constant Flow RateGroove Axial Position, A (mm)Groove Axial Velocity, VA (m/s)600 RPM800 RPM1000 RPMFigure 3.12: Bulk axial velocity versus groove position for flow groove at the operating point:G=0.75 mm and Q=300 LPM.housing periphery, at least near the arbitrary grooves being observed in the right window of theMRD as in Figure 2.8 on page Discussion of Stator Groove Velocity ResultsStator groove axial velocity results were presented in Sections 3.2.1 to 3.2.3. Negative axial veloc-ities were measured for all operating points. In LC refining terminology the negative flow tendencyis referred to as stator backflow and backflow velocity will represent the magnitude of negativeaxial velocity.Lumiainen (1994) was the first to quantify stator backflow velocities in a conical refiner. Back-flow velocity profiles along a primary groove (with constant cross-sectional area) for water flowwere similar to those presented in Figure 3.11 with the backflow velocity trend decreasing fromperiphery to inlet. Consequently, these results indicate net fluid transport out of the groove fromperiphery to inlet as supported by the equations of continuity and momentum shown in Appendix49250 300 350 400 450 500 550 600 650 700 750?7?6?5?4?3?2?10Groove Axial Velocity vs Refiner Flow RateFlow Rate, Q (LPM)Groove Axial Velocity, VA (m/s)600 RPM800 RPM1000 RPMFigure 3.13: Primary groove axial velocity versus flow rate for lines of constant refiner speed.F.Fluid transport out of the stator groove is support of the tertiary wiping flow speculated byFox et al. (1979) as a result of a pressure gradient from stator to rotor grooves. Nasab et al.(2013a) also noted groove transference in numerical studies of the refiner 2D cross-section due tothe formation of a pressure gradient during bar crossings. Kondora and Asendrych (2013) presented3D numerical simulation results that also depicted mass transport out of the stator grooves into thegap. It was found that for increased groove divergence angles, similar to that found in the flowgroove studied in this work, increased mass transport occurred as a result of a favorable pressuregradient. The presence of mass transport out of the stator grooves is beneficial for two reasons,those being:1. Increased gross refiner circulation results in increased homogeneity of the refining process.2. Pulp transport out of the grooves ejects fibres into the refining zone where the refining ac-tion takes place. This work has confirmed the presence of net mass transport out of the50?30 ?20 ?10 0 10 20 30 40 50?8?7?6?5?4?3?2?10Primary Groove Axial Velocity versus Refiner Differential PressureRefiner Differential Pressure, ?P (kPa)Groove Axial Velocity, VA (m/s)Figure 3.14: Primary groove axial velocity versus refiner differential pressure. Linear trend-line R2 = 0.99.stator grooves but it remains difficult from these experimental results to prove the underlyingmechanism.The presence of particles departing the stator grooves may also be a mechanism of the escaperadius re observed in the refining zone by Senger et al. (1998) for a C=20% water fibre suspensionin a laboratory scale single disc refiner. At the escape radius, stapled flocs were observed to detachand get swept into the gap and rotor obtaining a large tangential and radial component of velocity.This is a similar behaviour to that observed in particle trajectories when exhibiting groove departurebehaviour. Qualitatively, it was observed that groove trajectories had a tendency to depart groovesover a limited range of radial positions. Unfortunately, quantification of this behaviour was difficultand no further conclusions may be made.The cause of backflow velocity fluctuations along the groove at all operating points is unclear.For all cases, the 3 grooves from adjacent sectors where data was combined, the peaks and troughsoccurred at approximately the same groove positions. This indicates that the profiles are repre-sentative of a flow phenomena rather than experimental error. These fluctuations are indicativeof transport into and out of the stator grooves resulting in velocity increase when fluid enters and51decrease when fluid departs. This is supported by the the groove entry and ejection particle be-haviours observed in this work and depicted by the characterizations (4) and (2) in Figure 3.3.The lack of backflow reversal observed in the grooves in these experiments indicates that underthe operating ranges studied, the local pressure gradient opposed outward stator flow. Kondoraand Asendrych (2013) showed a linear relationship of pressure versus groove position showingdeviations from linear near the inlet groove opening. Minimal pressure fluctuations were presentedalong the grooves by Kondora and Asendrych which is inconsistent with the expected pressureprofile from these observations. It is expected that with fluctuating velocity along the length of thegroove the pressure profile would also fluctuate. In addition, the time-dependency of bar crossingevents will induce pressure fluctuations as shown by the numerical work by Nasab et al. (2013a) inan unsteady time-dependent simulation. Note that the bulk estimates reported here do not capturetime-dependency.The dependency of backflow velocity on refiner operating parameters was clearly depicted inFigure 3.13 with equivalent trends shown for the flow groove in Appendix E. Flow rate affectedstator backflow minimally in comparison to refiner speed. As refiner speed increased from 600 to1000 RPM backflow velocities ranged from 3 m/s to 6.5 m/s and 3 m/s to 6 m/s for the primaryand flow grooves, respectively. This trend is consistent with observations that were made by Halme(1962), Lumiainen (1994) and Wittberg et al. (2012).Even though the primary control variables N and Q were found to influence stator backflow,it was shown that backflow was linearly dependent on refiner ?P. This suggests that the refinerdifferential pressure ?P is directly proportional to the pressure differential that exists across thestator grooves from periphery to inlet.In turn, plate geometry also has a large impact on refiner ?P. This is indicative that increasedrefiner circulation due to backflow, and thus increased residence time, is associated with increasedpressure differentials across the refiner. The optimization problem introduced in Chapter 1 withplate geometry and refiner operating parameters is further exposed. Plate geometry parameters suchas groove width W, groove depth D, bar angle, etc. that are included in the refining theories (SEL,SSL, MEL, etc) are not only influencing the refining action but also refiner circulation patterns andresidence times. Ultimately, pulp transport in the refiner has been found to be a complex componentof the optimization of hydraulic capacity and homogeneous refining action.523.3 Effect of Refiner Operating Parameters on Groove RotationalFlowThis section presents a study of groove rotational flow for the primary groove. Groove rotationalflow in the stator grooves was quantified for a set of 9 operating points at G = 0.75 mm. Theoperating points were parametrically varied with Q = 300, 500 and 700 LPM and N = 600, 800 and1000 RPM. Groove rotational flow has been quantified to estimate groove turnover rate in relationto machine operating parameters.3.3.1 Results and DiscussionTo obtain a quantitative measure for the turnover rate of the fluid in stator grooves the mean an-gular velocity was calculated for each operating point. Temporal Lagrangian particle trajectorieswere processed in MATLAB to determine the mean period ??t of the helical stator groove flow asshown in Figure 3.15. This was performed by averaging data for upwards of 40 particle trajectoriesper operating point. Peak detection in the trajectories was automated in MATLAB and approxi-mately 200 periods were measured for each operating point to obtain a mean value with suitableconfidence. Angular velocity was then calculated as:Figure 3.15: Temporal Lagrangian trajectory showing an example period measurement atG=0.75 mm, N = 800 RPM, and Q = 500 LPM.? =2???t(rad/s) (3.2)where ? is the angular velocity in radians per second and ??t is the calculated mean period.53The angular velocity was found to be directly proportional to refiner speed, as expected (Figure3.16). This shows that the angular velocity in the grooves is mainly dependent on refiner speed. Asrefiner speed is increased the influence of flow rate becomes more pronounced.550 600 650 700 750 800 850 900 950 1000 105016001800200022002400260028003000320034003600Primary Groove Angular Velocity vs Refiner SpeedRefiner speed, N (RPM)Angular velocity, ? (rad/s)Figure 3.16: Groove angular velocity versus refiner speed.Turnover rate ? was previously described as a measure of fibre network turnover accountingfor the number of turnovers that fibres are exposed to along the groove. It is proposed to quantify? as the number of fluid rotations during a single groove passing, that is:? =(LGVA)(?2?)(3.3)where LG is the groove length, VA is the bulk groove axial velocity, and ? is the angular velocity.The first term is effectively the time it takes fluid to travel along the groove length. The secondterm is the frequency of turnovers per second. This number takes into account measurements ofgroove axial velocity VA and angular velocity ? .Plotting the turnover rate versus flow rate it is clear that the number of turnovers does not varysignificantly with flow rate nor refiner speed. For all operating points the number of turnovers inthe groove per groove pass remains between 5.5 and 6.5. There does remain a small distinctionbetween turnover rate for refiner speeds of 600 RPM and those of 800 and 1000 RPM. For 60054250 300 350 400 450 500 550 600 650 700 750012345678910Groove Turnover Rate vs Refiner Flow RateFlow Rate, Q (LPM)Turnover Rate, ? (rotations per groove passing)600 RPM800 RPM1000 RPMFigure 3.17: Turnover rate versus refiner speed with lines of constant flow rate.RPM the turnover rate increases, revealing a larger dependency on VA in Equation 3.3 for lowerrefiner speeds.55Chapter 4Conclusions and Recommendations4.1 ConclusionsLow consistency refining is a complex problem of great interest to businesses and research institu-tions worldwide. Many models to predict refining performance have been developed from the firstattempt at modelling SEL by Wultsch and Flucher (1958) to the numerous models that remain inuse today. The physical refining action, machine operation parameters and fibre transport in the re-finer are all critical components of the aforementioned refining models. This work experimentallyinvestigated the effect of refining operating parameters on groove velocities relating stator grooveflow to fibre network turnover and its influence on homogoneity of refining. The key findings forthis work are summarized below.? Six particle behaviours were phenomenologically characterized representing fluid flow pat-terns in the LC refiner. These included: helical stator groove flow, groove departure, gapflow, groove entry, bolt hole entrainment, and helical rotor groove flow.? Consistent with observations by previous researchers in experimental studies of refiner flow,stator backflow and helical groove flow were confirmed inside the refiner for all operatingregions under test.? Reasonable evidence was provided to support fluid transport in and out of both primary andflow grooves at all operating points. Qualitative observations of particle trajectories showedpathlines departing and entering stator grooves. Quantitatively, axial velocity profiles alongthe groove provided evidence of net fluid transport out of the stator grooves. In addition,56velocity profiles indicated fluid transport occurring both in and out of stator grooves. Thissupports the existence of a flow similar to that of the tertiary wiping flow reported by Foxet al. (1979), though the mechanism can not be confirmed.? Stator backflow velocities were found to increase with refiner speed and decrease, to a lesserextent, with refiner flow rate. Stator backflow was found to directly correlate with refinerdifferential pressure. A linear relationship between stator backflow velocity and refiner dif-ferential pressure was found for both the primary and flow groove types.? The fluid turnover rate in the stator grooves was quantified as the number of turnovers thehelical flow experiences in a single groove passing. The turnover rate was found to remainrelatively constant for all operating regions tested in this work.4.2 Strengths and Limitations of ResearchThis experimental research has yielded several important findings that will further the understand-ing of fibre and pulp transport in the LC refiner. The project provided many challenging obstaclesfrom start to completion. The particle tracking velocimetry technique in combination with themodified refiner door offered a viable method to study flow inside the laboratory scale LC refinerand overcome these hurdles. One important limitation of the methods and results presented in thiswork is the applicability of results to refiner flow with fibre suspensions. It is understood that fibresuspensions in LC refiners are generally fluidised though it remains to have an increased apparentviscosity in comparison to water. The phenomena and trends presented in this work should relateto fibre suspension flow within reasonable error though measurement magnitudes remain suspect.4.3 Recommendations for Future WorkThe work presented in this thesis represents a small subset of the potential studies that may beperformed with the newly designed and fabricated modified refiner door in the UBC Pulp and PaperCentre. The following recommendations provide potential areas of study that the experimentalsetup would facilitate:? Manufacture additional acrylic stator plates with different geometries to study the effects ofplate geometry on the flow field.? Make use of the modified refiner door to study fibre capture using UV fluorescent nylon57fibres. It is possible to take full advantage of the capabilities of the plate-scale observationby observing floc or fibre movement throughout the refining zone.? The modified refiner door having four large viewports facilitates execllent plate-scale obser-vation. As such, it would provide an ideal platform for studying gross-refiner circulationpatterns to gain further insight into fibre and pulp residence times in the refiner.? The effect of fibre length and consistency on groove mobility was an important issue beyondthe scope of this research. It is hypothesized that the fibre length to groove size ratio willhave an impact on fibre mobility and fibre suspension fluidisation conditions; this will be animportant factor in understanding transport phenomena in the refiner.? The method of energy imparture to fibres during the refining action is still uncertain. The fi-brage theory suggests uniform fibrage, though others have observed floc refining action. Thedevelopment of fibrage, if it exists under modern day refining conditions, and the movementof flocs or individual fibres within the refiner could be further studied by making use of thelarge field of view obtainable with the modified refiner door.58BibliographyW. Banks. Design considerations and engineering characteristics of disc refiners. PaperTechnology, 8(4):363?369, 1967. ? pages 6C. P. J. Bennington and R. J. Kerekes. Power requirements for pulp suspension fluidization. TappiJournal, 79(2):253?258, 1996. ? pages 2Cospheric Innovations in Microtechnology. Cospheric Fluorescent Yellow PolyethyleneMicrosphere Properties - 1.00g/cc. URL fluorescent orange spheres density100.htm. ? pagesvii, 24B. Derakhshandeh, R. Kerekes, S. Hatzikiriakos, and C. Bennington. Rheology of pulp fibresuspensions: A critical review. Chemical Engineering Science, 66(15):3460?3470, Aug. 2011.ISSN 00092509. ? pages 2L. Drain. The Laser Doppler Technique. John Wiley & Sons, Inc., Chichester, England, 1980. ?pages 23T. Fox, R. Brodkey, and A. Nissan. High-speed photography of stock transport in a disk refiner.Tappi, 62(3):55?58, 1979. ? pages viii, 6, 7, 8, 9, 50, 57W. Frazier. Applying hydrodynamic lubrication theory to predict refiner behavior. Journal of pulpand paper science, 14(1):1?5, 1988. ? pages 6U. Ghia, K. Ghia, and C. Shin. High-Re solutions for incompressible flow using theNavier-Stokes equations and a multigrid method. Journal of Computational Physics, 48(3):387?411, 1982. ? pages 7M. Halme. Visual observation on flow phenomena in a conical refiner. Tappi Journal, 44(12):658660, 1962. ? pages 6, 9, 52W. Herbert and P. Marsh. Mechanics and fluid dynamics of a disk refiner. TAPPI Journal, 51(5):235?239, 1968. ? pages 659R. J. Kerekes. Pulp flocculation in decaying turbulence: a literature review. Journal of pulp andpaper science, 9(3):86?91, 1983. ? pages 2R. J. Kerekes. Force-based characterization of refining intensity. Nordic Pulp and Paper ResearchJournal, 26(01):014?020, Mar. 2011. ISSN 0283-2631. ? pages 6R. J. Kerekes and C. J. Schell. Characterization of fibre flocculation regimes by a crowding factor.Journal of pulp and paper science, 18(1):J32?J38, 1992. ISSN 0826-6220. ? pages 1, 2G. Kondora and D. Asendrych. Flow modelling in a low consistency disc refiner. Nordic PulpAnd Paper Research Journal, 2013. ? pages 2, 50, 52P. Leider and A. Nissan. Understanding the disk refiner - The mechanical treatment of the fibers.Tappi, 60(10):85?89, 1977. ? pages 5J. Lumiainen. A new approach to the critical factors effecting on refining intensity and refiningresults in low consistency refining. In TAPPI Papermakers Conference, Atlanta, 1990. ? pages5J. Lumiainen. Plate pattern and fibre dimensions have an effect on the performance of the refiner.In TAPPI Proceedings - Engineering Conference, pages 265?271, 1994. ? pages 6, 9, 49, 52M Polan. Effects of Refining on Flocculation. PhD thesis, University of British Columbia, 1993.? pages viii, 4O. Marques. Practical image and video processing using MATLAB. 2011. ISBN 9780470048153.? pages 26, 27D. M. Martinez, K. Buckley, S. Jivan, R. Lindstrom, A. Thiruvengadaswamy, J. A. Olson, T. J.Ruth, and R. J. Kerekes. Characterizing the Mobility of Papermaking Fibres DuringSedimentation. In Proceedings of the Transactions of 12th fundamental Research Symposium,pages 225?254, Oxford, 2001. ? pages 2Mason. The Motion of Fibres in Flowing Liquids. Pulp Paper Mag., 1950. ? pages 1A. Melling. Tracer particles and seeding for particle image velocimetry. Measurement Scienceand Technology, 8:1406?1416, 1997. ? pages 23H. Moffat. Viscous eddies near a sharp corner. Arch. Mech. Stosowanej, 2(16):365?372, 1964. ?pages 8N. R. Nasab, T. Mithrush, J. A. Olson, and D. M. Martinez. Turbulent Couette flow between twoparallel corrugated walls: The case with motion of one wall perpendicular to the corrugationcavities. Submitted to Canadian Journal of Chemical Engineering, 2013a. ? pages 50, 5260N. R. Nasab, T. Mithrush, J. A. Olson, and D. M. Martinez. On the relationship between platepattern and the flow field in LC refiners: Insight into to the groove depth effect and no-Loadpower. Submitted to TAPPI, 2013b. ? pages 8, 44N. Ouellette, P. OMalley, and J. Gollub. Transport of Finite-Sized Particles in Chaotic Flow.Physical Review Letters, 101(17):174504, Oct. 2008. ISSN 0031-9007. ? pages 31D. H. Page, J. Kosky, and J. Booth. Some initial observations on the action of the beater: What arewe doing? BP + BMA Bulletin, 28:1?7, 1962. ? pages 9F. Pan and A. Acrivos. Steady flows in rectangular cavities. J. Fluid Mech, 1967. ? pages 8Y.-F. Peng, Y.-H. Shiau, and R. R. Hwang. Transition in a 2-D lid-driven cavity flow. Computers& Fluids, 32(3):337?352, Mar. 2003. ISSN 00457930. ? pages 8D. Radoslavova, J. C. Roux, and J. Silvy. Hydrodynamic modeling of the behaviour of the pulpsuspensions during beating and its application to optimizing the refining process. In 11thFundamental Research Symposium, Fundamentals of Papermaking Materials?, pages 607?639,1997. ? pages 6M. Raffel, C. E. Willert, S. T. Werely, and J. Kompenhans. Particle Image Velocimetry - APractical Guide Part 1. Springer, Berlin, Germany, 2nd editio edition, 2007. ? pages 24H. Rance. In Symposium on Beating. Proc. Tech. Sect. British Paper and Board Makers? Assoc.,32(2):360?370, 1951. ? pages 6J. C. Roux and G. Joris. Radial distribution of the pressure in an industrial refiner. In 13thFundamental Research Symposium, Bonding and strength development, pages 409?425, 2005.? pages 6J. Senger, D. Ouellet, and C. Bennington. Effects of pulp furnish and refiner speed on residencetime in a high-consistency refiner. Tappi journal, 81(4):152?158, 1998. ? pages 10, 51P. Shankar and M. Deshpande. Fluid mechanics in the driven cavity. . . . Review of FluidMechanics, pages 93?136, 2000. ? pages 7S. Smith. The Action of the Beater in Papermaking. Journal of the Royal Society of Arts, 3655(71):38, 1922. ? pages 9J. Stachiw. Handbook on acrylics for submersibles, hyperbaric chambers and aquaria. Best Pub.Co., Flagstaff, AZ, 2003. ? pages 65B. Steenberg. In Symposium on Beating. Proc. Tech. Sect. British Paper and Board Makers?Assoc., 32(2):388?395, 1951. ? pages 661C. Tropea, A. Yarin, and J. Foss. Springer Handbook of Experimental Fluid Mechanics. Springer,2007. ISBN 978-3-540-25141-5. ? pages 23L. P. Wittberg, M. Bjorkman, G. Hkokar, U.-B. Mohlin, and A. Dahikild. Flow conditions in thegrooves of a Low-Consistency Refiner. Paper Physics, 2012. ? pages 2, 6, 9, 44, 5262Appendix APreliminary Mechanical DesignA custom laboratory refiner door was designed and fabricated to accommodate the requirementsof the low consistency refining flow visualization project. It was required to implement a modi-fication that would not compromise refining performance during normal operation nor structuralintegrity of the refiner at any time. To comply, an entirely new refiner door was designed, fabricatedand assembled. The following sections include preliminary design details and material selectionconsiderations.A.1 Preliminary DesignA set of design requirements were determined in the preliminary design phase in order to maintainintegrity and direction of the design process. The performance and interface requirements that wereconsidered were as follows:1. The experimental apparatus design shall provide clear and unobstructed visual observationof the fluid flow in the stator grooves.2. The experimental apparatus interface with the refiner shall not compromise normal operationnor structural integrity of the refiner.3. Most importantly, the safety of the refiner operators shall not be compromised under anycircumstance.In order to meet these design requirements, five conceptual designs were developed and aweighted trade-off study was performed. A set of evaluation criteria was devised in order to assessconceptual designs in the preliminary design phase. The evaluation criteria included adminstrative63(cost, lead time and design time), research potential (viewing area size and attainable operatingregions), and other (ease of lighting, refiner interface and design complexity) criterion.The five conceptual designs that were developed ranged from small-scale modifications thatcould be implemented on the current refiner door to full-scale modifications that would require anew door to be manufactured. The weighted trade-off study revealed two options to be the mostfeasible.1. A small-scale modification to the current refiner door that would allow a borescope to beinserted through the refiner door housing. The stator plate would then have a transparentsection installed that would allow observations from the under-side of the stator plate.2. A large-scale redesign that would allow for plate-scale viewing of the refining area. Thestator plate was to be made entirely of a transparent material.In the end it was decided that research potential outweighed increased design time and cost as(2) above was chosen. This option would allow observations of gross-refiner circulation patterns,flow pattern changes along the bar length and fluid transport from groove to groove in the refiningarea. In addition, it would be relatively simple to provide external light in the camera field of view.A.2 Design Considerations for Acrylic as an Engineering MaterialWhen choosing a design material for any structural device it is important to be aware of materialcharacteristics, its resistance to the environment and have an accurate prediction of loading. Thissection provides a discussion on material selection, planar acrylic windows, crazing and waterexposure effects.A.2.1 Material SelectionAcrylic is a thermoplastic material made up of synthetic polymers. Acrylic was chosen as thedesign material for transparent components because of its excellent optical clarity and strengthover competing transparent materials. Cell cast acrylic was used to eliminate residual stress dueto the extrusion process used to produce most commercially available plastics. Cell cast acrylicalso offered adequate strength and machinability properties. One caveat of acrylic over competingproducts is its decreased toughness making it a brittle material.64A.2.2 Planar Disc WindowsIt was decided to use a planar disc window made of acrylic for the modified refiner door transparentwindow. Planar disc windows are the oldest design for pressure-resistant viewports in pressurevessels. The three main benefits of planar disc windows given by Stachiw (2003) are:1. Plane shape is economical for fabrication from widely available plastic sheets.2. The window seat on the mounting surface can achieve close flatness tolerances with commonmachine shop tools.3. Sealing a plane window in its mounting is simple and may be done with a simple sheetgasket.The main disadvantage of a planar window is its sensitivity to crack initiation on the low-pressure surface due to a tensile stress condition.A.2.3 CrazingCrazing is a phenomenon that affects glassy plastics such as acrylic, polystyrene, polycarbonateand polymethyl methacrylate when used as structural components. Crazing can be defined as alens-shaped damage zone containing induced microvoids with a highly-oriented polymer chain(Stachiw (2003)). It resembles, but is generally not classified as cracking.Crazing is of particular interest to the MRD design because it is caused by the presence oftensile strain on the surface of acrylic. The low-pressure surface of a planar window is predisposedto tensile stress. For this reason the acrylic window bearing surface was designed to provide aflat stable support for the acrylic components. It has been shown that crazing does not result in adecrease in tensile strength but may lead to crack initiation and a degradation of optical properties.An extensive review of crazing causes and results is given by Stachiw (2003).A.2.4 Water Exposure EffectsWith the acrylic components in the MRD being exposed to water under test, the effect of waterexposure was researched. Acrylic is a permeable material and as such will absorb water to someextent; it has been shown that acrylic will absorb up to 2.2% water by mass with thicker win-dows subject to a lower percentage with surface exposure. Excessive water absorption can degradephysical properties and may cause swelling of the acrylic component. (Stachiw (2003))65Appendix BMechanical Design DrawingsItem Drawing Number Drawing Title1 UBC Item List UBC Modified Refiner Door - Item List2 UBC 001 UBC Modified Refiner Door - Section Drawing3 UBC 002 UBC Modified Refiner Door - Operation Mode Section Drawings4 UBC 003 UBC Modified Refiner Door - UBC Feed Pipe5 UBC 012 UBC Modified Refiner Door - Assembly - Normal Operation6 UBC 013 UBC Modified Refiner Door - Assembly - Monitoring Operation7 UBC 005 UBC Acrylic Components - Acrylic Window8 UBC 009 UBC Acrylic Components - Acrylic Plate 019 UBC 011 UBC Acrylic Components - Acrylic Window - ModTable B.1: Mechanical Design Drawing List66676869707172737475Appendix CStructural Finite Element AnalysisA parametric static structural finite element analysis was performed to determine the housing back-plate design. The outer rim of the housing backplate was held fixed and surface loading was appliedon the interior face. Static pressure loads were applied, ignoring torque transmitted to the hous-ing through hydrodynamic fluid coupling action and induced thermal stresses. American Iron andSteel Institute 304 grade stainless steel material properties were used in the simulation. The conceptdesigns varied parameters such as thickness, spoke shape and window size.To model internal pressure within the refiner a maximum expected pressure under extremeconditions was determined to be 0.5 MPa (72.5 psi). The static pressure did not act directly onthe modified refiner door backplate as internal pressure loads are transferred via the steel mountingplate (normal operation) and the acrylic window (monitoring operation). To model the loading onthe backplate an equivalent static pressure load was calculated based on relative areas as shown inEquation C.1.Peq = PoAoAeq(C.1)Where P represents static pressure and A is the internal surface area exposed to P. Po and Ao are fora solid flat disk of diameter 450 mm (17.7 in) and Peq and Aeq are the calculated equivalent valuesfor the spoke backplate.To compare 8 concept designs, two parameters were used as criterion: maximum deformationand safety factor. The safety factor was calculated from material strength and the maximum von-Mises stress in each case. In the chosen design, the maximum deformation was 0.326mm, themaximum von Mises stress was 129 MPa and the safety factor was 1.6.76Appendix DIndustrial Operating Region Data77Industrial Operating Regions for Refiner Speed and Flow RateManufacturer Plate Diameter (in) Refiner Speed (RPM) Low Med High Low Med High20 900 150 250 400 568 946 151424 720 250 350 600 946 1325 227126 720 300 450 800 1136 1703 302830 600 375 600 1100 1420 2271 416434 514 475 750 1400 1798 2839 530034 600 550 875 1650 2082 3312 624638 514 650 1075 2025 2461 4069 766542 450 775 1250 2400 2934 4732 908542 514 900 1450 2800 3407 5489 1059946 450 1025 1675 3275 3880 6341 1239752 400 1300 2150 4300 4921 8139 1627752 450 1475 2425 4850 5583 9180 1835954 400 1475 2425 4850 5583 9180 1835918 1000 70 690 265 261224 750 70 690 265 261226 750 130 1380 492 522432 600 130 1380 492 522434 600 260 1980 984 749540 480 260 1980 984 749542 500 530 2110 2006 798748 400 530 2110 2006 7987Machine Parameters Flow Rates (gpm) Flow Rates (lpm)AikawaJ&L (Double disk)Voith Twinflo78Appendix EGroove Axial Velocity Plots?15 ?10 ?5 0 5 10 15 20050100Flow Groove Axial Velocity HistogramAxial Velocity, VA (m/s)Frequency?15 ?10 ?5 0 5 10 15 20050100Flow Groove Tangential Velocity HistogramTangential Velocity, VT (m/s)FrequencyFigure E.1: Histogram of groove velocities at mid-length of a flow groove at the operatingpoint: G=0.75 mm, N=800 RPM and Q=300 LPM.79?15 ?10 ?5 0 5 10 15 20?10?5051015Groove Tangential Velocity vs Groove Axial VelocityAxial Velocity, VA (m/s)Tangential Velocity, VT (m/s)Stator FlowGap\Rotor FlowFigure E.2: Scatter plot of tangential velocity vt versus axial velocity va at mid-length of aflow groove at the operating point: G=0.75 mm, N=800 RPM and Q=300 LPM.800 10 20 30 40 50 60 70?10?9?8?7?6?5?4?3?2?10Primary Groove Axial Velocity vs Axial Position with Constant Flow RateGroove Axial Position, A (mm)Groove Axial Velocity, VA (m/s)600 RPM800 RPM1000 RPMFigure E.3: Bulk axial velocity versus groove position for primary groove at the operatingpoint: G=0.75 mm, and Q=500 LPM.810 10 20 30 40 50 60 70?10?9?8?7?6?5?4?3?2?10Primary Groove Axial Velocity vs Axial Position with Constant Flow RateGroove Axial Position, A (mm)Groove Axial Velocity, VA (m/s)600 RPM800 RPM1000 RPMFigure E.4: Bulk axial velocity versus groove position for primary groove at the operatingpoint: G=0.75 mm, and Q=700 LPM.820 10 20 30 40 50 60 70?8?7?6?5?4?3?2?10Flow Groove Axial Velocity vs Axial Position with Constant Flow RateGroove Axial Position, A (mm)Groove Axial Velocity, VA (m/s)600 RPM800 RPM1000 RPMFigure E.5: Bulk axial velocity versus groove position for flow groove at the operating point:G=0.75 mm, and Q=500 LPM.830 10 20 30 40 50 60 70?8?7?6?5?4?3?2?10Flow Groove Axial Velocity vs Axial Position with Constant Flow RateGroove Axial Position, A (mm)Groove Axial Velocity, VA (m/s)600 RPM800 RPM1000 RPMFigure E.6: Bulk axial velocity versus groove position for flow groove at the operating point:G=0.75 mm, and Q=700 LPM.84250 300 350 400 450 500 550 600 650 700 750?7?6?5?4?3?2?10Groove Axial Velocity vs Refiner Flow RateFlow Rate, Q (LPM)Groove Axial Velocity, VA (m/s)600 RPM800 RPM1000 RPMFigure E.7: Flow groove axial velocity versus flow rate for lines of constant refiner speed.85?30 ?20 ?10 0 10 20 30 40 50?8?7?6?5?4?3?2?10Flow Groove Axial Velocity versus Refiner Differential PressureRefiner Differential Pressure, ?P (kPa)Groove Axial Velocity, VA (m/s)Figure E.8: Flow groove axial velocity versus refiner differential pressure with linear trend-line. Linear trend-line R2 = 0.96.86Appendix FFluid Transport JustificationBy applying continuity to a simplified model of the flow in the groove we can make a reasonableconclusion that fluid transport occurs out of the groove. The axes of the conduit are defined as inthe figure below. Assume that the width W is large enough that variations in the z-direction arenegligible.Figure F.1: Simplified groove coordinate system.The equation of continuity for any flow field is given as:??? t+(? ??v) = 0 (F.1)Assuming a two-dimensional incompressible steady laminar flow, the equation of continuity re-duces to:(? ? v) = 0 (F.2)87which when expanded exposes the two terms:?vx?x+?vy?y= 0 (F.3)If the x-axis is along the axis of the groove, ?vx/?x is non-zero, as shown by the measuredvelocity profile (Figure 4.10). It then proves true that ?vy/?y and vy are non-zero. From theequations of motion for a Newtonian fluid it can be seen that convective momentum transport inthe y-direction is non-zero by the left hand side of the equation below. This equation representsconservation of momentum in the y-direction.?(vx?vy?x+ vy?vy?y)=?? p?y+?[?2vy?x2+?2vy?y2](F.4)It follows that in the presence of a negative axial velocity gradient, there exists fluid transportin the positive y-direction. In the case of the stator groove, this indicates net fluid transport out ofthe top of the groove.88


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



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"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items