CALIFO~~IA STATE UNIVERSITY, NORTHRIDGE A VISCOUS FRICTION MODEL A thesis submitted in partial satisfaction of the requirements for the degree of Master of Science in Engineering by Phyllis Lynn Dahl May, 1982 The Thesis of Phyllis Lynn Dahl is approved: California State University, Northridge ii ACKNOWLEDGEMENTS I express a profound appreciation to my parents; Dorothy for her incessant patience and support, and Philip for his invaluable counsel and advisement during the course of this study. A deep sense of gratitude is also expressed to my husband Ron for the long hours he spent typing and editing this document. His endless encouragement, diligence, and support have truly made a difference. Finally, I express thanks to The Aerospace Corporation for the use of their laboratory facilities and personnel to perform the experimental portion of this study. iii TABLE OF CONTENTS page LIST OF FIGURES • vi LIST OF PLATES viii NOMENCLATURE ix ABSTRACT xi Chapter 1. INTRODUCTION 1 2. BACKGROUND ON DYNAMICS OF FRICTION 3 3. 4. 5. SOLID FRICTION 3 FLUID FRICTION 5 COMBINED FRICTION • 9 MATHEMATICAL MODEL DEVELOPMENT 12 THE SOLID FRICTION MODEL 12 THE VISCOUS HYPOTHESIS 15 A COMBINED FRICTION MODEL • 17 VISCOUS FRICTION MODEL VERIFICATION • 24 VFM SOLUTION 24 COUETTE SOLUTION 25 EXPERIMENT 34 EXPERIMENTAL RESULTS 44 CONCLUSION 47 NOTES • • BIBLIOGRAPHY 52 54 iv APPENDIXES A. TEST PROCEDURE 57 v LIST OF FIGURES Figure Page 1. Solid Friction Models 2. Fluid Friction Models 3. Maxwell Model Element 4. Dynamic Viscous Friction Models 5. Bingham Model Element 6. Combined Friction Models • 7. Block Diagram of the Solid Friction Model 8. SFM Predicted Hysteresis Loops 9. Block Diagram of the Combined Friction Model 4 .. ... ... .. ... 8 ....... 11 .. ... 6 9 11 ... ....... 14 15 .. ..... 19 10. Block Diagram of the Viscous Friction Model 11. Block Diagram of the Prandtl Model 21 12. Block Diagram of 22 13. Block Diagram of 14. Block Diagram of the Newtonian Model 15. Couette Flow Problem 16. Steady-State Velocity Profile of Couette Flow 17. Velocity-Time Profile Across the Gap 18. Torque Profile at the Fixed Wall • 19. Viscometer Design 20. Physical Characteristics of Cylinders and Annuli • 38 21. Experimental Set-Up 40 22. Dry Pivot Friction Torque . the Coulomb Model . . . . the Hooke Model . . . . 22 .. 23 ............ vi ... • • • • ......... 25 27 30 32 ••••••••••••• .. 20 ... ........ 35 42 ... 45 23. Experimental Friction Torque and Velocity 24. Experimental Friction Torque-Time History 46 25. Comparison of Experimental Results With Couette Solution • SO 26. Comparison of Experimental Results With VFM Solution • • • 51 vii LIST OF PLATES PLATE page I. VISCOMETER: CUP tVITH THREE DRUMS II. LABORATORY TEST EQUIPMENT ••• III. EXPERIMENT DURING TEST PROCEDURE • viii .. ........ ........ ..... 39 43 44 NOMENCLATURE English Symbols symbol units a spatial frequency of sine term rad/cm A coefficient of sine term em/sec th n A area of viscometer wall, steady state em B steady state viscous friction coefficient dyne-cm-s/rad c coefficient of cosine term em/sec f function F Force dyne Coulomb force level dyne G shear modulus dyne/em h viscometer gap em H viscometer height em i solid friction model parameter k Maxwell model spring stiffness dyne-cm/rad thermal conductivity of fluid cal/gm- C n w F c coefficient of sine term em/sec A 2 0 m eiqenvalue of solution m n n R radius of viscometer drum em t time sec th eigenvalue ix 2 T torque dyne-em T c Coulomb torque level dyne-em Temp max maximum centerline temperature between two parallel plates oc temperature of both parallel plates oC velocity of fluid element em/sec velocity of moving wall em/sec u spatially dependent velocity em/sec un n X linear displacement em y distance from fixed wall em z distance from wall Temp u u w w th velocity term em/sec Greek Symbols y elemental shear strain rad 9 angle rad viscosity dyne-s/em \) kinematic viscosity 2 em /sec p density gm/cm (J friction model rest stiffness dyne-cm/rad friction model variable stiffness dyne-cm/rad shear stress dyne/em first time constant sec transportation delay time sec T X 2 3 2 T T n w n th time constant sec shear stress at fixed wall xi dyne/em 2 ABSTRACT A VISCOUS FRICTION MODEL by Phyllis Lynn Dahl Master of Science in Engineering A viscous friction model is developed which represents the dynamic or transient behavior of viscous friction. Stated as a general rule of finite viscous friction, the viscous friction force lags the velocity that produces it and in steady state it is proportional to that velocity. An experiment is designed and performed which validates the analytical model. A new combined friction model is also presented which represents elastic-plastic or viscous behavior. This combined model exhibits the unifying characteristic of reducing to various historical solid and viscous friction models in addition to the proposed viscous friction model. xii Chapter 1 INTRODUCTION The analyst in search of a mathematical model for viscous friction has usually assumed friction force simply to be proportional to velocity. This model has been found adequate in nearly all control systems analyses and has gone unquestioned to almost the same extent as Newton's second law of motion. In cases where the system dynamic model appears to break down and fails to duplicate experimental data over the spectrum of interest, elusive non-linear effects have been blamed and an improvement in the fidelity of the model is often attempted by adding dynamic fudge factors. A brief exploration is made of the historical models that have been used to mathematically describe the elastic, plastic and viscous flow deformation of matter. Recently, the solid friction model was developed for use in simulations of dynamic systems which involved mechanical elements that are subject to sliding and rolling friction. 1 An approach similar to the one taken with the solid friction model is used to develop a viscous friction model which represents the dynamic or transient behavior of viscous friction. The improved viscous friction model is linear; it does not include non-linear effects. Stated as a general rule of finite viscous friction, the viscous friction force lags the velocity that produces it and in steady state is proportional to that velocity. 1 2 An experiment is designed and performed which validates the proposed analytical model. Dynamical relations are derived for the force on a stationary plate produced by laminar flow of a viscous fluid between that plate and another moving parallel plate. The response of the plate drag force to a step change in velocity is derived and shows the time lag behavior. Velocity profiles across the gap are presented to help explain the physical nature of the lag effect. The force velocity transfer function for the parallel plate or Couette flow problem obtained from the step response analyses is found to be an infinite series of first order lag terms. A reasonable approximation to this viscous friction model is obtained by using the first or longest time constant lag term. A new combined friction model is also put forth that realistically represents the elastic-plastic or viscous characteristics of materials and devices. This combined model exhibits the unifying characteristic of breaking down to various historical solid and viscous friction models in addition to the solid friction model and the proposed viscous friction model. Chapter 2 BACKGROUND ON DYNAMICS OF FRICTION The dynamics of friction in engineering systems and devices that use bearings of all types, including rotating bearings, sliding bearings, dashpots, and dampers have been represented in the past by equations of the form (1) The first term on the right represents classical Coulomb or solid friction, the second term represents classical Newtonian or viscous friction, and the third term represents turbulent and other kinds of non-viscous flow. in this thesis. The latter term and its concepts will not be treated Due to the high performance requirements of modern control systems it has been necessary to seek and refine high fidelity simulation models which improve the representation of the classical approach typified by Equation 1. SOLID FRICTION Perhaps the most rudimentary part of Equation 1 is the Coulomb friction term. 2 This term has been refined and improved upon by the solid friction model, referred to in this thesis as the SFM. The improvement was brought about through the observation that experimental 3 4 data showed rolling friction to be a function not only of velocity but also of displacement. This behavioral characteristic is represented in Figure 1 which displays friction torque vs. angular displacement for continuing positive velocity motion. The torque-displacement characteristic for positive velocity in this Figure is seen to be generally elastic-plastic. ! Coulomb ( 1784) St. Venant - Prandtl (1904) - - - ( Davidenkov (1938) Solid Friction Model ( 1968) DISPLACEMENT , 8 Figure 1. The model of Prandtl, displacement, 3 Solid Friction Models which is elastic up to a characteristic ec , and subsequently plastic at a constant friction level, T , was an improvement over the sudden step up to the constant c 5 friction level, T , of the earlier Coulomb model. c model The DavidenkoV 4 was a later improvement over the Prandtl model for elastic-plastic materials that mathematically represent a continuously changing function from elastic to plastic behavior. Both the Prandtl and Davidenkov models have been available for many years, but not generally used in the study of dynamical systems. The reason for this is that there was no simple mathematical way of representing the hysteretic behavior in the repeated or successive reversals in velocity for other than simple Coulomb friction prior to the introduction of the solid friction model. The development of the higher fidelity solid friction model coincided with the high speed and high capacity capabilities that were being developed in modern computer technology. FLUID FRICTION Not as crude, generally speaking, as the early Coulomb model of solid friction is the Newtonian model of viscous friction, 5 which is the second term in the general friction Equation 1. This simple Newtonian model depicted in Figure 2, is linear, i.e., 6 ---Newton ( 1687) I Blasius (1908) Nikuradse (1932) 1UJ :::J 0 0::: 0 1- • VELOCITY , 8 Figure 2. Fluid Friction Models Generally, it has been known that a non-linear dependence on velocity is present in the mathematical representation of liquid shearing friction. The deviation from linear behavior is referred to as the pseudoplastic or dilatant property of liquids and is a subject of interest in the field of rheology. In a steady state velocity sense, any fluid response not representable by Equation 2 may be termed non-Newtonian. In the past, a number of relationships have been used in an effort to model this non-linear behavior. The simplest relationship that fits experimental data being the power law model, 7 (3) For most materials the exponent n is less than unity and the material flows more easily the faster it is sheared and the viscosity thereby decreases with shear velocity. pseudoplastic. This type of behavior has been termed For a few materials, the exponent n is greater than unity and the viscosity increases with shear velocity. This type of behavior has been called dilatant. For most dynamic non-steady state modeling problems the linear Newtonian model has been adequate and used almost exclusively in the study of dynamics of systems with the usual engineering devices. However, it is noted in the Newtonian model of Equation 2 that viscous friction changes instantaneously with velocity. It has been known, for example in shear dampers and dashpots that operate in the laminar flow regime, that dynamical effects which are not represented by the simple linear Equation 2 are present. This is taken to mean that no dynamical time dependence is represented in Equation 2. An improvement to the Newtonian model can be made by employing the Maxwell model. 6 This model is represented by the spring and dashpot series combination, shown in Figure 3, and has the equation of motion (4) The steady state behavior of the Maxwell model is the same as the Newtonian model, i.e., T • = Ba. However, a dynamical term is present as a result of the introduction of the spring term to the model which brings in elastic behavior in addition to the viscous behavior. 8 Figure 3. Maxwell Model Element The dynamic response behavior of the Newton and Maxwell models is illustrated for a step change in velocity in Figure 4, where the Maxwell model shows a friction varying continuously with time, and the Newton model shows friction changing instantaneously with the step change in velocity. Thus, the Maxwell model appears to appropriately represent the observed dynamical behavior in shear dampers and dashpots. However, the solutions for the shear damper and dashpot equations for non-steady Couette flow do not indicate a physical 7 8 mechanism that would explain the spring in the Maxwell model. ' In fact the Couette solutions indicate that the dynamic term in Equation 4, (B/k)T, if it is an appropriate model, would be due to inertia effects of the fluid mass in the shear damper gap rather than a spring effect which is totally absent in a pure Newtonian fluid. Nonetheless, the Maxwell model would appear to appropriately introduce, as a first approximation, a dynamic term which represents the dynamic effects of 9 . . Ir ~ WI ;::) Ol 0:: ~I I (1965) I I I TIME , t T = J(9, 9,• t) Dynamic 9• Figure 4. Couette flow. = Function of Time u(t) Dynamic Viscous Friction Models Therefore, the Maxwell model appears to be a good model in representing the dynamics of viscous friction. COMBINED FRICTION When the solid and viscous friction effects are combined, it might be expected that the Maxwell model would adequately represent linear elastic and Newtonian viscous combined friction. This turns out to be the case in some situations but it should be pointed out that the elastic term does not allow for the plastic deformation characteristics of solid friction. The elastic portion is due to elastic properties 10 within the fluid and cannot be considered as part of the solid friction component. Bingham9 evidently understood this shortcoming and added a Coulomb friction element to the elements of the Maxwell model, as shown in Figure 5. The Bingham model has the equation of motion, (5) where, T = -k(e 1- e) 2 (6) The responses of the Maxwell and Bingham models to a step input in velocity are shown in Figure 6. The Bingham model exhibits a linear response with time due to the linear spring for the region of friction torque below the Coulomb friction level, T , and a response identical c to the Maxwell response thereafter. A SFM in conjunction with a viscous component of friction could be expected to behave in a manner similar to the Bingham model since the SFM has a steady state Coulomb level as well as a static spring effect. 11 T Figure 5. Bingham Model Element Bingham (1923) Maxwell (1855) TIME , t T = j (8, 8, t ) Dynamic Function of Time • 8 = u (t) Figure 6. Combined Friction Models Chapter 3 MATHEMATICAL MODEL DEVELOPMENT The solid friction model was developed for use in simulations of dynamic systems which involved mechanical elements that are subject to sliding and rolling friction. The foundation of this model was based on the hypothesis that solid friction resulted from quasi-static contact bonds which were continuously formed and subsequently broken. This SFM has been proven to accurately simulate transient torque behavior seen in tests in the laboratory on ball bearings and other devices that exhibit solid friction. It is desired to approach viscous friction from a similar viewpoint to see if it is possible to develop a viscous friction model (VFM) that more accurately represents the dynamic or transient behavior of viscous friction. THE SOLID FRICTION MODEL Solid materials that undergo elastic-plastic deformation as a function of time have a time rate of deformation given by; dy dt 1 dT = G dt (7) where we have considered the elemental shear strain, y, to be produced by the shear stress, T. The shear modulus, G, is taken to 12 13 be a non-linear function of y or T and so it includes elastic as well as plastic shear modulus characteristics. Now convert the elemental or microscopic Equation 7 to macroscopic terms by substitution of the torque T for the angular deformation e for T, y, and the angular stiffness b for G. Solving • for T we obtain dT Cit= • (8) 2:9 Equation 8 was the key relation upon which the non-linear solid friction model was built. The non-linear stiffness o is usually considered to be a function of the deformation angle e but it can alternatively be considered a function of the shear stress or torque. A simple general stiffness relation was empirically developed that is expressed as 10 T T •li T < Tc sgne c (9) where o is the rest stiffness or the slope dT/de when T : 0, T c is the saturation or running friction, and i is an exponent that describes the plasticity or brittleness of the material or the friction process. The stiffness can conversely be thought of as the slope of the torque vs. angular deflection characteristic, E dT (10) =de and the shape of the characteristic can be found analytically by substituting Equation 10 into Equation 9 and integrating. example, for i =1 • and 9 > 0 we obtain for 9 =0 at T = 0, As an 14 T T = -(~)9 T (1 - e a ) c • e > o (11) • When the rate reverses, i.e. 9 < 0, a new solution must be obtained, using as initial conditions those values of T and 9 existing at the instant of rate reversal. Successive rate reversals produce the hysteresis behavior of solid friction. The hysteresis behavior of solid friction and of solid materials is nicely represented by the solution of Equations 9 and 10 as depicted in the block diagram in Figure 7. Typical torque vs. deflection hysteresis loops obtained from the solution of Equations 9 and 10 are illustrated in Figure 8 • • • (J T 1 T s • .E(T, 9) • sgn 9 Figure 7. Block Dia~ram of The Solid Friction Model Examination of this solid friction or materials properties model reveals that there is a time dependence inherent in the solution or simulation of the process. This is the result of using •9 as the input independent variable and solving for the torque as a function of time in the simulation. Note that if torque is plotted vs. angle as in Figure 8, the time dependence does not explicitly appear but the rate 15 dependence is evident from the motion reversals. Note· also that the hysteresis curves are independent of the rate magnitude from Equation 9. These observations lead to the conclusion that the solid friction model represents solid friction and solid materials elastic-plastic behavior adequately in a manner which is independent of the rate magnitude as is the case from the facts of experimental observations. T Limit , Tc Figure 8. SFM Predicted Hysteresis Loops This representation is what was desired and expected of the solid friction model. Questions that naturally arise next are: 1) can viscous friction be represented somehow by the solid friction model, or 2) can the solid friction model be modified generally to include viscous friction? The following paragraphs will shed light on these questions. THE VISCOUS HYPOTHESIS In an heuristic approach to adapting the SFM to apply to viscous friction, suppose we look at the solid friction model equation 16 • T • • (12) T = a(l - T sgn e)e c where it is assumed that i=l and that the saturation level of friction Tc varies with velocity according to Equation 3. Equation 12 then • becomes, for Newtonian fluids with n = 1 (ie. T = Be) c B• -T + T 0' = Be• (13) • The sgn term has been dropped because the e which was introduced has the sgn property and so obviates the need for it. Equation 13 could be construed as a dynamical equation for viscous friction torque. It is identical in form to the Maxwell rheological model Equation 4 and is referred to in this thesis as the viscous friction model (VFM). It has been arrived at or deduced in this section independent of the Maxwell rheological model using the solid friction model as a basis. The VFM indicates that the viscous friction torque T lags velocity •a with a time constant of B/a seconds. The parameter B/a for fluids was thought to be related to the time required for the laminar velocity profile to change, as in Couette flow, however, it has a form identical to the Maxwell model with the friction stiffness a equivalent to the spring stiffness k in the Maxwell model. The steady state viscous • as could have been anticipated. friction torque is Be, The validity of this viscous friction model has not been established at this point but the groundwork has been laid by the hypothesis that a combination of viscous and solid friction is accomodated within the framework of the basic solid friction model. Because of the usefulness of the solid friction model in modeling 17 and simulating many friction processes imbedded in dynamic systems, it was deemed desirable to include or add the viscous contribution to the total friction of elements in a unified solid and viscous friction model. So before proceeding to the testing of the viscous friction model as postulated in Equation 13, we shall treat the combined effects of solid and viscous friction in a combined friction model. A COMBINED FRICTION MODEL The SFM was originally developed from Equation 7 which was applied in this thesis as an infinitesimal element material property relation derived from elastic-plastic laws. Following this same approach we depart from the macro world of friction producing devices and focus on the microscopic deformations of materials at this time to include both elastic-plastic and viscous properties. The deformation processes involve the fundamental properties of materials in a microscopic sense, and so the combined friction model (CFM) to be derived can be considered a combined property rheological model in the microscopic sense. Materials that undergo elastic-plastic deformation in a quasi-static fashion and at the same time undergo viscous flow deformation can be considered to have a total time rate of deformation gi ven by ll (14) where we have considered elemental shear strain y produced by the 18 shear stress T. The net shear strain rate dy/dt is the sum of the elastic-plastic strain rate (1/G)(dy/dt) and the viscous contribution (1/~)T. The shear modulus G is taken to be non-linear and so includes the elastic as well as plastic shear stiffness characteristics. Thus, G is a function of strain G(y) or alternatively can be modeled as a function of stress G(T). This is the approach taken in arriving at Equation 8 and is also taken here. The viscosity ~ is assumed to be a constant or a function of • • temperature but not T, y, G, T, nor y. We now convert the basic governing Equation 14 from elemental or microscopic terms to macroscopic terms by substituting T for T, e for y, B for ~, and E for G. In this context, T stands for torque, e for angle, B for drag or friction coefficient of angular velocity, and E for angular stiffness. Then, solving the resulting • equation for T we obtain dT • E dt=Ee-BT (15) where we assume the same form for the stiffness as in Equation 9. • I E(T,e) = a 11- T ., i T sgne T < T c c where again a is the "rest" stiffness and T c is the usual Coulomb or saturation level of torque in the quasi-static elastic-plastic portion of the deformation. Now that viscous flow deformation is present, under the usual stable condition that (dT/dt) approaches zero in steady state in (16) 19 appropriate dynamic situations, Equation 15 can be solved to yield • T = Be (17) which is the classical Newtonian viscous flow drag or friction relation. The block diagram of Equation 15 is presented below in Figure 9. This block diagram represents a math model for a visco-elastic-plastic material or a system component or element that exhibits such behavior • • • T 8 1 T s 1 B Figure 9. Block Diagram of The Combined Friction Model If we assume from Equation 16 that I is constant, this implies that T/T c is very small and can be considered negligible. Then I can be replaced by a pure elastic characteristic a and the block diagram can be configured as shown in Figure 10. This block diagram represents, in at least one case, the dynamic lag behavior of a viscometer or dashpot or other element as described by the Couette flow solution in Chapter 4. model of Equation 17. It is an improvement to the simple Newtonian 20 • 8 (J • T .. + \, - -s1 T ' (J B Figure 10. Block Diagram of The Viscous Friction Model It should be pointed out that the combined friction model is valid for a single physical macro or micro element and that it may not be applicable to describe two separate elements. As an example, if solid friction and viscous friction occur simultaneously but are independent, then a single model is not appropriate and two separate models should be employed and connected in series or parallel or as the physics of the system require. The combined friction model shown in Figure 9 encompasses most of the features of the various friction models which were described in the background section. This unifying characteristic is unusual and is projected to have many applications. Elastic-Plastic Special Cases When the steady state viscous coefficient B is increased to infinity the CFM reduces to the Solid Friction Model, represented in Equation 18 and Figure 7. 21 • I9 = • T .I i . all - - sgne e I T I c ITI < T c T (18) ITI = T c 0 When the steady state viscous coefficient B is increased to infinity and the friction function exponent i is set to zero the CFM reduces to the Prandtl model, as noted in Equation 19 and Figure 11 • ae• ITI < T • c (19) T = ITI > T c 0 - • • T 8 1 T s Figure 11. Block Diagram of The Prandtl Model If the steady state viscous coefficient B is defined to be infinite and the rest stiffness a becomes infinite, the CFM reduces to the Coulomb model, represented by Equation 20 and Figure 12 • T = T sgne c (20) 22 Figure 12. Block Diagram of The Coulomb Model Finally, if B becomes infinite and the Coulomb friction force, or saturation level of torque T , becomes infinite then the CFM reduces c to the linear elastic Hooke model, represented by Equation 21 and Figure 13 below. T = oe (21) ____8_·~-~~~---~--~----T______ Figure 13. Block Diagram of The Hooke Model Viscous Special Cases When the Coulomb friction level, or saturation level of torque T c is raised to infinity, the CFM reduces to the viscous friction model where the angular stiffness term rest stiffness o. ~ is constant and replaced by the The VFM is found from the CFM Equations 15 and 16 to reduce to Equation 13 and is represented by the block diagram in Figure 10. 23 When the saturation level of torque T in the CFM equations iS c raised to infinity and the time constant of the model B/a is defined to be zero, the CFM reduces to the classical Newtonian model, represented by Equation 22 and Figure 14 • T = Be• 6 Figure 14. (22) ·1. . _s_ _:-__T_,.,..._ Block Diagram of The Newtonian Model Chapter 4 VISCOUS FRICTION MODEL VERIFICATION This section presents both an analytic and an experimental validation of the VFM. An experiment was envisioned which would verify both the viscous hypothesis and the viscous friction model. The test would utilize a viscometer where a viscous fluid is contained in a gap between an inner cylindrical drum and an outer cup. The torque response to a sudden stop from a steady flow was expected to reveal the characteristic lag phenomenon. A solution to the Navier-Stokes equation for flow between parallel plates would relate the physical characteristics of the experiment to the solution predicted by the VFM. THE VFM SOLUTION The analytical solution of the VFM for the desired test conditions is obtained by assuming the experimental angular velocity of the viscometer drum to be a step change from a previously existing steady state velocity, stopped. •a, to zero, i.e. the viscometer drum is suddenly The resulting steady torque induced by the fluid flow prior to time zero is, from Equation 13 T = B\l• (23) The solution from the VFM Equation 13 for this viscous drag problem is, 24 25 (24) COUETTE SOLUTION The unsteady viscous fluid flow field resulting from time varying boundary conditions is referred to as Couette flow. To analytically determine the appropriate experimental parameters for the linear viscous friction model, it is desired to solve the one-dimensional Navier-Stokes equation. A solution for a non-steady Couette flow condition is calculated for the case where the relative motion of the parallel plates is suddenly stopped or reversed. 12 13 14 ' ' The viscous flow system is modeled as having two infinite parallel plates, initially one moving relative to the other with velocity uw' and with a viscous fluid of viscosity ~ or kinematic viscosity v between them as shown in Figure 15. Wall y FoRAG = 11 d u] dy Figure 15. Aw WALL Couette Flow Problem p ' 26 The assumptions which are made for this derivation are: 1. The properties of the fluid that we are dealing with may be treated as continuous and homogeneous. 2. The tangential stresses are neglected. 3. The temperature is considered to be constant, thus the loss of heat is ignored. 4. The density of the fluid is considered constant. S. The fluid velocity at a surface is tangential to that surface and equal to the surface velocity. Selection of the x-axis along the wall in the direction of u w leads us to obtain the one-dimensional Navier-Stokes Equation for viscous flow au at (25) -= The boundary conditions for the problem being considered are: 1) The fluid velocity at the stationary wall is zero. u(O,t) 2) =0 (26) For all t < 0, the moving wall moves at velocity uw prior to bringing it to a sudden stop at t = 0. Therefore, at t u(h,O) 3) = 0, = uw (27) The flow is assumed to be laminar. The geometry of the steady velocity profile of the fluid at time t < 0 is linear as shown in Figure 16. Therefore, for t = 0 u u(y,O) = hw y (28) 27 y __._-L-------------y=O Figure 16. 4) Steady State Velocity Profile of Couette Flow For t > 0 , u(h,t) =0 (29) Employ the method of separation of variables and try the solution u(y,t) = emtU(y) where U(y) is the gap dependent variable. (30) Substitution of Equation 30 into Equation 25 yields (31) or (32) If we let a 2 Equation 25 now becomes m \1 (33) 28 (34) A solution to Equation 34 is U = A sin(ay) + C cos(ay) (35) Satisfying the boundary condition u(O,t) c =0 requires that = 0 (36) In order to satisfy the boundary condition u(h,t) = 0, we must have A sin(ah) = 0 (37) or n1f = 0,1,2,3, n a=- h •••• (38) For each n we therefore obtain a solution to Equation 34 U n =A n sin(n1ry) h (39) By substituting Equation 38 into Equation 33 we find the eigenvalues 2 m = n (mr) v h (40) Using Equations 39 and 40 in Equation 30 (41) Summing over all n, we obtain the general solution 29 2 -(E.!) vt ... u(y,t) = f Ae h sin(n~y) (42) n=l n To satisfy the initial condition given by Equation 28 ... u(y,O) = f Ansin(n~y) (43) n=l We must therefore expand u(y,O) in a half-range series of sines, 2 h uw n~y An = -h I -h y sin(-h) dy (44) 0 and if we let n~y/h = z 2u A n A n = n~ ~ n 2~ I [z sin(z)Jdz (45) 0 = (-l)n+l 2u w (46) n~ By substitution into Equation 42 we obtain the solution to our problem 2 .., u(y,t) = r n=l n+l 2u (-1) -2:. n~ -(E.!) e h (47) A plot of the velocity profiles for several times, t > 0 is shown in Figure 17. The velocity profile, as a function of time, tends asymptotically with time from the initial steady state linear velocity profile to the final zero velocity as seen in the figure. 30 U " UWALL Figure 17. Velocity-Time Profile Across the Gap Finding the force at the fixed wall, TAw (48) au(O,t) ay (49) Fw (O,t) = where T = l.l is the shear stress at the wall, and Aw is the area of the fixed wall Then = r 2 n1T 2u w -(--h) vt au(y,t) = (-l)n+l ~ cos(n1Ty) --e ay n=l h h n1r (SO) 31 au(O,t) = ay Q) r (-l)n+l (51) n=l Using equations 49 and 51 in 48 Fw (t) = 2AwI.IUw h Q) r (-l)n+l (52) n=l The torque on the outer drum (from the experimental configuration) T (t) w = (R + h)F (t) w (53) where, R _ radius of the viscometer inner drum h _ gap width Finally, • 2A 9Rvp(R+h) T (t) = _w_~-w h ao L (-l)n+l (54) n=l where •e = u /R = viscometer drum angular velocity. w A plot of T (t)/T (0) is shown in Figure 18. w w In the experiment to be described in the next section, the viscometer drum was driven such that the steady angular velocity of the drum was reversed cyclicly from +u to -u to +u , etc. w w w The velocity profile solution given by Equation 47 is not valid for this input condition. The correct solution was found to be 32 Analytical Solution: T.w (t) = ~ LJ 2 (-1 o [ -a ~ ~ ...: -·-· ·-·-. .~ -- -. Tw(O) '· '· '·, 0 )n+1 exp [- (n1T)2 -·- vt n=1 '· '· '· ................ h '· -. - .............. __ ·-............... -·- 2.0 1.0 Figure 18. l Torque Profile at Fixed Wall 2 u(y,t) uwy 4uw m (-l)n+1 h "If n=l n = --+-r -(~) vt h e sin(n1ry) h (55) The resulting torque on the viscometer cup was then found to be identical to Equation 54 in form although a bias term is present and a factor of two in magnitude is introduced. If we equate the VFM solution, Equation 24, to the dominate first term from the Couette flow solution in Equation 54 the following approximating equation is obtained. 33 (56) T w We then obtain expressions for the VFM parameters in terms of the experimental configuration. B The viscous friction coefficient is 2A vpR(R + h) = ___ w__~------ ( h dyne-em-s rad ) (57) and the time constant of the VFM is (58) and the apparent shear stiffness of the VFM is C1 = 2(~) 2 ~ 2 AwR(R + h) ( ph dyne-em-s rad 2 ) (59) Equation 57 shows us that B has the proper units of a viscous friction coefficient and thus corroborate the VFM solution. Also, the B/a coefficient has the units of time and thus behaves like a time constant, as the VFM stipulates. a It is interesting to note that the coefficient seen in Equation 59 exhibits the units of a stiffness, but contains no terms which indicate elastic properties of the fluid. This is expected because no elastic properties were stipulated in the definition of the Couette flow problem. However, the VFM can be interpreted to contain what in fact is an inertia or mass property of the fluid and not a stiffness property at all. If one views the dynamic lag characteristic in the VFM as the time required for the velocity profile to change, this implies that viscosity and inertia 34 properties of the fluid are governing as Equation 58 suggests. The viscosity and inertia or density properties of the fluid were not explicitly modeled in the VFM and yet the VFM response shows their effect. This leads to the conclusion that the a or stiffness term in the VFM is related to the viscosity, density, and geometry of the Couette boundary value problem as Equation 59 indicates and is not a stiffness at all. EXPERIMENT Through experimental testing, it is desired to verify and validate the postulated viscous friction model. It is expected that the experimental data will show the characteristic lag for a given viscous fluid and test configuration. This subtle effect must be measured within the capability of test hardware and thus poses considerations for the experiment hardware design and assembly. We wish to observe the behavior of a liquid in a viscometer when it is viscously sheared under isothermal conditions in a narrow annulus between vertical coaxial cylinders, one cylinder rotating with respect to the other. The viscometer is composed of a fixed vertical outer cylinder, or cup, which is mounted on the torque measuring device. The experimental set up to accomplish this is shown in Figure 19. The torque on the stationary cylinder or cup is measured as a function of time after the cessation or reversal of motion of the inner rotating cylinder or drum. The torque is measured by a device which moves minutely with the torque being applied to it by the cup. The inner 35 Torque Motor Drive Shaft Pivot Bearing spotface :..) 20• drill angle I m \bpunch ·. : : . go· . . ,: Mounting Platform Torque Measuring Device Figure 19. Viscometer design drum is accurately aligned with the outer cup so that the annulus gap is uniform. bearing. The alignment during rotation is maintained by the pivot By fine vertical adjustment of the drum coupling to the 36 motor drive shaft which can support the weight of the drum, the pivot bearing solid friction effects are minimized. The pivot bearing was selected in the design to allow a vertical adjustment that would minimize vertical loading on the bearing and thus the pivot bearing solid friction contribution to total friction. The drum is coupled to the drive shaft of a torque motor which is driven so that any prescribed drum motion time function may be achieved. In order to minimize the viscous drag from end effects not accounted for in the Couette solution the following methodology was employed. A large space with tapering walls was machined on the bottom of the inner cup to trap an air bubble. The air has low viscosity compared to the oil in the gap and so causes a minimal drag contribution, which can be neglected, compared to the annular gap viscous drag friction. Several items of concern arise when the viscometer dimensions are to be finalized. First of all, the test must be operated in compliance with the assumptions made for the Couette flow solution. Second, additional geometric considerations must be recognized ie., the flow must be laminar, the temperature and end effects must be negligible, and the lag effect must be substantial enough to be measured. To determine whether the laminar flow condition prevails, we check the point at which turbulent flow may be anticipated in the gap due to shearing the fluid at very high rates. The critical velocity is given bylS u w (60) 37 where u w is the velocity at radius R of the inner cylinder, and v is the kinematic viscosity of the fluid. The laminar flow constraint is to limit the drum rotation rates to values below this velocity. Temperature control is important because the lag time constant is proportional to viscosity which is temperature sensitive. With the complicated equipment used it was deemed impractical to control the temperature of the entire apparatus and thus thermal influence was unavoidable. In evaluating whether the viscous sample will remain nearly constant at room temperature leads us to the analysis of the problem of heat generation for Couette flow. The solution to this problem of a Newtonian fluid in simple shear between two parallel plates with both plates held at temperature Temp u Temp - Temp = max w where Temp max 2 w n p h w is given as 16 2 ---==---8kt (61) is the maximum centerline temperature of the fluid in the viscometer gap and kt is the thermal conductivity of the fluid. If the maximum temperature differential in the fluid across the gap is small then our assumption of an isothermal system is valid. Otherwise, a large temperature differential would cause an error between the experimental and predicted time constants. An analysis of Equation 61 relative to this concern suggests that we minimize the drum velocity and the gap width for a given test sample. In order to assure that the time constant was sufficiently large to be measurable, a computer program was created to help provide the best design criteria for the sizing of the viscometer. The aforementioned laminar flow and thermal criteria were applied to the Couette flow 38 solution Equation 54 with the added restriction that the lag time constant be greater than 20 milliseconds. It was anticipated that 20 milliseconds would be within the measurement capability of the experiment equipment. In order to achieve the desired measurement for a large range of viscosities, three different drum diameters were considered for fabrication. The characteristics of the viscometer cylinder and annulus determined by this program are given in Figure 20. PLATE I shows the inner and outer cups used in the experiment. Cylinder Radius (em) Annulus Length, H (em) Annulus Width, h (em) - - Outer 2.5400 Inner I 2.3813 10.1600 0.1588 Inner II 2.2225 10.1600 0.3175 Inner III 1.9050 10.1600 0.6350 Figure 20. Physical Characteristics of Cylinders and Annuli The experimental layout which is designed for these validation tests can be seen in Figure 21. A periodic forcing function is achieved by a square wave output from a function generator which is used as a command input to an angular rate servo. consists of a servo amplifier, a D. c. The rate servo torque motor, and a tachometer. The torque motor output shaft angular velocity is measured by the tachometer, and fed back to the servo amplifier which determines the error between the command signal and the tachometer signal. It then 39 PLATE I. VISCOMETER: CUP WITH THREE DRUMS amplifies and applies current correction to the torque motor to reduce the error. The torque motor output shaft drives the inner drum of the viscometer through a coupling device. The inner drum rotates about a pivot bearing that was designed to minimize the dry friction torque contribution to the measured torque. The viscous fluid between the inner and outer cylinders is placed under shear by the angular motion of the inner cylinder. The torque on the outer cylinder is expected to give us the decay characteristic predicted by the VFM solution Equation 56 and / or the Couette solution Equation 55. The torque measuring device used at the base of the viscometer is a McFadden Electronics Co. 40 SERVO AMP. FUNCTION GENERATOR POWER SUPPLY VISCOMETER HP PLOTTER FLOPPY DISK O·SCOPE NORLAND DIGITAL t--'-T--; ANALYZER PWR. SUPPLY I TORQUE DYNAMOMETER I I Figure 21. Experimental Set Up Servo Torque-Balance Reaction Dynamometer Model llOA. The mounting platform is part of the torque meter and is air bearing supported for axial and radial loads. When a torque is applied from the viscometer test fixture to the platform, a small angular displacement of the mounting platform occurs. This displacement is sensed by a feedback 41 position transducer which in turn sends a signal to a control and power electronics unit that supplies current to a torquer. The torquer output torque reacts against the input applied torque to reduce the displacement error and thus provide a torque balance, and limits the platform motion to very small angular displacements. It is the output of the torquer which is measured to be opposite but equal to the torque produced by the viscometer outer cup wall. The current to the torquer is proportional to torque produced and is readily measured as a voltage which can be recorded. The continuous recording of the torque decay characteristic is achieved through the use of a Norland 3001 Digital Processing Analyzer. The three signals of interest, rate command, viscometer drum angular velocity, and viscometer cup torque measured at the stationary viscometer wall, are continuous signal inputs to the Norland. The Norland digitizes and stores over 1000 data points over a selected period of time and operates on these input signals. Peripheral equipment such as an oscilloscope, x-y plotter and floppy disc memory aid in the manipulation and recording of the data. A test procedure was created in order to standardize the testing process. An outline of the test procedure can be found in Appendix A. The dry pivot torque characteristics of the test hardware are presented in Figure 22. The viscometer drum position and rate measurements and viscometer cup torque response are plotted as a function of time. The torque responds to a step change in rate rather quickly, taking approximately 25 milliseconds to reach the maximum torque overshoot. Therefore, the 20 millisecond response assumption used in the design criteria computer program was not far off. Care had 42 Torque on Cup 0 I 0 0 0 Viscometer Drum Rate N I 0 20 40 60 80 Time, msec Figure 22. Dry Pivot Experiment, Friction Torque to be exercised to assure that the predicted decay time constant of the VFM be a magnitude or two greater than this. The validation test was performed using an oil with a kinematic viscosity rating of 164 centistokes, about the consistency of cooking oil. An inner cylinder with 0.6350 em annulus width was chosen for the selected viscosity oil. PLATE II displays the equipment assembled for the test, and the viscometer undergoing dynamic testing is shown in PLATE III. 43 PLATE II. LABORATORY TEST EQUIPMENT 44 PLATE III. EXPERIMENT DURING TEST PROCEDURE EXPERIMENTAL RESULTS The command rate was kept substantially under the predicted rate for the onset of turbulance for the test geometry. Figures 23 and 24 are presented as raw data obtained from the experiment. Figure 23 shows the experimental friction torque and velocity curves as a function of time. The torque responds to a step change in rate quite a bit slower than the dry friction torque response, as anticipated. The "viscous" torque response takes approximately 250 milliseconds to reach 45 the maximum torque overshoot, ahout two magnitudes greater than the solid or dry friction effect, as was desired. This same "viscous" friction torque versus time plot is presented on a log-linear format in Figure 24. The delay experienced as a function of the viscous fluid is clearly visible. 150~--------------------------------------------------~ - Velocit (,) Q) .5 ~ ~ Q) - -.... "'C N >. 0 !: Q) :::l C" (,) 0 ~ ~ ~ 0 0 -:5 -150 80 120 160 Time (msec) Figure 23. Experimental Friction Torque and Velocity 46 "" ' ' ' .1 .~ ~ "\ '~ N 0 ~ s:: 1 \ \ '\. \ ~ '\ ' .001 40 . 60 80 100 120 140 160 Time (msec) Figure 24. Experimental Friction Torque Time History The comparison of this raw data and the results predicted from the VFM are discussed in the following section. Chapter 5 RESULTS AND CONCLUSIONS The agreement between theory and experiment for the case of a Newtonian fluid is illustrated in Figure 25 which shows the experimentally obtained friction torque and velocity curves plotted as a function of time. A theoretical fit derived from the Couette solution equation is compared to the experimental data. The Couette solution predicted a slightly higher torque measured at the wall and a slightly smaller time constant. 3.5 and 8.0 percent respectively. The differences were determined to be This agreement appears to justify the neglect of end and temperature effects, as well as to indicate the accuracy achieved in this kind of experiment with the apparatus described in Chapter 4. Generally speaking the VFM models the viscous time lag to a relatively high degree of accuracy using the predominate, or first constant plus a time delay. If the VFM without the empirical time delay is used, a relatively poor approximation is obtained as seen in Figure 26. An increase in accuracy of the VFM without the time delay would be obtained with the use of additional first order VFM lag models in parallel that can be derived based on the accurate Couette solution. The delay effect was not expected nor predicted by the simple VFM Equation 13. Therefore, the VFM Equation 13 should be modified to include the transportation delay effect. The Laplace transformed Equation 13 with the delay included takes the form 47 48 (62) where the single time constant Tl is as before 2 T 1 = -(£) (63) '\1 1T and the delay time constant that fits the test data and the Couette flow solution is TD = Tl ln 2 (64) The validation of the viscous friction model presented above in addition to the solid friction model validations from 1968 to the present lead to the conclusion that these models are useful and that the combined friction model if and when similarly validated will be useful in the future. The special cases of the combined friction model were presented in section three. It is interesting to note that the mathematical reductions and assumptions made there have physical meanings. The common mathematical limiting conditions made for the elastic-plastic special cases had the effect of increasing the steady state viscous coefficient, B, to infinity. On a physical basis, this means the viscosity approaches infinity, or this viscous material behaves as a solid. Conversely for the viscous special cases, the Coulomb saturation level of torque was raised to infinity. This implies that the solid material characteristic with a saturation torque level no longer exists, and the solid material behaves like a viscous one. 49 It must be reemphasized that the combined friction model is valid for a single physical material element only. If solid friction and viscous friction occur simultaneously, but are independent, then two separate models should be employed and connected in series or parallel as the physics of the system require. Because the combined friction model does not contain an inertia term, the viscous lag effect seen in the Couette solution is not actually present in its first order differential equation. The VFM does not contain an inertia term either, however, the stiffness term a can be interpreted to be such if the solid material properties producing friction are negligible. The VFM then becomes useful in representing the dynamic lag characteristic of viscous dampers and other devices. so .10 --~ Experimental Couette Fit for T w = .166 & - T1 =.027 +150. N 0 I c: Velocity Q,) ::::l C" :... ~ c: - EXPT DATA COUETTE FIT % DIFF. .027 .025 8.0 .166 .172 3.5 0 u Tl :... {SEC) u.. - Tw {IN OZ) u Q,) If) 0 ' C) -Q,) 0 "C >. ·u 0 ~ -150. 40 60 80 Time (ms) Figure 25. Comparison of Experimental Results With Couette Solution 51 T(s) BO(s) h2 7j = rr2 " e-ros = 7jS+1 T0 = r 1 Jn 2 ~~~~~~~~~~----~~------~~----------------~--- 1Q) :::l 2" ~ c: 0 - ·-u ""' LL. 6 VFM without Delay T(s) • BO(s) = 1 TS+1 8~---L----L---~----~--~----~--~----~--~----~---100 Time (msec) Figure 26. Comparison of Experimental Results With VFM Solution 52 NOTES 1 P.R. Dahl, "A Solid Friction Model," The Aerospace Corporation, El Segundo, California, TOR-158(3107-18)-1, 1968, p. 1-24. 2 D.T. Greenwood, Principles of Dynamics (New Jersey: Prentice-Hall, 1965) pp. 111-116. 3 L. Prandtl, "tlver Flussigkeitsbewegung bei schrkleuner Reibung Verhandlungen,", IIIrd Inter. Math. Kongres Heidelburg 1904, (1905), 484-491. 4 N.N. Davidenkov, "Energy Dissipation in Vibration," J. Tech. Phys., 8, No. 6, (1938) P• 483. 5 Sir Issac Newton, Mathematical Principles of Natural Philosophy, trans. A. Motte and F. Cagori (Berkeley, Univ. of California Press, 1966), p. 385. 6 N.H. Polakowski, and E.J. Ripling, Strength and Structure of Engineering Materials (New York: Prentice Hall, 1965), 8, pp. 210-213. York: 7 H. Schlichting, Boundary Layer Theory trans. J. Kesten (New McGraw-Hill 1979), pp. 90-92. 8 J. Steinheuer, "Eine exakte Ltisung der interstation~ren Couette-Stromung," Proc. Scientific Soc. of Braunschweig, (1965) pp. 154-164. 9 Polakowski, p. 215. 10 P.R. Dahl, "Solid Friction Damping of Mechanical Vibrations," AIAA Journal, 14, No. 12 (1976), 1675. 11 Polakowski, p. 215. 12 Schlichting, PP• 90-92. 13 G.G. Stokes, "On The Effect Of The Internal Friction Of Fluids On The Motion Of Pendulums," Math and Phys. Papers, (1901), 1-141. 14 Steinheuer pp. 154-164. 53 15 D.C. Bogue, and J. White, "Engineering Analysis of Non-Newtonian Fluids," AGARDograph, No. 144 (1970), p.ll. 16 Bogue p. 11. 54 BIBLIOGRAPHY Bogue, D.c. and J.L. White. "Engineering Analysis of Non-Newtonian Fluids." AGARDograph, No. 144 (1970), 1-34. Dahl, P.R. "A Solid Friction Model." The Aerospace Corporation, El Segundo, California, TOR-158(3107-18)-1, 1968. Dahl, P.R. "Measurement of Solid Friction Parameters of Ball Bearings." Incremental Motion Control Systems and Devices Symposium Proceedings, (1977), pp. 49-60. Dahl, P.R. "Solid Friction Damping of Mechanical Vibrations." Journal, 14, No. 12 (1976), pp. 1675-1682. Davidenkov, N.N. "Energy Dissipation in Vibrations." 8, No. 6 (1938), 483. AIAA J. Tech. Phys., Draper, c.s. and M. Finston. "High Accuracy Mechanical Integration by Shear in Viscous Liquids." Proceedings of National Academy of Sciences, No. 4 (1959), 528-49. Eckert, E.R.G. Introduction to the Transfer of Heat and Mass. York: McGraw-Hill, 1950. Eirich, F.R., ed. Rheology, Theory and Applications. Academic Press, 1958. New New York: Goldstein, s. ed. Modern Developments in Fluid Dynamics. Oxford: Claendon Press, 1952. Greenwood, D.T. 1965. Principles of Dynamics. Lamb, Sir Horace. Hydrodynamics. New Jersey: New York: Prentice-Hall, Dover, 1945. Lazan, B.J. Damping of Materials and Members in Structural Mechanics. London: Pergamon Press, 1968. Marris, A.W. and J.T.s. Wang, eds. "Symposium on Rheology." ASME Applied Mechanics and Fluid Engineering Conference Proceedings, New York: American Society of Mechanical Engineers, 1965. McKelvey, J.M. Polymer Processing. New York: Wiley, 1962. 55 Newton, Sir Isaac. Mathematical Principles of Natural Philosophy• Motte, A. and and F. Cagori, eds. Berkeley: Univ. of California Press, 1966. Pipes, L. A. Applied Mathematics for Engineers and Physicists. York: McGraw-Hill, 1946. New Prandtl, L. Uver Fltissigkeitsbewegung be! schrkleuner Reibung. Verhandlungen. IIIrdinter. Math. Kongres Heidelburg 1904, (1905), 484-491. Polakowski, N.H. and E.J. Ripling. Strength and Structure of Engineering Materials. New York: Prentice Hall, 1965. Schlicting, H. Boundary Layer Theory. McGraw-Hill, 1979. Sneddon, I.N. Steinheuer, J. StrBmung. 154-164. Fourier Transforms. Trans. J. Kestin. New York: New York: McGraw-Hill, 1951. "Eine exakte LBsung der instationaren Couette Proc. Scientific Soc. of Braunschweig, XVII, (1965), Stokes, G.G. On the Effect of the Internal Friction of Fluids on the Motion of Pendulums. Cambridge: Math. and Phys. Papers, (1901), 1-141. Van Vazer. J.R., J.W. Lyons, K.Y. Kum, and R.E. Colwell. Viscosity and Flow Measurement, A Laboratory Handbook of Rheology. New York: Interscience Publishers, 1963. 56 APPENDIX A TEST PROCEDURE 57 1. Assemble test hardware 2. Verify proper equipment set up 3. Obtain scale factors for torque meter and tachometer 4. Obtain time response of test hardware 5. Obtain pivot dry friction torque characteristics 6. Test specified oil using a specified viscometer gap a) Check equation for the predicted parameters b) Fill annulus with fluid c) Record data in log d) 1) date 2) run number 3) sample interval of analyzer 4) torque range selection on analyzer 5) veloctiy range selection on analyzer 6) scale factor for torque data 7) scale factor for velocity data 8) dynamometer meter range 9) command signal frequency 10) wall velocity 11) n-point filter for data; n = ? Record data on floppy disc 1) command 2) rate 3) torque 58 4) e) f) time Plot the following parameters 1) torque vs time (linear scale) 2) velocity vs time 3) torque vs position 4) torque vs time (log - linear scale) Thoroughly cleanse viscometer