## Are you sure?

This action might not be possible to undo. Are you sure you want to continue?

A Literature Review of the Numerical Simulation of Mixers, S.S.E., and T.S.E. (1994-2003) Part 1

B. Berger Emeritus Professor Department of Mechanical Engineering 2181 Glenn L. Martin Hall University of Maryland College Park, MD 20742 Telephone: 301-405-5262 Facsimile: 301-314-9477 E-Mail: berger@eng.umd.edu

December 20, 2004

D E P A R T M E N T A.

O F

M E C H A N I C A L

E N G I N E E R I N G

J A M E S C L A R K S C H O O L O F E N G I N E E R I N G

INTRODUCTION A preliminary attempt has been made here to survey the state of the art in the computer simulation of flows in T.S.E.. Twenty-two publications spanning the period 1994-2003, dealing specifically with the numerical simulation of flow in 2 and 3-D mixers, single and twin screw extruders, have been reviewed. The inclusion of mixers and S.S.E. was predicated on the observation that algorithms developed for the analysis of these simpler geometrics have often been generalized effectively for the analysis of T.S.E.. A summary, in tabular form, listing essential reference data is provided for rapid reference. A brief section of conclusions follows. Individual reviews, grouped according to algorithm, and mixer geometry are then presented. Although the review is not exhaustive, an effort has been made to include recent publications of comprehensive F.E.M., B.E.M. and D.R.B.E.M. algorithms. The encouragement of Professor D. K. Anand is appreciated. Typing of the manuscript was expertly executed by L. Crittenden.

1

SUMMARY The following summary tabulates only essential data drawn from the references, where T.S.E. = twin screw extruder, F. B. = full barrel, TEMP. = temperature calculation, F.E.M. = finite element method, G. N. = generalized Newtonian fluid, B.E.M. = boundary element method, C. F. = creeping flow, N = Newtonian fluid, G. N. S. = generalized Newtonian fluid with solids loading, F.A.N. = flow analysis network, S.B. = starved barrel, MIX = mixer, F = fluid, F.A.T. = full acceleration term, T. D. = time dependent, D.R.B.E.M. = dual reciprocity boundary element, C. D. = circular drop, P. F. = partially filled, R. C. F. = reacting creeping flow, V. E. = visco-elastic fluid. Ref. No. [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] T.S.E., F.B., 3-D, F.E.M., TEMP., G.N., C.F. T.S.E., F.B., 3-D, F.E.M., TEMP., G.N., C.F. T.S.E., F.B., 3-D, F.E.M., TEMP., G.N., C.F. T.S.E., F.B., 3-D, F.E.M.,---------, N., C.F. T.S.E., F.B., 3-D, F.E.M.,---------, N., C.F. T.S.E., F.B., 3-D, F.E.M., ------, N., C.F. T.S.E., F.B., 2-D, F.E.M., TEMP., G.N.S., C.F. T.S.E., F.B., 2-D, F.A.N., ------, G.N., C.F. T.S.E., S.B., experimental, C.F. MIX, 1F, 2-D, D.R.B.E.M., TEMP., G.N., C.F. MIX, 1F, 2-D, DRBEM, TEMP., G.N., F.A.T. T.S.E., F.B., 2-D, B.E.M., -----, N., C.F. T.S.E., F.B., 2-D, B.E.M., ------, N., C.F.

2

2-D. C. REVIEW ARTICLE REVIEW ARTICLE REVIEW ARTICLE C.E.. V.. B.B..M. C.[14] [15] [16] [17] [18] [19] [20] [21] [22] [23] MIX. F. TEMP.D.F. 1F. N.. B. MIX.M. G.....B.F. 2-D flow of viscoelastic fluids with memory.F. 2F.N.E.. T.M.E..F. B. TEMP.M.S.A. N... ------. ------. R.E.F.M.. F. S... 2F.T. MIX. C. F. ------.E..S. P.E.C. ------..E. C.E.E. F.S. 3-D. N.. 2-D.. 2-D.F.N. 3-D. F. G. F. 3 . 2-D..M.

In B.E. generalized Newtonian fluids with solids loading..E..CONCLUSIONS Conclusions drawn here are supported by the data presented in the summary. rather than B.M. Existing B.M. respectively.E. respectively. much of the 4 .M.M. solution for the flow of a chemically reacting Newtonian fluid within a full barrel T. of Newtonian and generalized Newtonian fluids.M. [12 and 15] present B.M.B.M.E. algorithms could form the basis for the calculation of 3-D. Execution times for full size T. There is a possibility that through the use of large vector or parallel processors. solution for the 2-D flow of a viscoelastic fluid within a mixer is presented in [22].E. However..E.S.M.E. full barrel. could be achieved.E.M..M..S. when the necessity of using D. References [10. (MIMD). solution for the flow of a generalized Newtonian fluid with solids loading within a full barrel T..E.E. Already implemented F.E. reasonable execution times. within T. and chemically reacting Newtonian fluids.E.M. References [7 and 21] discuss an F. Such extensions would incur increased execution times.M.E.S. An F.. of viscoelastic fluids. the boundary element method. In view of this. for these algorithms.E. Reference [20] reports on F. 2-D.E. respectively.E.E. are more efficient than F.E.R. Nonlinearities appearing in both cases require the replacement of B.M.S. Newtonian flows..E.. reference is made to information contained in individual reviews.S.M.S.E. and a 3-D simulation of a S. with the far more complex D. From this it is reasonable to infer that: 3. Existing F. and a mixer.M.E. B. and nonreacting. it appears that: 1.S.E. These results suggest that: 4.E.S. Ref. arises.E. creeping. As an alternative to F. on possibly multi processor or vector machines are difficult to estimate.S. From the summary it is seen that ref. algorithms for the creeping flow of a variety of fluid models have been demonstrated which are applicable to full scale full barrel flows in T. B. flows in a 2-D approx. 2. These simulations were limited to short segments of full barrel T. formulation. 3-D. and performed on relatively modest single processor workstations. solutions for full barrel Newtonian. In some cases.M. of a T. 11] present analysis of the 2-D mixing of a generalized Newtonian fluid with thermal effects for creeping flow and for the full acceleration term including products in the velocity.S.B. The resulting codes would reduce memory requirements and execution times. algorithms and codes could form the basis for the calculation of 3-D flows within T.E.M.M.E.. analysis of full barrel 3-D creeping flows in T.R. 20] report successful F. only domain boundaries are discretized. reacting. provides a computational algorithm which is one dimension less than the corresponding F.E. In general. [1-6.E.

M.E.E.S. 5 .E. These results suggest that: 5.S. Possible avenues of further research into the modeling of flows in the difficult case of partially filled or starved T. over F. 7. Ref. 21]. [14] presents a B.advantage of B. [23] open up the possibility of constructing codes for the calculation of flows of viscoelastic fluids with memory in T. D.B. is feasible provided execution time could be kept within reasonable limits. in the presence of a rigid no slip surface is computed through 2-D D. possibly through large scale parallel processing.B. Ref. through full barrel T. 19. It is shown in [10] that parallel processing on a 32 node connection machine reduces single processor execution times by factors of from 10-15. is lost.E. The results of Ref. Large Multiple Instruction Multiple Data computers with distributed memory would be essential for the associated computations. The boundary between the fluids is computed and tracked as a function of time. [21] gives the creeping flow of a generalized Newtonian fluid with solids loading and temperature effects in a partially filled 2-D mixer. with thermal effects and including the nonlinear acceleration term.M.E. solution for the combined creeping flow of 2 distinct Newtonian fluids in a 2-D mixer. [14. solutions for the flow of generalized Newtonian fluids. From the foregoing it appears that: 6.R. [19].S.R. The motion of an initially circular 2-D region of Newtonian fluid surrounded by a different Newtonian fluid. are suggested by the approaches of Ref.M. in Ref.E.E. based on the Newtonian fluid solution.E.

analysis.785 degrees of freedom in the temp.p.6GHz are available giving a ratio of 7. and the material was probably high density polyethylene. except that η = η 0 H (T ) /( H (λγ&H (T ))1−n ) . An upper limit of 500 MHz was estimated by H. Exact specifications for the HP-J2240 were not available from H. Improvements in memory access and internal information transfer probably boost this ratio to 10:1.S. Modeled a TSE over 10 kneading discs for a total length of 9. The barrel diameter is 100mm.E. Full barrel of D=30mm.. ρC p v ⋅ ∇T = k∇ 2T + τ : ∇v . the field and constitutive equations are ∇ ⋅ v = 0. F = η 0 [1 + λ2 c ( 2 II d )] The solution is quasi-static.E. FEM similar to [1]. was modeled. The material is high density polyethylene. ( n −1) / 2 .2:1. τ = 2ηD . no-slip.. 3-D.25m. Only the rotor region of the second mixing stage was analyzed over a length of 0. There were 20.561 nodes with 60.P.011 degrees of freedom in the flow analysis and 41. Computer 6 . Full barrel. Computer used was an HP-J2240 without parallel processing.p. quasi-static model. In all three cases 30mm of the T.Summary of Reference 1 In [1]. max.3cm.P.470 nodes and 65. Desktop systems operating at 3. Cross model. Upwind Petrov-Galerkin applied to the temperature field. full barrel. rotational speed = 420 r. Summary of Reference 3 Field and constitutive equations are identical to those of [1. The residence time distribution. 3-D. and several mixing indices were studied. There were 42. no-slip.843 degrees of freedom. FEM identical to [1]. Velocity distributions are computed.m.m. with 480MB of memory.S. Carreau’s model.281 degrees of freedom in the temperature field. − ∇p + ∇ ⋅τ = 0 . The measured temperature 200°C < T < 290°C. Execution time for each rotor position would then ≅ 9 min. The rotational speed = 60 r.2]. max. rotational speed = 400 r. Experimental verification of predicted pressure and temperature profiles is good. = 200°C. barrel temp. Mesh dimensions and computer execution times are not given. η = H (T ) F (γ&H (T )) . nearest distance for each marker.m. 6930 markers were defined in the simulation and followed in the evolution of the flow.p. FEM Galerkin method applied to the flow field. Summary of Reference 2 Field and constitutive equations are identical to those of [1]. right-handed full flight screw and a right-handed full flight screw for a T. The modeled TSE is 2-stage. γ& = 2 II D . and H (T ) = exp[− β (T − T0 )] . 3-D. Studied a right-handed screw with lefthanded slots. In the flow field there were 22. Execution time for each rotor position = 90 min.

mixing coefficients. The pressure is found using a penalty function as follows: p = −λ (∇ ⋅ v) λ (∇ ⋅ v) + ∇ ⋅ (η [(∇v) + (∇v)T ]) = 0 λ = 108 where λ ≡ penalty number. t ). for the purposes of estimation. The motion of 1778 massless marker particles was studied.S.6 min.6GHz. The DS20E may be configured to run at 500.M.E. assume 667MHz. which is probably closer to 7:1 considering improvements in memory access and internal information transfer.M. method is employed with a 27 node tri-quadratic brick element for the components of the velocity vector. for convergence after 5 iterations of each rotor position.memory employed = 900MB and execution time = 60 min. 667.m. Pressure. The model is quasi-static with a no-slip boundary condition.E. is considered with pairs of two-tipped kneading discs staggered forward at various angles. Then on a current machine the solution for 3cm of a T. area stretch. method with a penalty function is used in the discretization with p = − βη (∇ ⋅ v) and β = 10 6 7 . The speed ratio of the new to the old is 5. Execution times and storage requirements were not provided. Summary of Reference 5 The fluid is modeled as an incompressible Newtonian fluid undergoing a 3-D isothermal creeping flow. Summary of Reference 4 Field and constitutive equations are identical to those of [3] except that the flow is assumed to be isothermal. 3 The Galerkin F. values of the stress magnitude were found for the three screw geometries. The Galerkin F.E.S. the analysis is quasi-static and non-isothermal. A total of 6930 markers were followed and the nearest distance between markers. Computed results are in reasonable agreement with experimental results.E. As in [1. Since the authors did not specify a configuration.4:1. A mesh composed of 1344 elements with 13. Field equations are: ∇⋅v = 0 ∇p = ∇ 2ηv v = v( xi . average residence time and concentration of markers in the z-direction computed.. i = 1. mixing element would require ≅ 8. Current desktop systems run at 3. Analyzed a set of 5 kneading discs each of width = 10mm. A 3-D T. rotational speed = 60 r. The material studied was polypropylene. per iteration on a Compaq Alpha Station DS20E. 2]. max.p.543 degrees of freedom was used which introduced an error of less than 4% as compared with an analysis based on a mesh of 2352 elements. or 833 MHz. The material is a low density polyethylene melt at 180°C. Max. temperature.

S. Summary of Reference 7 The analysis is based on a 2-D model of a T. residence times. Summary of Reference 6 This is a generalization of [5]. per iteration on three parallel processors. 12.46” lead screw with a helix angle of 16. A 3.432 nodes.861 degrees of freedom.The F. Lyapunov exponents. 2 8 . there are 25. Execution time was 73 min.600 brick elements.25 is modeled.S.287 nodes. model consisted of 1280 tri-linear eight noded brick elements with a total of 1347 nodes at each axial section. Several iterations were required for convergence. Chaotic dynamics are indicated in the T.E. j + v j .. Poincaré sections and tracer paths are found. For a 30° stagger angle. The 3-D mesh had 10.E. there is good agreement between the 3-D and 2-D models. Computer memory requirements and execution times are not provided. The energy equation in dimensionless form is 2 Pe (v ⋅ ∇θ ) = ∇ 2θ + Gη (| γ& |) where θ = (T − T0) /(Tb − T0 ) .i ) . γ& = II d . For screw geometrics in which the flight width is small in comparison to the channel width of the screw element. and 84. The solution is quasi-static with a no-slip boundary condition and a full barrel. The field equations are ∇⋅v = 0 ∇p = η∇ 2 v Again the penalty function is p = λ pη (∇ ⋅ v ) The no-slip boundary condition is assumed. A Bingham fluid was used for which τ y (1 − exp(− nb | γ& |)) ⎡ ⎤ η = ⎢m | γ& | n −1 + exp(− c 1 (T − T0 ))⎥ | γ& | ⎣ ⎦ T0 = entrance temp. 37. isothermal and the fluid incompressible. The 2-D versions of the 3-D field equations of [5.E. γ& is IId of the rate of deformation tensor.920 elements.. 28. with two-tipped kneading discs. The flow is quasi-static. G = Griffith member. d ij = 1 (v i. 6] are augmented with a solution of the energy equation based on the SUPG technique (Streamlined Upwind/Petrov Galerkin technique). Tb = barrel temp. d = deformation rate tensor.296 simultaneous equations and a global stiffness matrix with (22)·106 non-zero entries.

The flow field is described by a cylindrical coordinate system (θ. z. v ≡ fluid velocity.E. For an incompressible linear Stokesian fluid the constitutive is t kl = − pδ kl + 2µ d kl . This method is applied to a generalized-Newtonian flow in a composite modular machine.i i = 0 9 . the suspension well mixed and properly de-aerated. The material modeled was poly dimethyl siloxane filled with borosilicate hollow glass spheres. The 2-D F. The velocity vector is expressed as v = vθ (r )eθ + v Z (r )e z + v r er where v r = 0. The equation of continuity ∂p + v ⋅ ∇p + ρ ∇ ⋅ v = 0 ∂t with ρ = constant becomes ∇⋅v = 0 or v .S. Here the barrel is full. ⎞ α ⋅ (v − v s ) = β 1⎛ ⎜ nα : T ⎟ n ( ) Summary of Reference 8 The authors have generalized their Flow Analysis Network (FAN) method of analysis of flow in intermeshing counter-rotating T.The constitutive model is the generalized Newtonian fluid t ij = −ηII d d ij .E.E. devolatilization and kneading sections of the T.S. e z and er . The wall slip condition is expressed as: n ⋅ (v − v s ) = 0 ⎠ ⎝ Where v s is the solid boundary velocity. The solid loading level was 60% by volume. η is modeled according to the Herschel-Bulkley relationship. τ y (1 − exp(− nb | γ& |)) ⎞ ⎛ n −1 1 ⎟ & η =⎜ γ m | | + ⎜ o ⎟ exp(− c (T − T0 )) & γ | | ⎝ ⎠ The filled polymer is assumed to be a purely viscous fluid. Computer storage and execution times were not provided. β 1 = m0 w n −1 / Rs β is the Navier’s slip coefficient. Then v(r . The numerical modeling and analysis were applied to the regular flighted screw section of the extruder preceding the die. r) with unit base vectors eθ . The wall slip simulation was essential in bringing the computed and experimental wall pressure into agreement. the rheological behavior of which is described by the generalized Newtonian fluid.θ ) . e z is a constant vector while eθ is a function of θ . which includes positive displacement effects in the pumping action. model was not capable of predicting flows in the mixing.

1 = 3 − ⎨ ⎬v l ∂x1 ⎩ 31⎭ and ∂v ⎧ l ⎫ v1. then the equation of motion is t k l.1 ) 2 then ∂v ⎧ l ⎫ v 3.Assuming that the body forces vanish and the Dvi / Dt ≅ 0 .k = 0 Substituting the constitutive and continuity equations into the equation of motion gives k − p. follows from the constitutive equation 1 t13 = 2µ d13 = 2µ (v1.3 = 1 − ⎨ ⎬v l = 0 ∂x3 ⎩13 ⎭ and ∂v t13 = µ 3 ∂x1 ( 3) 1) In terms of physical components t ((3 ) and v 1) t ((3 g11 g 33 = t13 ) = t13 / and v (l) v (1) v ( 2) v ( 3) v 3 = ∑ g 3l = g 31 ⋅ + g 32 + g 33 = v ( 3) l g ll g11 g 22 g 33 Then ∂v ( 3) ∂x1 Or in the notation of the paper ∂v σ zr = η z ∂r Similarly ∂ ⎛v ⎞ σ θr = ηr ⎜ θ ⎟ ∂r ⎝ r ⎠ The equations of motion may be expressed as: 1 ∂ 2 1 ∂p ( =0 r σ θr ) − 2 r ∂θ r ∂r t (1) ( 3) = µ 1 ∂ (rσ zr ) − ∂p = 0 r ∂r ∂z − ∂p =0 ∂r 10 .l + µv l .k = 0 The shear stress.3 + v 3. t13 .

Plots of the length of the pellet bed as a function of turns of the screw thread are given for various flow rates and a screw speed of 200 rpm. These solutions are isothermal. two different regions of pellet melting were identified. Ts = temperature of the solid . may be modeled in terms of a power law ⎡⎛ ∂ ⎛ v θ η = K ⎢⎜ ⎜ r ∂r ⎜ r ⎝ ⎢ ⎣⎝ 2 2 ⎞ ⎞ ⎛ ∂v Z ⎞ ⎤ ⎟⎟ ⎟ ⎥ ⎟ +⎜ ⎠ ⎠ ⎝ ∂r ⎠ ⎥ ⎦ ( n −1) / 2 The regions between the rotors and barrel wall and between the rotors are modeled separately. the polar angle θ. The equations of motion were solved for vθ and vz. At high shear rates. Solutions for the Newtonian and non-Newtonian full barrel cases are presented and compared. Continuity was satisfied approximately with flux balances. Based on the experimental observations. Fill factors are found for starved screw sections. No experimental verification is provided. for several screw configurations is plotted. assuming a starved flow. λ = heat of fusion. dL/dt = rate of melting. one in the calendar gap and the second in the C-chamber. Gp =mass flow rate of the pellets. 11 . Modeling melting in the calender gap resulted in the relationship dL ρ mQ p = dt ρ sWH ⎤ ⎡ σ zx (2πRN )WRθ − 1⎥ ⎢ ⎢ G p (cs (Tm − Ts )) + λ ⎥ ⎦ ⎣ where Qp = π ( R1 + R2 ) NW∆ and cs = specific heat of the solid.These equations must be satisfied together with the equation of continuity.S.E. Neither is there any discussion of computer storage or computational times. The above iso-thermal solution is modified for the non-isothermal case and temperature profiles given as a function of z. A similar equation is derived for melting in the C-chamber. However. Tm= melting point. Summary of Reference 9 The model is based on an experimental program with the melting of plastic pellets in T. A screw pull out technique was used to study polymer pellet melting behavior along the screw axis. Pressure as a function of the axial variable z. η. The length of the pellet bed is found for a specific set of initial conditions. A discrepancy of about 25% is noted between the experimental and computed results. experimental results contained in a 1996 dissertation are cited.

creeping flow assumption.Summary of Reference 10 A 2-D solution for thermally linear and non-linear mixing flows is formulated utilizing the BEM and DRBEM methods respectively.αα − µγ&αβ γ&βα 2 1 (Uα . Evaluating the line integrals using Gaussian quadrature yields k CiU ki + ∑ ∑ tα j =1 m =1 N NG [( ) ijm (U α ) j ] − ∑ ∑ [(U αk )ijm (tα ) j ] = 0 N NG j =1 m =1 which may be reduced to a system of linear algebraic equations.α ).α + µ Uα . and heat flux vectors respectively. A heat transfer formulation is developed by including the energy equation 1 ρC v (T j + U α T. A DRBEM discretized boundary integral equation is 2 given which is recast in matrix form as where T ≡ temp.α = 0 P. inside. ∇ ⋅ U = 0 or U α . The field and constitutive equations are.5 and 0.β + U β . Ci ≡ 1.α ) = kT. Utilizing Green-Gauss transformations. 0. in rectangular Cartesian coordinates. [ A]{U } − [ B]{T } = 0 Parallel generation of [A] and [B] is possible. these equations may be expressed in integral equation form as α α )d Γ CiU ki + ∫ (t k U α − tα U k Γ where Γ ≡ boundary curve. respectively.0 for point on. ββ = 0 where the fluid is assumed to be incompressible and Newtonian and inertial terms are omitted. and outside the boundary α α and U k are Green’s functions and tα are components of the traction vector. t k The boundary discretized form of the foregoing integral equation is CiU ki + ∑ U α j j =1 N ∫ Γj k k tα d Γ − ∑ tα j ∫ U α dΓ=0 j =1 N Γj where N = number of boundary elements.and γ&αβ = ˆ ] − [ H ][T ˆ ])[ F ] −1{b} [ H ]{T } − [G ]{q} = ([G ][Q where {T} and {q} are the unknown temp. Computer: -32 node connection machine model CM-5 system 12 .

k ) and v k .β + U β . rotor with a single axis of symmetry. λ ≥ 0. single rotor. b. Constitutive model for sheer stress: τ αβ = η (γ&. Omitting product terms gives t k l = − pδ k l + α1d k l where d kl = 1 / 2( v k .No slip on the barrel and rotor. and tflow = characteristic time of the flow field. In a more familiar notation t k l = − pδ k l + α1d k l + α 2 d k m d m l for a non-linear incompressible fluid where α i = α i ( II d . Total running times.5 min. It is observed that Newtonian and generalized Newtonian fluids do not exhibit normal stress effects. Boundary Conditions: . circular barrel. Serial ≅ 110 min.Newtonian fluid with conductive and convective transport effects. of the rotor and barrel = 180°C. De = λ / t flow where λ = relaxation time of the fluid. III d ). Memory Effects: The Deborah number. T )γ&αβ where γ&αβ = 1 / 2(U α . Discretization: -379 boundary elements 209 internal solution evaluation points within the fluid Results for BEM Flow (Linear). surface temp. a.Program Language: CM-Fortran Mixer Geometry: -2-D. Parallel ≅ 1. Parallel ≅ 8. Material Properties: . a constitutive quantity.l = ∂v k ⎧ n ⎫ − ⎨kl ⎬ v n ∂xl ⎩ ⎭ Normal stress effects are discussed. Results for DRBEM Flow (Non-linear). Total running times a.l + v l. b. 13 .α ) . De.3 min. 2. Summary of Reference 11 The dual reciprocity boundary element method (DRBEM) is discussed and applied to the 2-D mixing of non-linear polymer rheological fluids. Serial ≅ 15 min. rotating about an axis through the center of the circular barrel and ⊥ to plane of the mixer. i = 1.

t + U β U α . β = bα .S.β + S ⎣ ∂T ⎦ σ αβ .De → 0 ⇒ memory effects are negligible De → ∞ ⇒ memory effects are negligible solid like behavior De ≅ 1 ⇒ memory effects are observed Extrudate Swell: Polymer exits the die and expands to several times the diameter of the die.α − T ⎢ ⎥ U α . Equation of Motion.α ) = −qα .β ) j = ( f j )α Then k k C lU kl + ∫ Γ (tα U α − tα U α )d Γ = ∫ Ω (∑ β j (σˆ αβ . Continuity. 636.ρ ) j )U αk d Ω The dual reciprocity method transforms the above into 14 . BEM modeling of flow within a S. Field Equations. t +U α T .α + τ αβ U α .β = bα .β ) Then the integral equations l k k k C lU α + ∫ tα U α d Γ − ∫ tα U α d Γ = ∫ bα U α dΩ Γ Γ Ω ˆ αβ k . bα = − ρgα + ρ (U α .α = 0 ρ C v (T . ** The 3-D.β ) Note the retention of the acceleration terms in the equation of motion. DRBEM Formulation Let ⎡ ∂P ⎤ σ αβ .β = − ρ gα + ρ (U α . Energy Balance U α . was accomplished by Biswas et al. Attributed to normal stresses and memory. ˆα k be a particular solution of σ Let σ β Let ⎛ 2( N + L ) ⎞ bα = ⎜ ∑ β j f j ⎟ ⎝ j =1 ⎠α and k ˆ αβ (σ .t + U β U α .E. Antec. in 1994.

simulation of the flow of a multiple phase. which is assumed to be constant for a given fluid. and volumetric strain rate were computed. double. The exterior boundary corresponded to the barrel of the T. incompressible. Uα = velocity in the xα coordinate direction. 2. multi-viscous polymer in an internal single rotor batch mixer. ββ = 0 where α.E.5 min. strain rate.S. Summary of Reference 12 The 2-D BEM is used to analyze and compare the mixing effectiveness of T. Memory requirements and execution times are not given. and triple flighted screws were studied. β = 1. Summary of Reference 14 Reference [14] contains a 2-D. The effect of varying centerline distances between screws on mixing effectiveness was studied (random distribution of traces). p = pressure and µ = viscosity. 90° out of phase.E. iso-thermal.S. of various geometries. 15 .S. Summary of Reference 13 This is a slight modification of Reference 12 with a more extensive set of references. Geometry: The 2-D fluid motion in a plan perpendicular to the axes of a 3-D T. Single. were not included in the solution.α = 0 and − p.k k l ˆk ClU α α ∫ Γ tα U α − tα U α d Γ = ∑ β j C l U j =1 ( ) 2( N + L) ( ( ) + ∫ (t (Uˆ ) − (tˆ ) U ) d Γ) j Γ α k α j k α k j α k Introducing boundary conditions yields: ˆ ] − [G ][tˆ])[ F ] −1{b} [ H ]{U } − [G ]{t} = ([ H ][U A solution for counter rotation of concentric circular cylinders with a gravitational body force is provided.E. A creeping flow is assumed. rotate about the screw axes of the T. was modeled.α + µU α .E. One assumes that the acceleration terms in the equation of motion.E. The field equations are: Uα .S. Computer: DEC 5000/200 Execution Time: 7. Two symmetric elliptic impellers. Time Step = 2° rotational step Fluid properties = Newtonian fluid. B. creeping flow. linear and non-linear. The travel or deformation of a tracer line full barrel assumption. for each time step.

A three domain simulation of flow between parallel plates is also given. The deformation of a small circular drop of viscosity µ 2 surrounded by a fluid of viscosity µ1 in a 2-D single rotor mixer is presented. Uα is given Γ2. k r . Γ1.No slip condition. respectively.The traction vector tα is given. or outside. These appear on the diagonal of [A] and are linked by a matrix [C]. Constant boundary elements were assumed where the values of Uα and tα are assumed to be constant on each element and equal to their values at the midpoint. of the specific domain. No information regarding computer execution times or memory size requirements is given. Solid boundary:. 0. Integral Equation Formulation k k CiU ki = ∫ [tα U α − tα U α ]d Γ = 0 Γ Ci = 1. and placing point i on each element center yields a matrix equation [A] {X} = {R} where {X} is a vector of unknowns Uα and tα. Interfacial boundary between regions 1 & 2.α r . Γ3. b. [A] is composed of submatrices associated with each domain. 1 2 tα = −tα on Γ3 1 2 Uα = Uα on Γ3 Superscripts refer to regions 1 or 2. In general [Ai] is composed of coefficients of the unknown parameters of the i-th domain. The discretized form is then α α CiU ki + ∑ U α j ∫ Γ j t k d Γ − ∑ tα j ∫ Γ j U k dΓ=0 j =1 j =1 NP NP Evaluating the integrals. 0 inside. 16 .ν r −1 / π Discretized Form CiU ki + ∑ NP j =1 (∫ α Γj t k U α d Γ − ∑ ) (∫ NP j =1 Γj α α t Uk dΓ ) where NP = number of boundary elements. c.α r. k 4πµ [ ] and k k tα = σ αβ ν β = − r .5. Free boundary :. [Ai].Boundary conditions: a. Here k Uα = −1 k δ α lnr − r. on. utilizing Gaussian quadrature. It appears that the new position of a domain is estimated from the previous position through a Uα ⋅ ∆ t calculation.

Computational requirements: 1000-1400 quadratic shell elements. High stresses are required.82mm. creeping flow. The barrel diameter and mixing section length are not given. Pure rotational flow. A number of mixing sections were considered. respectively. No free surfaces. 200 tracer particles.Summary of Reference 15 Reference [15] presents a 3-D computer simulation of the flow in a S. λ =1. Shear flow. Full barrel flow. where γ Flow number: The flow number = λ is defined as λ = γ& /(γ& + w) where γ& and w are the magnitude of the strain rate (does the author mean deformation rate?) 1 and vorticity tensors. single screw extruder. Mixer geometry: 3-D. The vorticity tensor.S. Pressure and Torque Analysis Torque defines the size of the motor and screw of the extruder and is important in calculating the efficiency of the mixer. The magnitude of d( n ) may be what is defined by γ a.5 Dispersive mixing: Breaking of the secondary flow into smaller particles. Allowing for improvements in memory access and internal information transfer would probably increase this ratio to 20:1. quadratic shell elements. n is defined as 1 D d ( n) = (ds ) = d kl nk nl ds DT &. circular barrel. Distributive mixing: Homogenization of a secondary phase within the fluid matrix and is dependent on the amount of strain imposed on the material. Then the magnitude of the vorticity tensor could be defined as the magnitude of the vorticity vector. incompressible. screw diameter. In general.0 b. I assume that this timing was for a converged solution and not a single step. 1D-4D. Material properties: Newtonian fluid. and 60 min. BEM. the lower the torque and required pressure for a given flow rate the more efficient the extruder will be.55-21.0 c. pitch. 6. Computational algorithm: 3-D.l − v l .l . Boundary conditions: No slip on all surfaces.6 GHz are currently available. respectively. Total strain measure is γ& (t1 ) = ∫ γ& (t )dt 0 t1 & (t ) is the magnitude of the strain rate. Desktop systems operating at 3. with a rhomboidal mixing section. 4-6. Pure elongational flow. λ = 0. giving a ratio of ≅ 18:1.E. constant viscosity. Then execution time would be ≅ 12 min. number of cuts.k ) which can be 2 expressed as an axial vector wk such that wk = eklm v m . λ = 0. which is dependent on the rate of strain. execution time on a Silicon Graphics Indigo 2 R1000 was 4 hours to solve for unknowns. Tracer tracking execution time was ≅ 20 hours. Computer specifications: The Silicon Graphics Indigo 2 R1000 operates at 195 MHz. 17 . 4200 nodes. The stretching in the direction of a unit vector. wkl = ( v k .

18 . ρ= density. 65-68. Internal Mixer: 2-D analysis. k k Case Studies: a. The 2D-2D section was shown to be the best distributed mixer in agreement with experiment. All of this material has been treated in greater detail in references [10-15]. respectively. with particle tracing. [16] is a review article which discusses FEM and. Rhomboidal Mixing Section: See reference [15]. Pipe flow: BEM analysis of pipe flow. τ = deviatoric stress tensor. Cp = specific heat. Summary of Reference 16 Ref. ∇⋅v = 0 ∇ ⋅ τ + ∇p = 0 DT & = k∇ 2T + η (γ& ⋅ γ& ) + Q ρC p dt Constitutive law: τ = η γ& & = where v = velocity vector. Field equations: Assume creeping incompressible flow. n = a unit normal to the domain. Ω . with much greater emphasis. c. Mixing Section for an Extruder: Example of design optimization utilizing BEM. comparison between FEM and BEM element distribution. BEM for optimization of the rhomboidal mixing section is evident. including mixing. pp. e. 2. & = other heat source input. integral equation is k k k ci v i + ∫ (τ α v α + v α τ α ) d Γ = 0 Γ where τ α are surface tractions. T = temperature. d. and comparison to known closed form solutions. BEM algorithms for flow calculations in plastics processing. The BEM. b. k = thermal conductivity. p = pressure. residence time calculation. Q Boundary Element Method Divergence Theorem forms a basis for the BEM method. vα and τ α are fundamental solutions for velocity and tractions.Conclusions Usefulness of 3-D. ∫ ∇ ⋅U d Ω = ∫ U ⋅ n d Γ Ω Γ where Γ = a boundary surrounding a domain. This article is a translation of Kunststoffe 89 (1992). η = viscosity. Static Mixer: CAD generation of boundary elements. FEM network becomes overly complex. γ rate of deformation tensor.

Applications of the finite volume flow method are discussed. He focuses on problems associated with moving boundaries. mixing. Review of Reference 18 This is another review article. The Hele-Shaw model and FAN analysis are discussed. A brief discussion of the Hele-Shaw model is included. Kunststoffe 10/2003. and extrusion dies. See references [10-18]. Describing Geometrics with Models A brief history of modeling and simulation in plastics processing beginning in the 1950s with the work of Bird. See references [1-7]. FDM History of FDM applications in injection moulding. The fundamental physical laws required in processing methods in plastics technology are mentioned. 3-D. Process Simulation Plays a role in the design of moulds for injection moulding. Stewart. 130-134. Finite Element Analysis A history of applications of FEM in plastics technology with a paragraph on 2 and 3-D simulations of mixing processes. die flow. FEM or BEM analysis. which contains all of the figures.E. Flow models based on simplification of the basic field equations and or geometry are compared to models based on FDM. Introduction: The author gives general observations regarding 3-D simulations. free and solid boundaries. compression moulding. and partially filled systems should be regarded as one of the grand challenge problems in polymer processing.S. compression moulding. A hierarchy of computational complexity in the numerical modeling of single and twin screw extruders is presented. including mixing. All of the material on BEM in mixing has been treated in much greater detail in references [10-15].” 19 . The Boundary Element Method A review of the theory and application of BEM to TSE. pp. FAN has been applied to flows in TSE.Review of Reference 17 This comprehensive review article is a translation of “Verarbeitun gsprozesse mit Modellen beschreiben”. each section of which will be summarized in turn. and onedimensional approximation. making the distinction between moving. See references [8. process including melting. Included is a discussion of FEM with much greater emphasis on BEM algorithms for flow calculations in plastics processing. 1. melt conveying. The Finite Difference Method. References must be obtained directly from the author online. 9]. “The complete simulation of the T. and Lightfoot.

m X L . Bingham fluid – See the Summary of Reference [6] d. The author considers generalized Newtonian fluids. j + v j .2. K xl .k + ρh = 0 ∂θ 2 where θ = temp. L Dt The material derivative of the Eulerian strain tensor is: D (ekl ) = d kl − v m. ∇θ . GNF. It is clear that the constitutive equations of a general incompressible fluid is expressed in terms of dkl . Again. θ ) ∈= Ψ (θ ) − θ ∂Ψ ∂θ η = −∂Ψ / ∂θ ∂2Ψ & ρ oθ θ + f kl d kl + q k . Carreau’s model – See the Summary of Reference [1] b. Herschel-Bulkley model – See the Summary of Reference [7] e. f and g satisfy various inequalities.l E KL X K . η = entropy. etc.l − v m . which is linear in the stress tensor and of the form t kl = − pδ kl + 2µd kl where µ = µ (θ .θ ) q k = g k (d ij .k X L . ∈ = internal energy density. Rheology: Caution: The author appears to conflate shear strain rate and deformation rate. Cross’s model – See the Summary of Reference [2] c. ∇θ . the rate of deformation tensor rather than any of the several strain rates. d ij ) . Models are discussed in which µ is modeled in various ways included in the following: a.i ) 2 The material derivative of the Lagrangian strain tensor is: D ( E KL ) = d lk x k . The deformation rate tensor is defined as: 1 d ij = ( v i .m Dt For infinitesimal deformations: D ( E KL ) ≅ d kl δ kK δ lL Dt The constitutive equations for incompressible fluids are of the form: t kl = − pδ kl + f kl (d ij . tkl is given as a function of dkl..k E KL X K . Power Law – See the Summary of Reference [8] 20 .

immersed in a viscous fluid with different physical properties and moving towards a rigid wall.. Summary of Reference 19 This paper describes a boundary element formulation for the deformation of a viscous. Not suited to complex geometries. Ω1 is bounded below by a rigid wall at x2=0 and apparently unbounded in x1 and for x2>0. 11.2 Finite Element Method: Capable of modeling much more complex geometries than F. 21 . [] denotes the jump across the boundary. jj − p. from from Ω1 into Ω2. Modeling: A general discussion of numerical modeling including F. Green’s functions or fundamental solutions for the relevant equations are required. and σ ij = − pδ ij + µ l (u i . j + u j .D. µlui . Γ. of viscosity µ1 and density ρ1. 20. At the boundary. 16.M. circular drop. Upwinding techniques reduce instabilities for non-linear problems. 2-D.i ) .2 for the external fluid and the interior of the drop.E. respectively. Boundary Conditions: No slip boundary condition on the wall x 2 = 0. β = ρ 2 / ρ1 .E.M. Particle tracking is facilitated by the method. See Summaries of References [10. Field Equations: Incompressible Navier-Stokes in 2-D..D. This leads to the reduction of an N-dimensional problem to one of dimension N-1. 3. 21. Boundary Element Method: Only boundaries. 12.1 Finite Difference Method: Lends itself to modeling nonlinear problems. and B. 4. F. of the bubble [ui ] = 0 and σ ijη j = φρ1 (β − 1)g i + γkni [ ] where η is the outward unit vector normal to Γ. i = ρl (ui . Difficult to apply for problems with moving solid boundaries. 3. See Summaries of References [1. and φ = 0 or 1 . 7. Geometry: An initially circular. k is the curvature of Γ. and numerical results. − ∞ < x1 < +∞ . However. 19. surfaces of physical geometry are discretized.t + u j ui . relatively simple programming and fairly efficient algorithmic. valid for Re ≤ 200 are presented. 3. 2. Particularly effective for linear problems.3. 13. 5. Surface tension effects are included. 22] 3. 6. Γ. 17]. liquid drop Ω1 . j ) where l =1. 14. gi are the components of a uniform gravitational field.M. 15.3.M. 2-D.

Formulation The integral representation formula for the exterior velocity is ui ( x0 ) − ∫ Γ k ij ( x. non-linear partial differential equations in u1. x0 )d ij ( x)dΩ An equation of similar form applies to the interior of the fluid drop.E. H. ui .M. x0 )(u j ( x)1 )dΓ = − 1 / µ1 ∫ Γ uij ( x. through the dual reciprocity approximation. x0 )(σ jk (u ( x)))1 nk ( x)dΓ + 1 / µ1 ∫ Ω1 uij ( x.i ) where µ = constant. where d il = ρ l ( µ i . G.t + u j ui . and p. ∫ ∫ Ω1 ( )d Ω is an integral over a 2-D domain which is approximated by a boundary Γ ( ) dΓ . A first order explicit time integration is employed in which ∂ui ≅ uim+1 − u im / dt ∂t ( ) 22 . and T are matrices of known functions.Continuity for an incompressible fluid. Constitutive Equations: σ ij = − pδ ij + µl (ui .i = 0 There are three coupled. u2. α is an unknown. 1 ⎡1 ⎤ u − a1 Hu + a2 Gt = a4 ⎢ U − a3 HU − GT ⎥α 2 ⎣2 ⎦ where u and t are the velocities and tractions. B. Assuming that ρ1 and ρ 2 are constant and transforming the domain integrals in the previous integral representations yields in matrix notation. j ) Note that integral. U. j + u j .

T = temp. k∇ 2T − τ ⋅ ∇v = ρc p ( v ⋅ ∇T ) Material Properties Constitutive Equation: τ = −ηγ& where ln(η ) = ln K (T . b. for 100 time steps.E. melt density = 750 kg/m3. Γ. was discretized by 40 nodes. channel depth = 3. from an elliptical shape with no initial velocity or gravitational field. M )) ln γ&0 .8.7mm.000. Integrations employed 20 Gaussian integration points. starting from a circular shape with zero velocity. due to surface tension. = 33. Typical computational times were of the order of 1. entering peroxide distribution and pressure to drag flow ratio on the mixing of steady.S. M ) − (1 − n(T . The drop boundary.E.7mm.Numerical Examples Numerical results are provided for several examples. melt thermal conductivity = 0. no. incompressible. screw centerline separation. Screw o. n-a molecular weight = 43.. M )) ln γ&0 . pitch = 20mm. Motion of a drop towards a rigid wall. channel aspect ratio = 2.d.185 w/m°C. Summary of Reference 20 The authors have applied 3-D F. Computational Requirements: A parallel processor with two 2. in a gravitational field.428kJ/(kg°C).0 GB of RAM were employed.M. Mixer Geometry: The 3-D flow in a single fully filled conveying element channel of a TSE is modeled. M ) − (1 − n(T .420. including: a. Specific heat capacity = 2.0 min. Field Equations: Conservation of mass assuming incompressibility: ∇⋅v = 0 Momentum equation for steady flow: ∇ ⋅τ + ∇P = −∇ ⋅ ( ρ vv) Energy assuming that thermodynamic properties are constant. non-isothermal reactive flows in a forward conveying element of a T. entering w-a molecular weight = 330. Recovery of circular drop shape. 23 .0 GHz Pentium III processors and 1. M = average molecular weight. γ& > γ&0 where γ&0 = 10s-1. γ& < γ&0 ln(η ) = ln K (T . of screw flights = 2. modeling to the study of the effects of screw speed.

barrel wall temp.15) ln K = −0. For the peroxide weight fraction. 5di-tert-butylperoxyhexane. an additional dynamic viscosity relaxation scheme was used to stabilize the solution over the first 20 iterations. 5 dimethyl-2.Reactive Processes Described by species mass conservation equations (no molecular diffusion) for the peroxide as well as for the three moments of the polymer MWD.6294 ln M − 0. 24 .1364(lnM ) 2 n-1 is given as a function of T. 2 Operating Conditions Inlet temp. Cross-sectional area in the translation region = 22. Computer Requirements FIDAP version 7. Number of nodes = 12625 and 15121 and number of elements = 10560 and 12740 for 100 and 250 rpm. Computer Simulation: A mixed discontinuous pressure formulation was used in which one pressure DOF was solved for per brick element. with an element based dynamic relaxation scheme for all DOFs. & = −k I I D & Q = 2 fk I ((Q − 3Q ) /(Q − Q ) 0 D 1 0 1 0 & = −2 fk I ⋅ 2Q /(Q − Q ) Q D 1 0 1 0 & Q2 = −2 fk D I (−Q3 / 3 + Q1 / 3 − 2Q0 ) /(Q1 − Q0 ) Q3 = 2Q2 (2Q2 Q0 − Q1 ) /(Q1Q0 ) The average molecular weight of the polymer is given by M = M 0 Q2 / Q1 The peroxide decomposition rate constant is ln k D = (ln (1 ⋅ 98 ⋅ 1012 )) ⋅ (−14947 /(T + 273.0mm2. 250rpm. screw speed = 100. pressure to drage flow ratio = 0. average inlet peroxide level = 0. = 240°C. 0. For the reactive extrusion simulations.62 was run on an IBM Risc/6000 workstation with 128MB RAM.c. -0.2mm. were used for the screw and barrel surfaces. A fixed streamline upwinding factor was used for temp.25. zero flux b.6% wt. respectively.0243T + 0. Boundary Conditions: No slip condition on the barrel and screw surfaces. Axial length = 61. Execution times were not provided. Peroxide data used for 2. = 220°C.

non-isothermal. j − p.t + v j v i . Incompressible phase v i . j ) = K (T. d ij = ( v i . 2 b. jj ) + ηγ& 2 e. j + v j . peroxide conversion profiles.5Φ where Φ = effective volume fraction of carbon black. Field Equations: Continuity Equation: a. internal mixer with 2 rotating long blades and partially filled. Energy equation ρc(T.Results Sixteen cases were considered for which average temp. 25 . pressure profiles. Mixer Geometry: 2-D.t + ( ρ v i ) .t + v j T.i = ρ ( v i . flow efficiencies. generalized Newtonian flow and mixing of a polymer-filler compound is studied in a free surface regime consisting of incompressible and compressible parts. internal mixer utilizing intermeshing blades with very long tips. Compressible phase ρ . Summary of Reference 21 The authors model a partially filled.i = 0. Carbon black concentration equation C . j ) d.i = 0 where C = concentration of carbon black. Equation of motion τ ij . i = 1. Material Properties: 1 2 ln(η ) = ln (η 0 ) − ((1 − n) / 2) ln (1 + (λγ& ) 2 ) − b(T − Tr ) τ ij = 2η d ij . η r .t + v i C . 2-D. of the rubber-carbon compound at different times is given by ηr = 1 + 5.i = 0 c. and average molecular weight profiles were computed. The transient. profiles. No comparisons with experimental data were provided.i ) The strain rate λγ& is given by (λγ& ) 2 = 2d ij d ij The relative viscosity.

F = 0. stress sub-elements and upwind Petrov-Galerkin were employed to compute viscoelastic limits.i = 0.i = 0 In sections filled with rubber. i = 1. A nonlinear. Computer Requirements and Execution Times: These are not given. cylindrical. Results: Solutions are given for an initial configuration of two rotating blades in an 81% full chamber.5 indicates a free surface boundary between rubber and air filled or vacuum regions. Predicted free surface distributions are given for rotations of 30°. 26 .M. F = 1. internal. internal single bladed mixer. D = matrix of rate of deformation components. Equation of motion – (matrix notation) dv ∇ ⋅ (T1 + 2η 2 D − pI ) + ρ[ − f ] = 0 dt where I = identity matrix. in air-filled regions or a vacuum F = 0. free surface and mixing dependent viscosity equations are solved successively at each iteration cycle. regions. Mixer Geometry: Fully filled.M. an iteration method is employed in which flow.021s −1 Boundary Conditions: Free surface equation – Tracking of the free surface is carried out via the solution of the probability distribution density distribution equation for the free surface function. incompressible. analysis of a fully filled.t + v i F. Solution Algorithm: The Galerkin F. viscoelastic fluid model was used with parameters appropriate for a dough-like material. Conservation of mass – v i . Dt S [Φ. Since the resulting equations are non-linear. F. T1 = matrix of non-viscous stress components. C ] ≅ K [Φ − C ] where K = 0. 60°. Field Equations: a. energy. p = pressure.E. 180°.i = S[Φ.E. F. 2-D.t + v i Φ . C ]. Summary of Reference 22 Reference [22] presents an F. 2 (incompressible) b. together with the upwind Petrov Galerkin method is utilized. together with contours of high temp. 90°. The no-slip condition applies to all solid boundaries. 2-D.DΦ = Φ . Elastic-viscous stress splitting. single bladed mixer.

the spatial coordinate was given for a range of relaxation times for mixing speeds of 1 and 100 r. This was done by reducing the mixing speed or the relaxation time. Constitutive equations – ξ ξ ~ 2η1 D = exp(ελ tr (T1 ) / η1 ) T1 + λ ((1 − ) T1 + T1 ) 2 2 b.c. Stress splitting – T=T1+T2 where T2 = 2η 2 D is the viscous component. E.p. Solution Algorithm: A mixed Galerkin formulation of the governing equations was used. and p were made in terms of the F. The field equations were recast in integral form involving the basis functions.m. effects are neglected. Velocity fields were plotted for various relaxation times and mixing speeds. basis functions. Material Properties: a. 27 . Energy Equation – Not used in the calculation. 4x4 subelements for the stress were used together with the elastic-viscous stress splitting. T is the total stress tensor and T1 is the residual component. Memory requirements and execution times were not provided. A comparison was made with results from the Bird-Carreau viscous model. Stability problems were resolved by gradually introducing the source of the instability into the computation. c. Computer Requirements and Execution Time Meshes of from 360 to 3360 elements were tested. Finite expansions of T1. Temp. A 2080 element mesh proved to be optimal. Results: Pressure vs.. dT1 ξ ~ ξ = T1 + (1 − )T2 2 dt 2 e. Viscosity ratio – η1 /(η1 + η 2 ) = 1 / 9 d. v. Bird-Carreau viscous model – η1 = η 4 + (η 0 − η 4 )(1 + λ2γ 2 ) ( n−1) / 2 Boundary Conditions: No slip boundary conditions on all surfaces.

b. Equation of motion – ∇ ⋅ (T − pI ) = 0 (creeping. Ct (t 1 ) . t . respectively. α . A detailed analysis of the development and implementation of the algorithm on Multiple Instruction Multiple Data (MIMD) computers is given. Field Equations: a. flows of viscoelastic fluids with memory. Ft (t 1 ) = ∂χ 1 / ∂x. Conservation of mass – ∇ ⋅ v = 0 (incompressible case) where v = velocity vector b. Viscoelastic Stress – TV [ x(t )] = ∫ [Ct (t 1 ) − I ] t −∞ M (t − t 1 ) S t (t 1 )dt 1 S t (t 1 ) = φ1 ( I1 . the Cauchy-Green strain tensor = FtT (t 1 ) Ft (t 1 ). I = unit tensor. and µ i = viscosity coefficients. 2-D. incompressible case) where T = stress tensor. p = pressure. Stress Splitting – T=TN+TV where TN = Newtonian and TV = viscoelastic contribution to the extra stress. M (t − t 1 ) = ∑ ( µ i / λi2 ) exp (−(t − t 1 ) / λi ) H (t − t 1 ) i =1 n is a memory function where H (t − t 1 ) = α /(α − 3 + I1 β + I 2 (1 − β )) where I1 and I 2 are the trace of the Finger and Cauchy-Green strain tensors respectively. An adaptive strategy that reallocates the workload to the processors is given based on the history of the nonlinear iterative process. I 2 ) [Ct−1 (t 1 ) − I ] + φ 2 ( I1 . I 2 ) where Ct−1 = Finger strain tensor. β = material constants. x(t 1 ) = χ [ x(t ). Material Properties and Modeling: a. λi = relaxation times. t 1 ].Summary of Reference 23 In [23] an algorithm is presented for the simulation of steady-state. 28 .

E.Results: The numerical solution of the 2-D flow for the memory-dependent viscoelastic fluid through an abrupt construction is presented with a detailed discussion of the computer implementation. No-slip boundary conditions at the wall. It has been found relatively easy to transfer the Intel hypercube code to other MIMD computers such as the Convex Meta Series. The finite element mesh has 256 elements.M. and fully developed Poiseville flow at the inlet and outlet sections were conditions imposed on the flow. Computer Implementation: The parallel algorithm was implemented within POLYFLOW using the communication library of the Intel iPSC/860 hypercube. 29 . symmetry boundary conditions along the axis of symmetry. The viscoelastic solution follows from the solution of the Newtonian flow through an incremental loading procedure.. A nonlinear iteration is started from a Newtonian flow field. The solution of the pseudo-Stokes problem with the viscoelastic extra stresses taken as body forces is solved with the mixed velocity-pressure formulation and the standard Galerkin F. The finite element mesh has 256 elements.

Kalyon. Lawal and D. Flow Patterns and Mixing Mechanisms in the Screw Mixing Element of a Co-Rotating Twin Screw Extruder. Osswald. M. Ishikawa and T. K. et al. ASME. M. Amano. Lawal. White. 1995. A. Lawal. P. 40. 1325-1337. Polymer Engineering and Science. Mixing Mechanism of Three-Tip Kneading Block in Twin Screw Extruders. 1146-1150.S. 2. S. 2000. May 2002. L.. Vol. of Filled Polymers. Kihara. 43. Feb. M. No. Yoshinaga. T. ANTEC 95. 42. 3-D Numerical Simulations of Nonisothermal Flow in Co-Rotating Twin Screw Extruders. I. 1996. Davis and T. Polymer Engineering and Science. Polymer Engineering and Science. A. M.. Melting Model for Intermeshing Counter-Rotating Twin Screw Extruders. [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] 30 . No. White. Vol. 2003. Ishikawa et al. 1. 42.. Davis and T. 5. Parallel Computing in Multiphase Flow Systems Simulations. 199. A. Hong and J. eds. Vol. Polymer Engineering and Science. Simulation of Flow in an Intermeshing Modular CounterRotating Twin Screw Extruder: Non-Newtonian and Non-Isothermal Behavior. 10. SPE ANTEC. Polymer Engineering and Science. G. 1999. 17. Comparative Study of Mixing in Twin Screw Extruders.E. M. 40. Ishikawa. K.REFERENCES [1] T. Modeling and Simulation in Polymer Processing Using Parallel Computing. 2. Polymer Engineering and Science. C. S. 136-143. 3-D Analysis of Fully Flighted Screws of Co-Rotating Twin Screw Extruder. January 2000. 2. Vol. J. Karniadakis. A. Kim. L. D. 1994. Funatsu. Feb. Rios. International Polymer Processing XIV. T.. No. Polymer Engineering and Science. S. No. and K. Vol. B. 39. No. ANTEC 95. A. H. 2000. Vol. Mechanisms of Mixing in Single and Co-Rotating Twin Screw Extruders. FED-Vol. Mathematical Modeling and Experimental Studies of T. A. A. No. A Non-Newtonian Simulation Method for Moving Boundary Problems. A. et al. and T. Gramann. 1999. Osswald. 61-67. A. Vol. Wilczynski and J. 6. Numerical Simulation and Experimental Verification of Nonisothermal Flow in Counter-Rotating Nonintermeshing Continuous Mixers. Railkar. Vernon. 11391151. Kalyon. 1715-1726. 155-159. 35. 108-112. 40. Osswald. B. No.

D. Polymer Processing Simulations Trends. 1. Osswald. J. Weiss. A. Aggarwal and R. No. et al. 56. BEM in Plastics Processing. Simulation of Viscous. International Journal of Numerical Methods for Heat and Fluid Flow. Vol. T. J. No. Vol. 132-145. P. No. 2. A. 2. T. Osswald. 197-202. SAMPE. V. Nassehi and M. 20. Osswald. Stradins and T. 698-719. Vol. A. 2. R. Comparative Study of Mixing in Corotating Twin Screw Extruders Using Computer Simulation. Vol. Rios. Tzoganakis. Engineering Analysis with Boundary Elements. 2003. Mixing Analysis of Reactive Polymer Flow in Conveying Elements of a Co-Rotating Twin Screw Extruder. C. L. Advances in Polymer Technology. Journal of Rheology. 89. Osswald. Kunststoffe plast europe. 107-113. Comparative Study of Rhomboidal Distributive Mixing Sections Using Computer Modeling. No. Keunings. 38. 17. Simulating the Flow of Multi-Domain Polymer Blends During Mixing Using BEM. Advances in Polymer Technology.. R. A. 22-33.. 2000. Osswald. Gramann. 10. 65-68. 2003. C. 19. Connelly and J. and D. [14] [15] [16] [17] [18] [19] [20] [21] [22] [23] 31 . A. et al. Conf. Rios. Kokini. J. 1998. 16. No. 2-D Numerical Simulation of Differential Viscoelastic Fluids in a Single-Screw Continuous Mixer: Application of Viscoelastic Finite Element Methods. K. 130-134. 405-419. T. 1998. 2-D Planar Cylindrical Deformation Using DR-BEM. 1994. 2001. T.[13] A. 2003. A. 22-41. Advances in Polymer Technology. Ghoreishy. Hernandez. A. P. and T. 22. C. 1. Vol. Vol. Erlangen. 6. Kunststoffe plast europe. Simulation of the Flow of Integral Viscoelastic Fluids on a Distributed Memory Parallel Computer. Germany. 1995. Osswald. R.. Vol. Gramann. Advances in Polymer Technology. 2001. Describing Processing Operations with Models. 1. P. A. 13. L. H. Strutt. Proceedings of the Society of Plastics Engin. 1999. Modeling of Mixing in Internal Mixers with Long Blade Tips. 2.

- Lecture in Transport in Arbitrary Continuauploaded byBryan Pajarito
- Gaurav Shankar 2013 PoFuploaded byVaratha Rajan
- Fluid Mechanicsuploaded byMohammedSafuvanKazhungil
- 7. an Unsteady Flow.fulluploaded byTJPRC Publications
- The Marsh Cone as a Viscometer Theoretical Analysiuploaded bysonny nebbi
- Fentrans-2uploaded byMuhammad Adli Pratama
- ADPuploaded byAdnan Mehmood
- JCE_67_2015_8_6_1204_ENuploaded bynicholas_j_vaughan
- Fluid structure models for dynamic studies of dam-water systems.pdfuploaded byhamid
- Fluid Mechanics Streeteruploaded bypiticmic
- s 0022112084000318 Auploaded bymariomato
- T42.06uploaded bykhturan
- Pszvba - Piping Pressure Drop Functionsuploaded byRafael Reyes
- lec1.pdfuploaded byakbar
- Lectures in Fluid Mechanicsuploaded bynasadosa
- notes GDuploaded byms280674
- pub-mb_3909uploaded byKrishna Anand
- SERRINuploaded byNikolay Gusev
- FEMConvergence.pptuploaded byNaveen Raj
- 13082417514303934uploaded byregil01
- CI3_321_Finite Element Method LM Notes 02uploaded bytallenuk
- Mmi 20018786(Densitometer)uploaded byMayur Patel
- 43789e8b74423uploaded byJohn Marisa
- 113851176 Fluid Mechanics and Machinery Notes 2uploaded byVeera Reddy
- Basic_dynamics Analysis User Guide (FEM) - Copyuploaded byKurra Srikanth
- Preliminary Examuploaded bySunil Garnayak
- Introduction to Finite Element Methodsuploaded byaap1
- ME 2204 FMM 2 Mark Question Bankuploaded byRajha Rajeswaran
- Mcmaster MECH ENG 3O04 - Fluid Mechanics Notes by Dr. Mohamed S. Hameduploaded bycttiger7135
- Review of Biofilm Modelsuploaded byMunir Wahab Rumie Grez

- FEAP Manualuploaded bykingmajor
- Impact Initiation of Explosives and Propellants via Statistical Crack Mechanics [Sd]uploaded bykingmajor
- WEPS-(21) Nuclear Weapons and Effectsuploaded bykingmajor
- Mohr Coulomb Modeluploaded bykingmajor
- Chapter 1uploaded bykingmajor
- Role of Gas During Explosive Claddinguploaded bykingmajor
- Comparative Study of Fragment Geometries [PHD Thesis]uploaded bykingmajor
- COMP L02 MacroModelinguploaded bykingmajor
- 10 Impact and Shear Ignition by Nonshock Mechanismsuploaded bykingmajor
- Simulations of an Underground Explosion in Graniteuploaded bykingmajor
- Preloads in LS-DYNA®uploaded bykingmajor
- Robert K. Mertonuploaded bykingmajor
- IVV Handbookuploaded bykingmajor
- #Derivation of Mechanical Properties for Sanduploaded bykingmajor
- An Illustration of the Concepts of Verification and Validation in Computational Solid Mechanics.pdfuploaded bykingmajor
- #Derivation of Mechanical Properties for Sanduploaded bykingmajor
- Thermomechanical Analysis of HE Binder With Time-Dependent Propertiesuploaded bykingmajor
- Military High Explosivesuploaded bykingmajor
- 2012LLNL Annual Reportuploaded bykingmajor
- MIL-STD-2105Cuploaded bykingmajor
- Lecture 3 Transport Phenomenauploaded bykingmajor
- Mil Std 2105cuploaded bykingmajor
- High Explosive Equation of Stateuploaded bykingmajor

Close Dialog## Are you sure?

This action might not be possible to undo. Are you sure you want to continue?

Loading