Multi-component Injection Moulding

PROEFSCHRIFT

ter verkrijging van de gruad van doctor aan de Techuische Univcrsitcit Eindhoven,

op g-ezag van de Rector Magnificus, prof.dr: J ,H, van Lint, V()(}[" "''''" commissie aangewezell door het College van Dekanen in het openbaar te verdedigen op

vrijdag 7 april 1995 am 14.00 uui

door

WILHELMUS FRBDERIKUS ZOETELIEF

geboren te Baarn

Dit proofschrift is p,opdgnkcurd door (k promot.orrm:

prof.dr.Ir. H.E.H. Meijer prof.dr.ir. F .P.T. Baaijens

en (k copromoter:

dr.ir , G.W.M. Peters

Multi-component Injection Moulding

This research was filIfllI(,ifl.lly supported by the E.E.C_. 1lI ('11(' form of the 1)I,rrF:/rmRAM project BE-i07(j, Contract nREU·49f).

CtI'·DATA KONINKLIJKE BIJ3LIOnmEK, lJ8N HAAG

Zoctelicf, Wilhclmus Froderikus

M ulti-component injection moulding / Wilhelmus Frederikus Zoetelief, - Eindhoven; Eindhoven University of

Technology

Thesis Technische Univcrsiteit Eindhoven. - With ref. ISBN !·JO-:)86-0016-X

Subject headings: injeel.i(II' moulding ; simulation / particle tracking.

Druk: FEJ30 druk, Ensched«

Voor Mur'Fnd

Contents

Summary

v

Notation

vii

1 Introduction

1.1 The multi-component. injection moulding process. 1.2 Literature overview

1.2.1 Multi-com poncnt injection moulding 1.2.2 Modelling aspects .....

12.;:\ Particle tracking

1.2_4 Fountain flow phenomena 1_3 Research objectives _

1.4 Outline of the thesis

1 1 3 3

5

7 8 10 11

2 Modelling aspects

2_1 Fundamental equations

2_1_1 Balance equations _ 2_1_2 Constitutive equations

2.2 Injection moulding .

2.2.1 Thin film approximation .

2_2_2 Pressure and temperature problem 2_3 Particle tracking

2.4 Front fiow . . . . . . 2.5 Multi-component flows

2.5.1 Interface modelling 2_5.2 Inverse mapping. 2.6 Piston driven flow.

2_7 Conclusions _

13 13 13 14 16 16 18 19 21 24 25 25 26 27

3 Numerical solution strategies 29

3_1 Temporal discretization _ _ _ . _ _ _ _ _ _ _ _ _ 29

II COl'Jt()fltS

:1.4

Conclusions , ,

:\,2 Mulf.i-compoucnt injection Jllol,llding :1.2. j Spatial di~(:n't,iza.i.i()11

:1.2,2 Pressure and temperature problem :;'2,3 COHV(!d,ion problcm , , ,

;3,2.4 Bifurcations of the midsurface :3.2,5 Fouutain now .. ,

;3,2,G !nLerf;H:(~ modelling

Mo(h'! flow problems , ,

,1.:1.1 Di~crd.iza.tioT1.. . .. ,"

3.:\,2 Piston driven flow.

J.:Ll Pist(HI driven contraction flow

:u

4 Material characterisation 1.1 Introd uci.ion . . . . .

1.2 Rheological eli aracterization

,'1. 2, 1 P()lys(yr(:TI(~ .... ,

11,2,2 Acrvloutrile-butadienc-styrene

4,2.:3 Polyamide ""'"

1.:3 11I/I'-d»,1.", and thermal propert.ins

4.,1. j Polystyrene

11.:1.2 Polyamide ,1.4 C()(l(:lu.,i()lls .

45

5 Particle tracking: two test problems

[),l Introduction , . , . , .

5,2 Piston driven flow.

5.2.1 Material and methods

G,2.2 Validation ,

;:',3 Piston driven contraction flow 5.3.1 M,li(>!'i,,1 and methods 5.32 Validation

5.4 Conclusions , ,

57 ;i7 G7 :)7 oo W m 7! 71

6 Multi-component moulding: some examples G.l Introducf.ion to the case studies

6.2 Co-injected strip with ribs

(;,2,1 Experimental conditions

(),2,2 Validation or t.ho numerical simularions 0.3 Bifurcatiou or the rnidsurfacc

G.1 Inverse Ttla(,)ping .... , , . , ,

6.4,1 Multi-colour moulding ... G,4,2 Multi-component moulding. 6,,5 Conclusions

7 Conclusions and recommendations 7,1 Cunei us ions ' , , ,

7,2 Recommendations"

A Fountain flow approximations

B Solution methods for the convection equation

References

Samenvatting

Curriculum Vitae

COIl tell ts III

97 99 101

103 103 - 105

107

111

115

123

127

iv Cont,cnts

Summary

In multi-component injection moulding, two Or three polymer materials arc sequentially or simultaneously injected into a mould to make products that contain e.g. a layered structure. The distribution of the materials in the product depends on the position of the gate, the geometry of the injection nozzle and t.he method of injection (simultaneous and/or sequential). Applications of this technology can be found in: shielding against electro-magnetic interference, barrier products, recycling of thermoplastic materials, and inmould painting of products but even ball-joints could principally be realised inside a. product when immiscible materials are co-injected.

The advantage of the injection moulding process with its versatility in geometrica.[ design of the products can be extended to include the ability of combining different materials with their specific properties in one product. 10 order to reach that target, Some limitations of the existing technology should be overcome. First, the number of different. materials has to be eXtended from two to three to be able of combining polar and apolar polymers in one product with the third component typically acting as a cornpatibilizcr. Second, the geometrical design of a product. should not. be limited by the process. These facts ask for a technique which is capable of predicting the material distribution in a product. In this thesis numerical tools are developed for calculating the positions of material particles during the flow in the mould cavity, With these tools it is possible to solve the inverse problem of predicting the injection sequence in multi-component moulding given a required material distribution in a product.

Starting with the balance equations for mass, momentum, moment. of momentum, and energy the fundamental equations are derived. Further simplification is established by assuming that. the mould cavities consist of narrow, weakly curved channels which results in the thin-film approxirnat.ion or 2~D approach. In order to complete the set of equations constitutive relations are chosen. The material behaviour of the molten polymer is COIl~[dered to follow a generalized Newtonian law and the density as a function of t.ht:' pressure and temperature is modelled by the Tait equaiton. A particle

v

VI SrtrllllWI",Y

tracking method ha~cd on t1H~ «ouservation of identity ill matcrral part.iclcs is added for usn ill multi-component moulding. The unique partid(~ idmtt,ity is ddin(~d by its entrance position and injection time whicb is p],(~HerV('d during the flow in the cavit.y Special uttention is focussed to :-;ati:-;(:yillg Uw local lIHtSS balance at the melt front and (j,(, now split.ting».

The modelling it; k~tcd by comparing numerical ami oxpr-rimental results for two well defined flow problems: the now uelw()(m two pistons adv:wcillg iJI a (,I)l>e with and without. a contraction. Using differently coloured slices of the same OJ 01' dissimilar materials combined if! one sample, the deformatioIl patterns are visualized. The deformation patterns thus obtained proved to be :sellsitive to all aspects that influence t.h€' flow kinematics. Dilferences in the: fluid behaviour between polystyrene (PS) and acrylontrilc-butadicnestyrene (ABS) could be observed easily. The flow through t.he contraction is modelled using a fixed domain approach, The pistons are modelled wi I,b an artificial high viscosity material that slips at the wall. This ~IW(:ia.1 Cil.SC of mnlti-componcnt flow gives a good approximation of the flow in a time d(ljl(md(mt domain,

Several applications of multi-component injection moulding arc demonsl.rat.ed. 'I'he colour distribution in a co-injected strip with ~ti m~Il(~r fibs is calculated ami is validated with experiments. TIl(, colour patterns ill th,' midplane and the bre'lk-lhl()Jlgh of the second material t.hrough the first ~l)()w(~d that. all aspects present in the experimental results could I.H' n:vn;},iNj in dw nu rnerical simulations. The effect ()f t,b" presence of" bifurcar.ion oj' th« midsurface, as occuring in flow splittings, is investigated numerically. Finally, two examples arc given of th« i nverse mapping technique. III t.ho first, a. relatively complex product is considered for use ill rnult.i-colour moulding whereat; the application of an iterative technique is domonstratcd for the filling ()f a ~impk rectangular mould with different materials.

Tho conclusion can be drawn that the a .. ssistancc of computer sj mutations is indispensable to attain tho required material distribution in the product. Moreover, it. Lurnerl out that modelling of multi-component flows, ill trst. problems at; well at; in injection moulding, proved to be rather complete alit! th1ls predictions can be made succesfully fegarding all those asp"el,s thai. can be considered important. in practice.

Quantities A, (L

(7

A

;), g

6 I

Operations and functions

A" a.b,A.B A:E A-',4' 11511,IIAII !,r(A)

Ad = A - lir(A)I

T 'T' ~

g , J

V

o. da Ti,-

Notation

scalar vector

second tensor column matrix

second unit tensor

conjugation inner product

double inner product inversion

norm

first invariant of a second order tensor, or trace

dcviatoric part of a second order tensor transposition

gradient operator

material time derivative

spatial time derivative

vii

vi u Notnt.iou

Chapter 1

Introduction

1.1 The multi-component injection moulding process

In almost all technical products that we know of, parts made of polymers can be found. The main reason for this interesting findings is that polymer materials are easy to process and, moreover, possess a number of advantageous properties compared to other often used materials such as steel and aluminium. Amongst the possible processing techniques for polymer materials, injection muulding is One of the most commonly applied. This technique can be used when mass-manufacturing of complexly shaped, highly integrated products is required. Different applications and classes of polymers have led to the development of different injection moulding techniques. Some rather specific exam ples arc:

• gas assisted injection moulding

• foam injection moulding

• thermoset injection moulding

• (structural] reaction injection moulding

• (resin) transfer moulding

• two-shot injection moulding

• multi-component (or sandwich or co-) injection mouldtng

For the sake of dearness, injection moulding will be explained briefly.

Figure 1.1 shows a schematic representation of an injection moulding rnachine, It mainly consists of two parts: an injection and a clamping unit. The

~ ChfJ.pt.nr I

injection unit contains a, screw rotating ill a, heated barrel pla,kt,it:i~,ill(!; th« gr<.lIlula(,('d poly 111(']' fncd stock conveyed from the hopper. When s\lflici(~[ti, material has been pla~tkl'I,('(l, tho screw ads as a, piston and pushes I,ll(~ melt. through ,I 1)()1':7.18 and runner syst.em inl.o the mould cavity. Backflow in!.o tho screw channel is gC[l(~nl,lly prevented by <.I. non-return valve, Tho clamping uni! supports the two !w)\Jid halves and prevents the mould from op('ningdespite of the high pressure that occurs during tho process.

An injection moulding cycle can roughly be divided into (,lime different kl.a.gE1S: th« injection, the p8,('king and holding, and Ow cooling st.agc. 1 n i.he injection ~ta.ge ,hl" molten polymer is injected in the mould. A rl(~I' complete filJiug or the cavity, extra material will b8 added to compensate for shrinka1'/\ the so-called packing and holding stage. From the moment l.h<~ !4il,l,(~ is completely rrM,nn, no more material can he injected and th(~ cooling stage st.(l.d.s. When the temperature is dropped below the ejf:rtion temperature, the mould is opened and the product. CUll b,~ taken out for further cooling- to till: ambient temperature.

Figure 1.1: Injection moulding machine

Ill. thE' multi-component injection moulding process, two or (,hl'(1(' polymer materials arc sequentially Or simultaneously injected ill the mould t.o make products with e.g. 11 layered sLnICI,IIl'n. The distribution or t.ho layers ill tho product depends on the position of the gaLe, gfmmd.ry of the injcr.tion !l ()/,/, !I' anrl tho method of injection (simultaneous and/or ~cqu(:'!lt.ia.l). By cha.nging the sequence of Injection or two components A ,)(,,1 B (A,B; A,B,1\; AlB: A,A/B,A; dr:.) t.hc thickness and also 1.11(' pla.ell (centric or e<:c(~ul.[·i(:) or the layers can be controlled (Sflll figllre 1.2), The rnult.i-eomponcut. injcr-

Introduction 3

Figure J.2: Schematic representation of two component injection moulding with sequential (A,B,A), simultaneous (A/B), and combined (A,A/B) injection

tion technology differs from the multi-shot technology, where after complete solidification the first shot is placed into a larger mould-cavity followed by the injection of a second shot against or around the previous shot.

1.2 Literature overview

The literature that reflects the current, state of research concerning the subject of this thesis covers the following four topics: a history of multi-component injection moulding, modelling aspects, particle tracking options and fountain flow phenomena. From these, particle tracking plays a major role in the modelling of thi' multi-component injection moulding process, since it is required t.o know were the material particles end-up in a product even if it possesses a complex shape. A major influence OIl the final particle distribution throughout the product, especially in the outside layer" is governed by the so-called fountain flow. As a result, material injected later in the injection period may even break-through previously injected material.

1.2.1 Multi-component injection moulding

The multi-component. injection technique was developed by leI in 1967 (Garner and Oxley, 1969). The process aimed at. the production of large, light.. weight, and stiff products, The sequential injection of two components (A,B, A,B,A) limited the practical applications to simply shaped sandwich constructions with three layers. The component B, the core layer, typically consistad of the A material with a physical or chemical foaming agent. J II this way, large parts with hard, flat, and glossy skin layers could be moulded without the need for high damping forces since shrinkage was compensated for by the expansion of the core material (instead of by packing at high pressurcs as in the conventional injection moulding process). Applications of the

4 ClliI/ItGI'

thick-walled (> 4 rnrn ] products could ]1(' Ioun.l in til" auromotivc. housing, and sanitary br;mdlf"- T 11 applications where thin-walled prod IJ('I,S ar(! more suitable, light-weight, high-stiffnrss products cOllld Iii! moulded mor« economically IIsiu)1; r;'l"foning ribs.

Initiated at th« DSM (;,'1I(.ra,1 Laboratory (Vcrlieijcn ci (1/" 1 08:.l) , dllring t.he las!' de,;ad(~, the multi-component injection moulding t.Achnology has also been evaluated for thin-walled products using the Battcnfchl tWO-(·ha.1I1Ifd (,echniqllil (figmn I.:) (after Eckardt, 1(87)). With this V~(:hniqll() not only

F'iguro 1.:1: Principal of mould filling; (A,l3,A) with the Ibtt(,nkld system

Introductioll 5

sequential injection is possible as in the leI technique, but also simultaneous injection of two components (A,A/B,A; A,B/ A,A)- In this way, thin, eccentric layers could be realized in a product.

Some examples of applications are:

• Shielding against electro-magnetic interference (EMI) of e.g. computer housings, where a Faraday-cage is created by a conductive layer inside the product (Eckardt, 1985, 1987).

• Barrier products, where a thin layer of a material with a low permeability for gases is enclosed by outer layers that provide properties like stiffness and visual appearance combined with water resist:ency (Eckardt, 1987; Tung and Kudert, 1990).

• Recycling of thermoplastic materials. The scrap material ends up ill the COre of 3 product surrounded by virgin plastic (for this application the original literature goes back to Donovan ei al., 1975)-

• In-mould painting of the products. With the sandwich moulding process a thin layer of paint will COVer the surface. No finishing of the product is needed.

• A thin (eccentric) layer of special polymer can be put just below the surface t.o provide for e.g, laser-writability.

Besides the development of the multi-component technology, structure development during moulding has been studied in order to gain more insight in the moulding process itself. White and Lee (1975) and Young et al. (1980) performed several tests wit.h different material combinations and injection configurations. In the case of isothermal conditions, they found that the most uniform skin-core structures occurred when the melt injected first possessed a lower viscosity than the second melt. Moreover, in case of simultaneous injection of two materials, the melt with the lower viscosity did encapsulate the melt with the higher viscosity during the flow in the runner section (macroscopic phase inversion).

1.2.2 Modelling aspects

The last. t.wo decades have shown many contributions to thc development of the numerical simulation of the injection moulding process. In the seventies the analysis of the filling stage, and thus the prediction of filling patterns, weld-lines and vent gates and of the pressure- and temperature distributions, was the main goal. Only simple geometries, of circular or rectangular shape,

() Cl1ltpl;nr 1

could Iw analyzr«]. TIl(~ flow was mainly Ulli-!iir(>d.ional with a t(~lllJwral.lIn~ Iicld t.hat was typically two-dimnnsional (one coordinate in t1H~ How dircction and (.IlA other ill the thickness direction of the product) kading 1.0 1.11(' so-called qD approach. In 1.1i(~ simulations, generalized Newtonian fluid bchaviour w(lS (\::;s~IlI\('d. Papers of Harry and Parrot (1970); Wu ci if! (I D74), Williams and Lord (1975); Lord and Williams (lD75) ,UT worth 1.0 hn mentioncd.

During t.he eighties, the simulations were extended t.o include ![JO!"!' mlll~ plcx geometries. Ridlil.rdson el. al. (1980) combined ba8ic 1 ~I) flow gcornet.ries, such as rectangular channels, circular pipes and discs, in paril.]k] and series to form arbitrarily shaped cavities, The basic, useflll and genorally applicable approach that. is used in all more ~pri0llS programs that il.t'(~ still in development at present, was originally developed by Hieber and Shell (l !l80; 1983). They «ombined the finite clement method few solving tho pn~SSIlI'f1 fif~ld in two dimensions with til", Iinit« difh'H,nc(' method for solvillg the temperature and velocity field in three dimensions. This so-called 2.~ D approach was later illlprovcci ami/or extended by many others (~"" ".g. Sil.krs, 1988; Boshollwel's and van del' Wed, 1988; Douven , 1991; Chiil!ll', ct. al.. I !'J91a).

In th(' s;unf' prl'iod, the 1 ~D approach was extended 1.0 include tho packing and holding stages of the process. Kamal and Lafleur (1982); Kanml r;/ al. (1986); Lafleur and l(allla.l (1986) modelled the cntir. ird!'(:(.ioll moulding (:yc:I(~ introducing viscoelastic material behaviour.

Today, the tendency is di!(>ctrd towards prediction of (,fa' {i na] IH'OJlf~I'tics of' injection moulding products rather than ~oldy analyzing t.ho mouldability of a product. Nearly all (commercial) codes usc the 21D <I.ppnl!)dl for all stages of the process combined with generalized Newt.ouian Iluid behaviour during filling (Boshouwers and van der Werf, l!}8tl; Duprct ,.nd Vand,'r8ciIlln'n, 1988; Chiang et al., 1991a, 1991b) and/or differential (LCOIlOV) models or integral Wn.f(nrr models to account for the vj~Cl)-('!asl.ir: ~I.rnssns in a. direct (Baaijcns, 1991; Papathanasiou and Kamal, 1993; Kamal and Paputhanasiou, 1D93) or indirect, docoupled (Douven, 1991) approach. The product. properties that can be predicted are:

• flow-induced stresses (Isayev and Hieber, 1980; Flaman, 1990; Baaijcn», 1.991; Douven, 1 ~J9I)

• thermally and pressure induced stresses (Titornanlio ct al., IDS7; 1.),)U. wn, 1991; Baaijens, 1991)

• density distributions (Greener. 1986;), 198Gb; Caspers, 1995)

lntrodnctlon 7

• dimensions and shape of products (St. Jacques, 1982; Isayev and Hat" iharan, 1985; Douven, 1991; Caspers, 1995)

In literature not much attention is paid to the modelling of the multicomponent injection moulding process. Turng et al. (1993) modelled the sequential injection of two (different) polymers by calculating the residence time of particles that enter the mould cavity. By doing so, they were able to predict the position of the interface between the two materials. However, their method is restricted to sequential injection of two materials with one switch-over time only. In Peters et al. (1994) examples are given of both sequential and simultaneous injection, Their method, which uses a conservation of identity, can be extended to model all types of multi-component injection moulding.

1.2.3 Particle tracking

For obvious reasons, in the particle tracking technique adopted in this thesis, IlO real particles are present that exert hydrodynamic interaction to each other as is the case in the How or the motion of particles in e.g. slurry flows. The 'particles' are abstract distinct points in the flow that have to be followed ill time and space. Using similar definitions, particle tracking recently got considerable attention in numerical simulations of non-Newtonian flows employing memory integral equations (sec e.g, Crochet ei al., 1984; Luo and Tanner, 1986; Luo and Mitsoulis, 1990). Since integral models take into account a part of the deformation history to calculate the Cauchy-Green and Finger strain tensors, it is essential to find the particle position as a function of tit,· (residence) time. J 0 two-dimensional steady flow situations, particle tracking was simplified by the formation of special streamline elements in which the element boundaries follow the streamlines (Luo and Tanner, 1986). However, these methods failed in recirculation regions. To solve this, Luo and Mitsoulis (1990) tracked particles individually for every nodal point in the finite clement mesh. Based on the expression for particle acceleration in steady now, they derived a step-by-step scheme with third-order accuracy, provided that the velocity gradients are exact For the same purpose the particle paths can be constructed by integrating the reversed velocity field starting from every nodal point in the mesh (Crochet ei al., HJ84). Caspers (1991) used this method in injection moulding simulation ami solved the initial value problem with a Runga-Kutta method. However, in transient calculations, as in mould-filling, this requires storage of the complete velocity field for every time step. Thus, tremendous storage capacity is required. To get around this problem, Peters et al. (1991) developed a method of

8 C/lilptor I

conservation of- identity for mult.i-cornponent injection moulding hy ddilliTlg material particles by their unique identity (e.g. material, colour, plac« and time 0[' injection}, Tho transport of tilt, idelltity during the llow call l~e described by a ])111'(' convect.ion equation, which call lw solved in either <I Lagrangian 01' Eulerian way,

Particle tracking call abo be used for other purposes. Matsuhiro and Shiojima (1900) determined the moving free surface by transporting marker part.iclos In the front region. Those markers arc convected !1I1111(,riulJ ly by integrating; Ol!' Vl'locity Ellie! denoting the newly fill(1d clements at t.h« [ront.. Hanna-chi a.nd Mitsoulis (1090) showed that particle tracking mav provide it better understanding of How kinematics by calculat.ing the residence Lim" distribntion of material points through the integration (If (.h(, vol()cit.y field and by plotting isochrones and the deformation field 01" Iluid elerncnts, These kinematic patterns provide a Lagr~.!Il\ian representation of' t.h(~ flow ,1)1(1 (](oscribe till' ddOrTnl-l.l.ion history of fluid clements itS they travel through U}(' field. The patterns thus obtained can be compared ('<lKily wit.h exporimcntal n'~ult., if (:(>1orrd I.racnrs are used (sec e.g_ Vos et. al.. 1991).

1.2.4 Fountain flow phenomena

In flow::; with (flIP or more frp(~ boundaries and a no-slip condition at. 1.11(' walls, fluid clements adjacent to the moving front. experience t.ll(' ~o-callnd fountain .flow (Rose, 19G1) _ Tilt' fluid ncar tho center moves at a higher speed (ball 1."(' local :-l.vpl'il.gn speed across th« channel, and when reaching the Ironr., ~i)!'C'Lds toward, the wall« (ligll!'\~ 1 _4). In experimental mould filling studies, (,h(,s(~

.~ .. - ~~-:------

-=-=;;! -- ---"

-~

------

velocity profile

t'igHf(~ 1.4: Fountain flow phenomena

specific Ilow patterns behind the advancing fl'~e surface were observed as W(~l1 (Schmidt., E174, Gop,o:c; ei al., 1986; Coyle rot ill" J.987)- The fountain now din~d.ly aH-(~cts transport characteristics such as the temperature distribution, conversion distributiou and fibre orientation not only at t.he front hut. ill (.)u' whole domain that is d{)s(, to Lh" walls, thus the (import.ant) outsides of I.h(' product. Beris (1987) showed that L1H-~ observed V-:;hap(' dclormat.ion of

Introduction 9

fluid elements is independent of the constitutive equation used, since it is completely determined by the balance of rna:>s.

Studies dealing with the modelling of the fountain flow describe either an interface between two (immiscible) liquids or analyze a free front (thus one of the liquids is inviscid: air or vacuum) advancing in a capillary tube or between parallel plates. Bhattacharji and Savio (1965) made the first attempt. to arrive at a semi-analytical solution of the Navier-Stokes equations for the liquid/liquid interface lor Newtonian fluids in pipe flow under action of a piston. Castro and Macosko (1982) derived, analogously, the expressions for flow between two semi-infinite parallel plates. MOTf~ recently, Haiqing (l993) gave an analytical expression [or characterizing the fountain effect in rectangular shaped moulds with finite dimensions. In his analysis, an arbitrary dynamic slip boundary condition at the walls could be imposed, in contrast with the work of Bhattacharji and Savic. Unfortunately, no explicit expression can btl found in J·l(liqing (1993) for this specific case.

Numerical solutions for the 2D flow between parallel plates, containing a transient free surface flow, were obtained using the Marker-and-Cell technique (Gogos et ol., 1986; Kamal and Lafleur, 1982; Kamal et al., 1986; Lafleur and Kamal, 1986), or finite element methods (Coyle et ol., 1987; M"v,idi" ei ol., 1986, 1988). Behrens et al. (1987) investigated numerically t.he fountain flow "ffeet. in a tube. In these studies, special attention was given to the motion of the free surface and the movement of the contact point at the wall.

Several models are proposed that can be used in combination with the 2~D approximation in injection moulding (Manas-Zloczower et al., 1987; Duprct and Vanderschuren, 1988; Sitters, 1988; Garcia et. 0.1.,1991). In these models, the contact line between melt and air is a straight line that is normal to the mould walls and moves with the gapwise-averaged fluid velocity (figure 1.5). The essence of the flow kinematics in the flow region is captured by transporting particles that enter the front domain instantly from the high. velocity region in the core to the low-velocity regions near the walls based 011 a mass balance along the contact line. The point that, separates the highand low-velocity regions is travelling with the average front speed. However, this 'flat-front. turnover rule' (Shen, 1992) neglects the residence time of the particles in the front domain. Peters ei al. (1994) incorporated this residence tim" by parameterization of the results of particle tracking in the front domain according of the model of Bhattacharji and Savie. The residence time and the position of leaving the front are calculated only once and are fittE'd with a high order polynomial function. In this way, a sufficiently accurate df's(TiptioTl of the fountain effect can be established.

Recently, Friedrichs and Giiceri (1993) developed a hybrid numerical tech-

v

~---.- .

v

Figure J .5: Simplified front model in injection moulding

nique 1'01' capturing the flow kinematics in injection moulding for w'ollidrical simple domains. They NHIlbilH1d it 2D and a 3D finite element mesh t.o ,,)()di'l tile .3D fountain flow kinematics. The 2Dnrm~sh covers t.he, 1"<"l-\io11 behind the advancing front (Hole-Shaw approximation) and the .'3])-llHdl is used ill t.hrfront domain. With this approach the ma-in benefit of solving t.he flow kinemaries in 2D gets lost and thus this approach rep[(!Sr~llts it method (,IJ<~t is only a small st.ep away from a complete 3D analysi» of 1.1)(' whol« domain which would give a much i<u'g('" improvement,

1.3 Research objectives

The advantage, of thl, injection moulding process with its versatility i!l 11:(,0- metrical design of products can be extended to include the ability ()f COlJlbining dlfl'"(~r(~nt. materials with specific properties in one product. III order to reach that target, ,~(lTrH' limitations of the existing (,!:chnology should hi' OWl" com('. First, the ability of inject.ion of two components has to be cxt('llIlccl to three components 1.0 be able of combining polar and apolar J!)al.!:ria.ls ill ()II(: product with the third component typically acting <18 (I. compatihilizer SAcoll(i, the present limitation of the complexity of (,h" products to l"eiativ(!iy sim pic g(:omdrie8 has to I')(~ removed,

Giv(~TI these nHI.jOI' drawbacks of the existing technology, 1'1)(: research objective of this thesis can be summarized as:

Thr: doveloprncnt of numerical tools for calculating the llow Pil,1,h of particles. With these tools it will be possible to solve IfJ(' invorrr«] problem for predicting UHl injection :-;(:qIH'nC("H in multi"OI"I'OIl("lt injection moulding (Iiglll"f! I.G). Tho results shollid provide all I nlormat.ion needed for progr;;unmin?; the control syst(~m of the injection moulding machine.

Introduction 11

During the research project considerable effort is put on the development. of a three component injection moulding technique including the machine and the control technology to remove the first limitation mentioned above. However, the set-up suffered from some problems and experimental results obtained with this machine will be expected to be accomplished in the near future.

",_.". ~ ,._.,_ _ , .. """"'." " "' ,~.,, ~ 1

MAPP'Nif

Figure 1.6: Inverse mapping

Apart from the direct practical applications, the multi-component technology can be used to produce tests on the quality of computer models developed (.0 describe the dillE'rent stages of the process. JJy using multi-colour injection, the experimentally obtained deformation patterns can be compared directly with numerical results. The total deformation history of every particle in the mould is important for all properties that are history dependent, such as e.g. the calculation of the local temperatures when the Peclet number is sufficiently high, or of the transient viscoelastic stresses for high Deborah numbers or the prediction of the local conversion in reactive systems for high Damkohler IV numbers

1.4 Outline of the thesis

III order to make the multi-component technology more suitable for producing complex shaped products, VOH ci al. (1991) proposed a more flexible

12 Chapt",. 1

injection met.hod that circumvented the relatively long opening aud dosing times of valves in a more traditional distribution device. Using an ar.cumulator (sec figure 'l.7) in between the distribution device and tho mould, any conliguration requested can th~n be composed during the cooling ,)("i.i!,P of the previous product rather than during injection. The contents of t.JH! (heated) accumulat.or is subsequently illj('c«('d into the mould. In this thesis, ,~('VPrltl

Fi~ur(: 1_7: I~a,sic d",sipl of all i\(cumul",t.or for the usc in rnu lti-cornponcut iu_j,,,,Lion moulding

st.('ps in the dcvclopmeut of the multi-component, injection moulding l.;;cilnnlogy will he investigated separately, Starting with a. piston driven flow, tile: flow dHl'-':I(,L"ri~i.i(s will be studied !'sprrially in the (foullt.n.in) rCf\i()n~ dosl' t.o the pistons. Nf7xL, tlln piston driven flow through a cont.ractiou will I.w analyzed, providing the knowledge of the process of fillillg alld emptying <1[1 acr.umulutnr. Finally, multi-component injection will be inve~tig;It.,·d lIfiinf!; it distribution (hlVir(~ which can be Iound in the existiru; multi-component i njoction moulding machines.

In Chapter 2 or this thesis, the set of equations needed to model the mul ticom pouent iujcction moulding process is summarized wit.h special (,xi,e'lsions for modelling part.iclfl tracking and the fountain flow phenomena. Chapter:J deals with the numerical aspects, involved in solving tho problems (,IIC{)lIl1" tered. The mat.erifl.l data necessary in the simulations me given ill Chapter 4- Subsequently, Chapter 5 presents two well-defined flow problems which an-l IlRPd for testing of til" modelling. Some examples of multi-component injection moulding arc shown in Chapter G- Finally, Chapter 7 summarizes t.he main results of the research and gives suggestions for futuro research.

Chapter 2

Modelling aspects

Modelling of the multi-component injection moulding process is almost identical to modelling of conventional injection moulding. Therefore, the derivation of t.he basic equations is not repeated in detail. The reader can be rcferred to existing literature (e.g. Hieber and Shen, 1980; Isayev, 1987; Dupret and Vandsrschuren, 1988; Sitters, 1988; Chiang ei al., 1991a; Tucker III, 1991; Douvsn. 1991).

In Sscticn 2.1 the rnain balance and constitutive equations will be summarized. Section 2.2 deals wit.h the injection moulding process. In Section 2.3 particle tracking will be discussed followed by a treatment of the fountain flow phenomena in Section 2.4. The chapter ends with the formulation of two test problems for particle tracking.

2.1 Fundamental equations

The set of equations needed to describe moulding processes consists of the balance <'qllat.ions and the constitutive equations, completed with the proper set of boundary and initial conditions.

2.1.1 Balance equations

From continuum mechanics, which deals with the determination of fields of density, temperature, and motion for material points considered, the balance equations can be derived. In their local form they read:

Bolonce of m(~ss (contin'U'ity equation)

fi + p~. V= 0

(2.1)

13

l4 Cilal'l;or 2

WIH'ri' {) denotes the density, '11 the velocity vector, and ~ tho gradi('Ht, ()ppr· at.or

~ ..

\1 . a" + IJI ""'" p.i!

with a' the Cauchy sLn~~S tensor, and f the specific body rorC('.

Bolance of riunnent. ()J nunneniuni

(2.a)

Bolancc of cnngy (first law o] lhn·t/u)dy'fl.ami(s)

(li,=a-;D-V·'~+pr

where c is the specific internal energy, D iR tho rate of deformation V'JI,,;or, II. the heat flux v(~d()r, ~I.nd 't is a specific heat HOUlT('.

2.1.2

Constitutive equations

The ~(!L or k),hw(;c cqll'l,ti()n~ can only be solved if relations arc known for o , h and !: as function of density, \'('Illl.l('ra,tllrr and motion. In t.hi« ;-wdlon Lhc cOHstit.utivI! (!qllaLioIls for tit\) Cauchy stress tensor and th(, t.horm»! properties will he given in their final roull. For a more detailed description tlw reader is rch.'tTI'd t.o (Douvon. If)91, Chapter 2).

Cauchy stress tensor

M 'my ddr(~f(mt constitutive models exist that relll.tc til!) Cauchy Htr('~~ k!l~or to the independent variables p, T and .T. Siner: WI: arc b;\~ic;j.ll'y interested in the flow kinematics, in this thesis only the ge[l(1ralizod Newtonian fluid model will he 1.I::;e<1. This model i~ commonly employed in ~illlulal.illg t1w injection phase of thn injection moulding process, although it n(~gkcts all tho viscoelastic effects of the polymers; this proved to be not too ~(!rious (~(,e Baaijcns, 1£)91; J)OIIv(->ll, J991) since boundary and initial conditions arc given as prescribed velocities and thus thr kinematics dominate tlli' flow

Ccneralized N ewt.onlan fluid behaviour The Cauchy S(.I·PSH tensor ais split into a hydrostatic part ···pI and an exl.ra stroHS tensor 7":

a =-'pI -t- T.

Modelling aspects 15

The extra stress tensor is assumed to satisfy the generalized Newtonian fluid model:

T = 2ryD'I,

ry = ry(p, T, 1'), l' = JIDd : nd

Dd = D - ~tr(D)I, D = ~(V'rJ+ (Viir),

(26a) (2.Gb) (2.6c)

when: 1/ is shear viscosity, p (.he hydrostatic pressure, T the temperature, I (.Ii" unit tensor, nil the dcviatoric part of the rate of deformation tensor, and l' the shear rate (second invariant of Dd). Two models lor t.he steady state shear viscosity are widely used in simulations of moulding processes (Hieber, 1987): the Cross model

7/('" T "') ,:;;; T/O

r» " 1 --I ( . / .)l-"

- 1/0"( T

(2.7)

and the Ccrreau model

(2.8)

where 1'/0 = T/O(p, T) is the zero shear rate viscosity, T' and ti are constants. The Cross model fits the viscosity Curve well for polymers with a broad molecular weight. distribution, whereas the Carreau model can be employed successfully for models with less polydispersity,

The pvT- relation For amorphous polymers the double-domain Tait equation is widely used (Zoller, 1982);

( T) {(oo,+a',('/'-1'9))(1-U.U8G4111(11 -!,))

11 p, = ~;

(<<Om +"1»> (7' - r.))( I-O.OS94 In( I + -;f.;;:))

'/~ = Ii,o + sp,

8m = BOme-H'm'/', E" = Eo"e"B,,7',

(T;:;7'o) ('/'/1;' )

(2.9a)

(2.9b) (He)

where '!~ is the pressure dependent glass transition temperature and ao"" a'H', Bom, Elm, ao." (4[" Bo" J]ls, and 8 are constants.

ThennaJ properties

Finally, Fourier's law for the heat flux vector is used:

h = -A,. VT,

(2.10)

l() ella£) I,cr 2

with)" Lhe therma] conductivity tensor. The specific internal PTIf'q-\y c: call I)(~ r.;q)J(~~sed as (Sittns, I DES),

(:2. t 1;1')

(2.11 b)

.II ',,'I:+F/P,

(2.1·1c)

where (:" is tho thermal capacity at constant hydroxl.at.ir: prcssun: awl g is I.he ~p()<:i!i(: unt.halpy, Equation (2.11<1) is derived under a~~llmpl.i()i1 thai t.hr: elasticity has it negligible contribution to ttll' TIIi'chanica.l dissipation.

2.2 Injection moulding

The simplifications commonly applied i n injection moulding will 1)(' SIIl11111<lri~,(:d, with (:TIIphll,;;is on I,h() thin film approximation and the definition of 11\(, pressure and temperature problems to bo solved.

2.2.1 Thin film approximation

In most practical upplicatious the thickness or a injer.t.ion moulded product is small (0.;' " mm ) compared to the other dimensions. CiI,vil,i('s thus consist or narrow, weakly (,11I'v(~d channels. Therefore the Helc-Shaw or I,ldl' fil1!l approximation can b() employed (sec c.g, Hieber arid SIl(')I, t llRO; Sit.ters, 1988; B()sh()llw('r~ aud van dpl' Werf, 1988; Douvcn, l~)~Jl, Tucker II I ,1991).

A local Cartesian vector basi) Of : {f"'~I'';<} call he ddirwd ill every point of the midplane of' the mould. The vecl.nr I~-: is normal wi th rcspect (0 (,11(' midplane (figllJ'e 2.1). An arbitrary vector (1 can be d('c()mposl,1I i II a. vector il: parallel to the midplane and it component in t.he di['(~ctioll of I=:,

(2.12)

Ili~('<\IIN() "I' (,II,' 1Irgh viscosity of polymer 1Il81ts, the Reynold" numlxr (mtin between viscous and inertia forces) is small and, therefore, t.JH! inortia ill the momentum equation (2.2) may be ncgJr.ded. The main assumptions made in the approximations arc:

• the pressure is constant in the gapwisc c-: direction .

• I,hr.' wl()l:i(,y gra.di('lI(,~ parallel 1.0 the midplane ani small compared In those in E" direction.

Modelling aspects 17

Pigure 2.1: Local base OE

.. t!i(c thermal conduction parallel to the midplane may be neglected with respect to conduction in E, direction .

.. the velocity components in the f" direction arc small compared to those tangent to the midplane .

.. the mould is rigid, so the pressure field docs not interact with the cavity dimensions.

The mould is cooled below the glass transition temperature causing the solidification of the melt that contacts the walls. These solid layers grow inwards from the mould walls during the process. In the solidified layers the velocity components parallel to the midplane arc assumed to be zero while small non-zero velocity components in the E'z direction may exist due to density changes. The local solid-liquid interface positions are given by h-t, and h+tz for the lower and upper solid layer respectively. The total local mould thickness is 11.(.1).

As it consequence of these simplifications, the deviatoric part of the rate of deformation tensor reduces to

Dr! = ~~(ijt' + S_?r). (2.13)

21Jz' z •

18 Clmp!or 2

Then, t.ho shear rate, ddinnd as rj = V2D,J : D,l, is given by I :::: I. With these simplifications tlil' set of equations for a generalized NI'wtonian fluid j'(~ad

'<"\. .... Dv;

v 'v +-

Dz" p

~. a ( Oil')

'V po" - .,]~- ... : ,

Oz" i)z'

. i) (fJT) . T (i)I)

(if' 'I' = A- + 1/"" ____L 1).

I' '»' "".(" ! '0'/'

( ,~(h fI 0. l'

Op -=() a;:' '

('2.14")

The ('qll<},tions (2,11a)-(2.14c) n,!'(~ instationary, non-linear and cOllpil!d beI:iWS(~ viscosity and density arc both tcmperat.ure and pressure c11~p(,Bd('nt"

2_2.2 Pressure and temperature problem Pressure problem

TIH' pressure problem (PI') can be derived by iJlt<'i~ratillg (~qllat.ioll' (2.l'la) and (2.14") over the thickness (, eliminating the VI!l(l(:it.y components (DollVilli, ID!H, Chapt.or 2). 'I'h« resulting (!(jIl(j,I,iotis defining 1,1)1' PI' rf';ld:

hl'J; 11./2

V', (,q~'p) - I /';'11r1( = - .I «t «,

-·/,/2 --"/1

with

~ Jr

.'i =.12 - -] ,

• II

h+

IX'

J, ,= --_ ...... d(,

, "~I

h

(2,1(i)

and the boundury conditions:

• Injection area:

{llll-lle, q/8

P = pm",

11··I!i.·'·11 = 0

iM'

P = 11"},,,

if P < prn;,x otherwise

(2.l7)

• Mould walls:

• Flow fronts:

where S is th! iluid it.y, q the prescribed volume flux per unit of" i(~Ilf.\"th, ii" the normal vector in midplane or product and fJrmx and j)"lm tho maximum machine pressure '111d t.he atmospheric ]Jr<~S8111r. respectively,

Modelling espects 19

()nre the pressure is known, the velocity if' can be evaluated with (Donveri, 1991)

h+

-.. J 1 ( .11) -',

1) (z') = - X - - dX \l p.

rl Jo

,<

(2-18)

Temperature problem

The temperature problem (TP) can be summarized as:

- 8(of) -2 T(&P)_

pepT = &i ,\§;~. + 17{ ~ p aT p P,

(2_19)

with the boundary conditions:

• Injection area:

• Lower mould wall:

• Upper mould wall:

• Or: cooling to air:

T=7j

>.g;; = H-(T - T~-;) ,\:;; = H+(T - T,;;)

&'J' )

'\()z' = Ha(T -- Ta

(2.20)

(after ejection)

where 1j is the injection temperature of the melt, T;;" T~ the lower and upper mould wall temperatures, T" the ambient temperature, and H>, H:r , H a the heat transfer coefficients of the lower and upper mould wall and to the air respccti vely,

2.3 Particle tracking

In simulating the multi-component injection moulding process it is necessary to know where and when a material particle has to be injected to end up at the desired position in the product, Particle tracking can be carried out by constructing the flow path f for a certain particle, The track of a particle during the time span [t, t - 5J can be constructed by using:

£" = 2,,(£, t; t - s).

(2-21)

Equation (2.21) states that i, is the position at t - 5 of a particle which occupies the position :r at time t. Connecting all positions X8 with S E [0, t), its total particle track (or flowpath] is obtained. Summarizing, the trajectory f of a particle which occupies the position x at time t can be expressed as:

f'::(;s CO :ts(.£,t;i -~) v~ E [in,t],

(2_22)

Because tho velocity vcctmu(£, L) is tangent. to P, the Ilow pat.h call he obt.ained by intl'grating

el:1'

.- ..... = 'i1(:?, I.) with initial condition: :i'o":7(1, = 1,,,). ell

This initial value problem call he adequately solved by a Runga-l\lltta inI.ngril.tion method (Cro('.hd et. al., H)81, Ca::;P(TS, 19(1). Employillg tlli" method, the velocity field (,ll" ""v("al l.im8st(~p~ in t.b(~ past. has to Lc st.orr-c], which !'('qllirns extra storage capacity.

An alternative method of particle tracking i.~ to ident.ify particles, when entering I.he flow domain, with an unique ~f)t of labels E, n~pr(~s(~nl.ln14 their 'idcol.ity' (1\'(,1'1'.'>, 19WJ; Zoctcliof', 1(02). During How, 'iJH: identity of each part.iclc docs not. change (figure 2.2). Using the formal ,[(,f[nit.ion of it matoriul

fl",

,----------- I-

~ ------

... : -. - ------------

---...:-----

--_

---------

li'igllnl 2,2: Conscrvatiou of idenrity of a particle (lidwlillp;)

deri vat.i V('

IN) dd' iN' ~..f1t) - (/)'" -+'11' V<P

DI. iJt. '

(with t.lte spfl.l.i'·I.1 derivative [hl)/ iJt defined as t.he rate of dliUl~(' of Ii) 'J.1. il. lix(!d position in space) the conservat.ion of identity cnn h(~ (!XPJ'c!ss<"d as

. D~

~ = Tif = () ill lim, with ~();, t) = f(:Z, t) on J',,,

(2.20)

whor« ( is for instance a column containing the sturtiug position and time (:J:o, 'I/O, -"0, to), I', i.h« injor.t.ion area and Om the (:,wiLy W'otiidry, The pl'ohl(,rt1 of partir.l« tracking can then he solved with either 11 Laglilngin.11 (lI\ovillg frame of \'d(T('W~(') Or an Eulerian approach (Iix(,d frame of rcfcrcnc«)

Modelling ;)'I5PO<;tl5 21

Eulerian approach In a fixed frame of reference the conservation of idcnCity can be expressed as:

(2,26)

In this representation the change of label values in space is calculated instead of tracking pre-defined particles The actual tracking can be carried out by determining the pcsitions of a given set of labels in time if materials with Identical material properties arc used- For use in multi-component injection moulding, label values at the end of filling provide all the necessary information: i.e. where and when all particles had entered the mould. I3y defining a specific, desired shape and position of a second component inside a product, only the label values of these particles have to be determined to know what configuration should have been injected to realize this desired multi-component product.

The main advantage of this method is the uSE' of field information rather than tracking a set of particles, For every given material distribution the injection SHqUf'T1Cf' can be determined easily by looking for the label values, the actual particle tracking has not to hE' performed again, In the Lagrangian approach every newly defined set of particles result in restarting the particle tracking procedure, Moreover, velocity data. at every time step must be available which requires large storage capacity. Another advantage is that the method is somewhat less sensitive to erroneous velocity values, e_g_ a velocity that points out of the mould, When trying to determine the particle trajectory in this case, the particle will end up outside the mould and the method fails (Caspers, 1991),

2.4 Front flow

In mould filling, two flow regimes can be distinguished: the main flow domain and a front llow. In the last, the fountain effect occurs while in the main flow a simple 2]) stratified flow situation exists (sec figure 2,3). The fountain flow affects the particle distribution in moulded products, especially for material elements close to the mould walls. Also, break-through of core layers towards the surface of a product is completely controlled by the fountain region. For those reasons it is important to incorporate the fountain flow in UtE' modelling of moulding processes.

22 Cbnpt.er 2

","in flow

fountain region

Figure 2.:1: Fountain region in the Hclc-Shaw approximatiou

In (.h(' id0~1.1i~('d fountain flow, the residence time is neglected and a S[It1 pin model, based on conservation of HIU::iS, is applied (Manas-Zloczowcr ci ai., ID87; Duprot and Vundersehuren, 1988; Sitters, 1\)88). In t.hCHC approaches, Uw fountain flow region simplifies to a single straigh; line t.lia.t mnvr-s with t.ho a.v(~mgn front velocity 'i\ (soc figure 2.1). Particles I.ha(. "uV,,' 1.1". frOllt. region at. a position "'i will be transported instantaneously t.o a position z; where they leave the front .. The relation between t.h« ingoing positiour, and I.h(~ outgoing position .c" is only governed by a local ma~1) ba];\1Ic(' (HI till' front. line, Depicting the fountain flow ill th(' Lagrangian framework with tlw mould wall moving with -n" the front flow can be divided in two pm(.s (ligan) 2.4): a high-velocity region () < Z < zJ and" low-vulor.i tv rngi'H' .cJ < z < h/2, where ;;J is tile position wbenlll.,(Z) =iI.,. According to IJw

/frorlt

h!2

o

x

Figure 2.4: Lagrangian rcprescntution of simplified Iront model (only OJl" mould half is considered]

balance of mass (assuming an incompressible fluid), 1'.1'(' relal.ion !H\(.WP<l1l tho

Modelling aspects 23

inflow Zi and the outflow position Zo can be expressed by;

I! ~~)

.I (IJ,() - v,) d( = J (V; - v.(()) d(

(2.27)

In the case of a Newtonian fluid (in an isothermal flow) it can be derived easily that the folding line is positioned at Zt = 2~' Substituting t.he main flow velocity profile into equation (2.27) and integrating, results in the relation

Z<~ - Zo + Zi - z; "" O.

(2.28)

In 11 similar way the expression for a Power Law fluid reads (Manas-Zloezower ei ai., 1987);

(2.29)

In this case the position of the folding line is z/ "" (~ + 2) ;:;~~. For models that predict the viscosity more accurately (e.g. Cross model or Carreau model) the folding line is situated between the extremes given by the Newtonian and the Power Law fluid. The actual position depends on the shear rate and the material constants n (the power law index) and T' (denoting the transition from the Newtonian plateau to the shear-thinning region). For these viscosity models the folding line can best be determined by evaluating equation (2.27). Using this approach, an estimation can be made of the gapwise distribution of particles wit.h material quantities attached. It should be noted that it is assumed that these properties arc not influenced by the (fountain) flow.

For some applications, the residence time in the fountain flow region is of importance, e.g, in the case of multi-component Or in reactive injection moulding (Castro and Macosko, 1982). For a Newtonian fluid, the semianalytical model of Bhattacharji and Savic (1965) gives the opportunity to derive a relation between the residence time in the front flow and the inflow or outflow position (Zoet.elief, 1992; Peters et (L/_, 1994). 'I'he original Bhattacharji and Savio model gives an analytical expression of the velocity field ill the fountain region (see Appendix A) from which the flow paths can be constructed (figure Vi). The residence time as Iunction of the inflow or outflow position can be computed by integrating along a flow-path until the particle re-enters the main flow. This representation can be parameterized by fitting the Curves obtained with a polynomial function. In doing so, not much extra computing effort. is required for incorporating the residence time distribution into a front model.

24 (;h;l.pt()T 2

I), I _ . ....:_===---- ... -_-_._-_-.--=....: .. __

() . ~ .. ,---- ........ --__._-~~.--'- .. -

~) Il.~ OA (1,(') ~J.~ I 12. 14 1./, I.H

"'of-I

Figure 2_5: Flow paths in the fountain region for a u:rt'l.in t.imo interval

For non-Newtonian viscosity models, the gapwise velocity profile doc~ not. correspond to the parabolic prof! Ie of a Newtonian fluid- Therofore, t he model of Bhattacharji and Savle can only be applied when IlO precise information is needed.

In U1S(eS where it is desirable to incorporate fountain effects II I< .. >n , a«curatcly, particle trajectories can he determined separately by ~olvinp; t.h(-) fountain flow problem gi ven the flow field at the line that t;(·parat.As LIlt' main flow domain from the fountain flow region- Moreover, such a procedure giws the opportunity to alter the shape of the flow front itself, In Uw most existing models, the flow front is considered as a straight line. In practice til(' How front. is of course more like a semi-circle (see (J-g_ Mavridis d al., 1986)_ Changing this flow front shape do('~, however, only affect the n,si(kTIC(~ t.imo in the front, since the relation between the ingoing and outgoing position is completely determined by a local mass balance. The final comparison of two different. fountain flow models is present.ed in Appendix A.

2.5 Multi-component flows

In multi-component. flow!; a. discontinuity ill material properties is present. at the interface between the components. Instead of modelling the i uVThw(" discontinuous functions can be used for the material pJ.-OP(,r(.i('R_ I n order to satisfy the conservation of mass, the glob3.1 conservation laws have to be adapted for incorporating the discontinuous functions,

Modelling aspects 25

2.5.1 Interface modelling

Consider a time-dependent volume V(t) that is divided into subvolumes Vk(t). The subvolumes correspond with parts of the total volume that have specific material properties, like density Ph and viscosity r]k. The local mass conservation yields (Muller, 1984)

fJpk ~ ~

Dt + V . (Pk'(lk) = 0,

(2.30)

and the local balance of momentum with discontinuities over the material interface is given by

(2.31)

For simplicity, the contribution of surface tension at the interfaces is neglected here.

2.5.2 Inverse mapping

The ultimate goal of simulation of the injection moulding process, is to predict the process conditions given the required product properties. In the case of multi-component injection moulding it is important that one can predict the injection sequence on beforehand, since for complex geometries it is almost impossible to do this by trial-and-error. The task to complete can be summarized as:

Given the particle distribution at the end of the filling stage and the definition of a required layer distribution, determine the injection configuration and seq uence necessary in order to attain that pre-set definition.

In our case, the particle distribution is given by the label field f

For one-component flow, like occurring in multi-colour injection, the label field contains all the information necessary to determine the injection configuration and the inverse mapping can be carried out straightforward (as a post-processing task). For multi-component flows, however, the rnaterial distribution influences the flow kinematics. In that. case the injection configuration has to be obtained it.eratively:

L Perform a filling simulation with one component.

2. Determine the injection sequence given the required material distribution.

26 (;,118])I;ll1" 2

3. J\'rform a multi-component filling simulation with the injection scquence determined under step 2 as input.

1. Repeat steps 2 and ~ until the injected and calculated inject.ion sequences correspond to a sufficient degree.

2.6 Piston driven flow

In ('hfl piston driven flow problem, which is drawn schematically in figllt'() 2.G, it fluid is advancing between two pistons in a tube. In contrast to t.he mould-

Figure 2.u: Schematic representation piston drivel) flow problem

ing processes dealt, with in the preceding sections, the piston driven How, with or without contraction, is modelled via the instat.ionary Navier-Stokes equations for incompressible fluids. These equations can be derived from the equations for conservation of mass (2.1) and balance of momentum (2.2) using Nrwton's constitutive equation (2.5),(2.6a) and read:

J.;] _- __ -

fl fJt + fl(V . \7)11 \7 . (J = pi,

V·'u :;., 0,

(J = -pI + ",(V'ii + (Vi!)").

(2.::12il.) (2.12h) (2.:r2c)

Substituting equation (2.32c) in equations (2.32:'1.) and (2.32b) and int.roduuing the dimensionless variables

_ 1·, I-_' pi _ L I

\7 = L \7 , u -,,:: V Ii , P = pV2' f = V2 f , t ,," oi', 'rJ = rIo 1]' ,

(233)

and omitting the accents yields

(2.34a) (2.31b)

Modelling a.~pects '27

where tltI" Reynolds number RI' is defined as Re = I2X..!:. and the Strouhal

. %

Humber Sr as .':IT = O~I with () being a characteristic time.

The boundary of the computational domain n is supposed to consist or two non-overlapping parts I'" and r" on which the velocity (essential boundary conditions) or the stress (natural boundary condition) can btl prescribed:

17= § (T -rl = ;;

on r.,

(2_.35) (2.:16)

On r.,

An initial condition at t = 0 has to be applied to complete the set of equations:

ii'(£,O) ""UO

in n-

(2.37)

The variables ij, ii, and 170 are arbitrary functions that depend on the problem considered.

Since the Reynolds number is typically 0(10-') - 10--~), the inertia term (iY - 'il)-v = 0(1) is negligible small compared to the term preceded by Rc-1 and therefore may be omitted.

The thus obtained instationary Stokes equations read as:

_ a·v 1 - - - - '<")'- - -(v· 1)\1)ii + \11';;;;: I,

at Rc

~·v = o.

(2.38a) (2_.38b)

with the initial condition (2.37) and boundary conditions (2.35), (2.36). Further simplification can be established by neglecting the time derivative. for the probkms encountered in this thesis this is justified, since S)' is negligible small compared with Rc:', The piston driven flow problem eit.her without or with the contraction arc: modelled llsing the thus obta,irwd instationary Stokes equations.

In non-isothermal flow situations, the Navier-Stokss equations are coupled via the viscosity with the energy equation (2.4) or (2.11):

pCpi' = (~. :\~)TI rr'/, (2_39)

where the shear rate is defined as'Y = V2DJ : D,i,

2.7 Conclusions

The balance and constitutive equations given in this chapter can be applied for simulating the conventional as well as the multi-component injcction moulding process. In the latter case, an extra conservation law is

28 CJlI1ptu· 2

added to solve the particle tracking problem. A st.rategy is giWII i.o mode] multi-component flows using different material properties for pa.rh COIll!H)IWll(. w hich results in an inkrf<1"'" represented by t!le discontinuity of I.h()se propcrt.ics. TIl I.lil-l next chapter. LlHl numerical aspects of solving those (·q!la.l.iol1s a.n· discussed.

Chapter 3

Numerical solution strategies

In tid" chapter, the numerical solution of the equations defined in Chapter 2 is discussed, Section 3.2 deals with the numerical methods used for solving the problems defined in multi-component injection moulding. Special attention will be paid to the method used for particle tracking and for the fmlf!taill flow phenomena. In Section 3.3 the numerical solution of the piston driven flow will be described. Here, the combined flow of a polymer melt in between two (extremely high viscous) pistons, will be emphasized. Finally, Section ,3.4 gives some concluding remarks.

The E:'xt.ensiOIl of the filling ~ta!,/~ of the conventional injection moulding process to multi-component injection moulding is implemented in Vlp, a package for the simulation of moulding processes developed at Eindhoven University of Technology. This program make use of the finite element package SEI'RAN (Segal, 1984).

3.1 Temporal discretization

To solve the sets of non-linear partial difTerentiaJ eql)a.tion~ derived in 38r:. tion 2'2 numerically, the problem has to he discretized in both space and time to get an approximate solution. For the temporal discretization the time domain T = [0, tend] is divided in Ttt. intervals according to

nt,

T = U /'H1, f,.,+1 = [/.n, tn+1], 6.tn+1 '" 1,,+1 - t«

n=O

(3l)

The time derivatives that occur in the equations for the pressure problem (2.15) and the temperature problem (2.19) are all material derivatives. The

29

:10 C}'''l'tUI 3

material derivative of all arbitrary variabk: {~ is formally defined by

I ht (~"r I' == lim (~(], II (J!::"t) n(ii', i;)

Vt I~O O!::,.t

where ,; denotes the position at time t of the partick that is located at. :r aL time t + !::"t. Equation (3-2) ca,11 hr.! approximated as:

(:\.:J)

where «11.+1 = n(Y, t. + ()!::,.t) and I~(' C Il{;;', I,). Now, it remains t.o d"t'TJrJiII" the convected variable (.t,. Consider a" '.~.:'f a(.;'(:if, /'" I I), t,,) as a fidd variable. This field (:i111 hI' obl.aincd by convecting the variable field a1 I til' (~n '~f n{or, t",) by th(~ know II vnlority Iiekl ',l(:i:, t). Hence, t.h« variable 0,. is dctcrmiucd by solving

DI'" -

- + 'I; - \7u =0 ()

ilt .. C ,

(34)

Using this method, the convection operator can be ~plit.(.,-,d [rom I,h(, f('iIt"jlldel' of the equation. Although it dillers from t.hn operator s]llittill('; (.cchnique, where the convection operator is part of a larger operator as appears in ('_g. the convection-diffusion equation or differential models ill vis(:()-(~lastic flow calculations (see c.g. Donea, 1991; Baaijens, 1992), tilt> procedure is 11)(' same. The procedure can be summarized as follows: The solu t.ion field at I _. !::"t is convected by il known velocity field. The outcome of j,bh (:<;.kl']atioll provides t.hr: init.ial c()T)di(i(l1l ior t.hp remaining part of tho original equation for I.lw current time step. The advantage of this method is that different numerical schemes call he used for solving different terms in the equa.li()ll~. Esp!'cia.iiy in the case of convection dominated problems this is ,I, gT'f-'ilt benr-lit., since t.h« classical schemes often show spurious oscillations.

3.2 Multi-component injection moulding 3.2.1 Spatial discretization

Th(~ sp;),ji"l disCH·t.i"Mi()1l is performed by using a finite elcinen: 'lIlJ'lho(/ Tho difl'crcnt. finite element. meshes used are depicted in figure ;;.1. TIll' pressure problem requires a finite clement mesh u,,~(, (cf)v<>rs I.h(· m idpl anc of Lhn prodll(:(. (figure ;}.1,;,)- The discretization if> performed usim; biquadratic quadrilateral clements defined in lR,~ (Segal, 1984) witb 0111, d~gr'(~n of freedom

Numerical solutioll strategies 31

a)

Figure :1.1: Spatial discretization in midplane (1)) and perpendicular to that (h)

... the pressure ... per node, Since realistic injection moulding products are thrE''' dimensionally shaped, the elements are situated in a 3D space,

The temperature problem is solved using one-dimensional finite clement meshes consisting of linear elements in the gapwise direction of the product (figure ;Ub), These l Dvmcshcs arc present in every vertex node (nodal point) of the 20 mesh of the pressure problem, which are named the gridlinss. The nodal points of the ID-meshes are called the gridpoints,

Finally, the convection problem . for determining the convected variables am! particle tracking - is solved on meshes of bilinear quadrilateral elements in lH? (Segal, 1984), These meshes arc formed by the gridplanes (figure 3.1h) which are the 2D planes spanned by the grid points.

Although the total product is discretizcd in three dimensions, the cornputation differs considerably from II complete 3D analysis, In II 3D computation the systems to he solved are normally very large, especially when temperature gradients ill the thickness direction demands for a fine discretization which usually is the case in injection moulding, The systems to he dealt with in the SO-Called 2}D approach are much smaller, since the gapwise and the midplane discretizations are decoupled, which is computationally much more efficient.

3.2.2

Pressure and temperature problem

Pressure Problem

Applying thl' spliU,ing technique of the c(Hlv(!dioll oporator and a11 illiplicil Euler time-stepping t.o (!q uat.ion (2,1;)) yields:

~. , (8,,+ I 'V"Pn+ I) + ii"n I ,J)" I I = fn-I-I,

whnro t.hn S11 bscript n -I- 1 denotes the new time step. The ('o(:Oi('i('111 ,.-" I I and the right-hand sid« In+ I are defined as:

In II

h!2

,J- ,,"n I I d(, 6.tn+ 1

. hj2

h/2

/ (~,,+ I (1;,+ I - 1;) .u; + hjl

"/2

6. t" I 1/ '~"n I 1 P. d( , --·"/2

CUi)

6.1'n I I

(:l.7)

in which T", fir· d(1notn tile convected temperature (wei pressure rcspcct.ivclv. Fquation (:1,;)) is it ~f~cond order ('lIiptie, differential equation fwd is sol vr«! Ilsing t.1I1! Bubnow-Calcrkin finite dement. IIldh().! with tho prcssur« as the Ill) !(IJ()WII. Ak I,h(' pressure is biquadratic per clement, th(1 pressure gradi(!nl, -<;]'p and thus the velocity 'ii', which is a function of -<;]'p (~('(' (·(1I1a(.jon (:2.lB), {l,r(! hilinuar per clement. Obviously, the velocity pa,ra.lloi t.o t.hc midplane 'Ii" and the pressure gradient. V·p have t.he same direction,

For calculating tlw integrals over the thickness, the grid is divided (II 'll.d'lrp parts, 11",1,,"1' = 2 for symmetrical problems and TLd!)rp == 3 for ,,~'y[llItli'tl'ici\,1 (.(!mrwm.t.ure boundary conditions, These pares coincide with the solid and fluid parts of the grid lines. The fluid layer is discretizcd w;ingnJ pidpnilli.s that form all oquidistant grid and th« solid layer is divided in t.ht: ~n.t\\(; Wily ill '/"I., Kridpoints. For convenience, double 'I()(h~~ are present at. Ill(' fluirlsolid interfaces, The int •. ,grab in Ow z' direction ill the pressure prohktrl :t.n! approximated by

h/~

I I(() a: eo:; ±= ''f' f,;1II)"

"1-1 ,"i~'l

(:Ul)

-·"(l

Numerical solution strategies 33

where 1)1' = n; + ti; or n/8 = n; + 2n, for symmetrical Or asymmetrical problems respectively. The weighting factors for the symmetrical case arc defined as:

{ h-2h+

n.1

1~J' . = ."-4,,1

~,~ n.,

o

if 1: = 1 or i = n; if 1 < oj < 1IJ otherwise

(3.9)

ifi=n,ori=nf8 ifnJ < 'i < n/s otherwise

(3.10)

and for the asymmetrical case as:

{!l::..=1l

n,

2 h--h

'((1[,= ~

, n.1

o

if'i = 1 or i = ft. + 1 if 1 < i < n, + 1 otherwise

if i = TIs + 1 or i = TIs + T!J if ti; + 1 < i < n. + nf otherwise

(3.12)

ifi = n., +nJ ori = Il;, ifn, + tit <i < nt. otherwise

(3.13)

Assembling the weighting vector 111 and calculating the dot product (w ' f) gives the approximation of the integral according to the trapezium rule. In case of calculating the fluidity, the weight factors in the solid layer, w~" for the symmetrical and l1h,i, 103,i for the asymmetrical case, are set to zero,

Temperature problem

Analogous to the pressure problem, equation (2.19) can be discretized with respect to time and reads

(:-1.14)

34 Chapter :i

where {/"+I and l.he right-hand side /,.,+1 an: given by

J/p. I I

( 8:) ) PHl:.!_~" pc] ,

aT /),7'.11 P,j··I··1

(:lI;"i)

= --;;:----11 [PH+ I (:1',"+ I + '-' 'n+ 1

. 2 (1,,+ 1 Cp."'11 'J;.

.1", I I = Tty --I- ...... -.-

6.1.71, II

(,1.IC)

Equa.t.ion (::1.14) is a second order elliptic oquation which is solved Ilsin)!, the Bubnow-Calcrkin finite dement method. The transport 01 heat is incorpor:il,!'d by Lito convective variable T;,.

Since the pressure- and Li'llIp('ril,tnn~ problems form a coupled, non-linear s('L of dilforcntial equations, they have: to bn solved iteratively. Mor.-ov«r I.he computntional domain cli,wges with time because of clIP OIovillg flow front. Th« propn,ga,ti()ll of the flow front is carri('d 0111. with the S()UlTI~ propap;nLioll method (Sil-l,('r",1988; Caspers ci ol. 1994). J)I this method, every poiul Or! tho flow front acts as it (point) sourco and fills a certain circular al'l',1 with radius Ll.:r = 'il7,6J Thr envelope of all these ciJ'(~II.'~ giw th., nnw flow front position. '1'1)(' nnw mesh is constructed by (:olln(~cting the intersect.ions of OIl' onvelope with the clement edge~ of a background mesh that, r:.oVIT' UH' cornplcto product. Only that part of the background ll.Il'sh \.1111.1 cOllt.itillS intersections wit.h tho flow front envelope hat; to b(' rellll'~h('d tompurary. the part that contains completely filled elements only, remains UlH:haJlfv~d.

'1'11(: «alcu la.tion of the filling stage i.~ executed in 1:1)(> following stops:

1, detcrrnin(' the How front position based on the volocit.y li(~ld of the: previous tim« step ami perform a local remcshirig of (.h(' ha.<:kgl'()llIl!1 "'[(:,,h.

2, map the solution field from the old mesh (previolls t.ime step) to the

mesb constructed in step 1.

:t determine the convected variables TO' and p.: 4. ~()I ve the P!,(\SS1ll'(' problem,

5, solve the Lcrn pural.u re Ilt'ob18IYL

r;, update the material function Ilsing the calculated pn~~Sljr"-, 1.<:IIII'Nat.ure-, a.lld xhca.r rate fields.

"1. rf'pNI,t step :'. to Ii until convergence is achieved.

8. p('flonn particle tracking.

9. gew~rfd,(~ output fol' post-processing.

10. update t.hc Lime seep if the product IS not, yet filled (:ompkt.<'ly and 1(\1.1l1'11 to stf'P I.

Numerical solution strategies 35

3.2.3 Convection problem

The particle tracking problem as modelled by the conservation of identity and th" splitting technique of the convection operator are based on the same instationary scalar convection equation. This partial differential equation is hyperbolic of nature and needs special treatment to avoid spurious oscillations in the solution (see e.g. Donea, 1991). The time derivative in equation (2.26) is approximated by a two-step O-method (van de Vosse, 1987). Application of this method to the equation

DCt

at = 1,

(3.17)

results in

(3.18)

with

1nH = 01n+l + (1- 8)1,,, t"H = tn + (J6.t.

(.3.19) (3.20)

The value at the new time level is given by

1 1 - e

(~n+l = (jun+o - .. - =r >

(321)

Here f) is a parameter in the interval 0 < 0 ::; 1- This scheme is unconditionally stable for 8 ;: 0.5. For () =0; 1 the scheme reduces to the Euler implicit method which has an accuracy of O(6.t). The well-known Crank. Nicholson scheme appears for i1 = 0 .. 5 which is O(6.t2) accurate, The two-step scheme can be considered as a Euler implicit step to time level n + 0 followed by an extrapolation to the required t.ime level tn+1•

Applying the O-method to equation (2.26) yields

(3.22)

followed by the extrapolation step according to equation (3.21). Because of its relatively high accuracy the Crank Nicholson (J = 0.5 scheme is chosen in the calculations.

The solution of equation (::1.22) is carried out with the Streamline Upwind Pctrov Calerkin (SUPG) finite element method (Brooks and Hughes, 1982), The SUPG method provides stable solutions in case of convection dominated

flows with discontinuities II) I.hCl solution 11:'; may occur in the particle tracking problem. However, mmn numerical dispersion may occur both dOWJl- and upstream of such 11 discontinuity. the performance of the ~d)(.:ltI!~ chosen is demonstrated in Appendix B.

The cOllved,i()Il equation i:-; solved on 11 finite elemeur mesh (:ollsisting of bilinear quadrilateral clements coinciding with thl) gritiplanc:,; that arc situaled in t.he fluid layer (lIgllce :~.lb), The gridpiall8H arc sliglu.ly curved in the direction of the flow (figure ;1.11» following the shapu of' the already mlidifi('d luyr-r. In this way it rough approximation of the (small) vdocity COI1lPOllCllt. v~\ i:-; csta.blishnd (without expliciLly calculating it), since tho gridpoinU; aro oqually divided hdw('!'Il the midplane nnd the solid/liquid interface,

3.2.4 Bifurcations of the midsurface

Whnn a product cOIl~isLs of thin walled sections with their normal vectors not pointing in the ~n.rnc direction (ab is usually the ca~c), bifurr.ation» of 1.11(' midsurfru:c ('xists. In nearly all existing simulation codes one prossur« value is present at thc intersection points of those branches but, of COUrI;(!, d. different prossur« gradient inuv (!xist for each separate branch, All oLlwr' properti(,s like e,g, tern perature arc equalized in the intersection points. In <:a:;(' or ('.g. T-.i unctions, whose' upsl.n-a.in part is connected with two downstrc.un parts, this [(·a.ds to all unrealistic distribution of properties of material part.ir.lcs. Couniot ct al. (199,1) showed numerically that, when deal! with COlT('C(,jy via. a proper splitting of Llw flow, an initially xyrnmctric gapwis\! 1,(1Jllp(:l'at,\ll(! rlist.ributiou in the ItpsLr!'am part of the connection results i n asyinmct.ricul distributions in the downstream pacts (figure 3,2) inst.ead of the: symmntrical distribution that is usually taken for granted.

FigllH) 3.2; Temperature distribution ill a T-jllndioll (after Couniot et. (1,[. (1993))

The origin of this erroneous distribution sterns from the fact that <:OIlS()l'vation of mass is only taken into account in the calculation of the: pressure

field when employing the 2~D approach. The mass balance is only weakly satisfied by imposing that at flow splittings the sum of the downstream fluxes equals the upstream outflow flux. In order to deal with these splittings properly, local conservation of mass has to be satisfied for all those properties that are dependent of the flow history (such as temperature, density, stresses, orientation, conversion, etc.) or for the particle identity. In the case of bifurcations this demands for a special treatment ..

The improved strategy in resolving the convection problem is to split the product geometry into parts that contain no bifurcations, but that can be considered as 2D flow geometries. At the intersections, the values at the inflow boundaries of the downstream branches are considered as inflow boundary conditions (Dirichlet condit.ion). Then, the convection problem is solved for each part separately, irrespective of the order of filling. After convection, t.he values at. t.he inflow boundaries of the downstream branches are updated using the calculated values at. the outflow branches of the upstream sections. Hereby, it is assumed that the time steps taken are small enough so that the material particles do not cross the opposite element. boundaries, which is in fact. the well known Courant-Friedrichs-Lewy (CFL) condition (Hirsch, 1990).

During updating the values at the downstream inflow boundaries, local mass conservation has to be satisfied. Suppose that there are n outflow boundaries and m inflow boundaries. For every inflow boundary i the fractional t.hickness };,J of the connected outflow boundanesj that fills the inflow part is determined by a local mass balance. Let fi.J = [1-, f+], then the fractional thickness can be det.ermined by solving

j+

./ V; (z)dz = q; 1-

(.3-23)

where 1);(Z) is the velocity profile of the outflow boundary j and qi the volume flux at the inflow boundary i. An additional relation has to formulated to determine the positions .r-,.f+ completely. For that purpose, the spat.ial sequence of the element groups around a connection node is taken into account. For example ill the case of two inflow nodes connected with one outflow node (see figure :3.2) hI = [-h/2,,t+l and 12.1 = [J+,h/2]. Whet.her a connection curve is a inflow or outflow boundary is checked by evaluating the dot product (V*p. fi) with ii the outward normal vector parallel to the midplane-

The gapwise values of any property to be convected can be distributed over the connected downstream points by simply mapping the values in the

38 Clutpter 3

fractional thicknesses of the upstream points to (hp, inflow gridlin::s. I n this Wily, lo(:a,l conservation of mass is only weakly satisfied. A fnrj.hnr improvemont is to preserve mass conservation by evaluating

;;:'/1: .~

./ v;;(() .u; "" l'IJ~(() d(

k: = 2, .. , nI

(.3.24)

when' 'u;, 'u~ i~T"~ I,hn gapwisc velocities at. the down- or upstream nodr- f('~p,,(:I,ivnly and s the gapwisc position ai, which the value ha~ (,0 ill' d('l.cl'lllincd. This latter mr-thod can also be applied when :\II abrupt change of tbi(~kI"~~S ~x ist <lot, the interconnection between two parts. This could Iw ll;-;dlll when tho shape of the velocity profile is expected to dl<\!)~(' much with in- OJ' d('creasing flow rate, ("g. whr-n IIsing shear-thinning liquids or materials with a yield stress.

3.2.5

Fountain flow

In the thin film or lubrication approximation the fountain rq.\iOll is l(~dll(:(~d to a front perpendicular to tho midplane, Since in th,~ midplane no pressure !!:r:1,diput normal to the flow direction exists, the front. call 1)(' further sim plifir«! to il. single straight lin« ill every node situated Oil tilt' mi,lt. front. When (1<>li[lilli'. I.he folding point as the position where the velocity equals t.11" aV('I,lg(' front velocity, the front. line can I)(! divided in two n'l:)i'm~: ~ridp()int.:-i that arc situated between th<.' f()ldiul:\ points and those that all> bNw(~(!n t.he folding poi nts and the walls. Only the latter gridpoint.s have experienced 1-11" ri mntain ('ffm~t., the others arc entering t.he front. region. TIl(' parillll('t(1l'iz(~d front model giv(~~ ("(II' t.hoso points the: residence time in t.h" front and the "lltmll('(~ position.

How to incorporate the Iount.ain efrl'r:t depends on whether UI(' l<'si<il1I1Ce time is included or not., III hoth cases particle tuwki II)!, is carrier! out first ill the whole computational domain before applying the front model t.o (.11(, gridpointi> situated at the How front. The gridpoints hdwe(~lI th(~ folding points gd their values naturally by solving thl' particle tracking problem in the fluid. Vi/hen l1('gkd:illg t.he residence time, a material part.icl« a.t it certain gridpoint between the folding point. and the wall originates from a position Z, between the folding points on the front line, The position z, is given by t.lH~ parameterized model or by a local mass balance according to cquatiun 2.27.

In case of incorporation of t.h« residence time, 111<:, VnhI('8 cannot b(~ taken from th« front line dired.ly anymore, hut have tv f_", fOlllid at. a spatial position :/:/1', till! position of front passage Up). This position corresponds wil.h the place where a certain part.icl« that leaves the front ht t.he 0.(:1.11".1 time i enters

Numeilcel solution strategies 39

the front region at a time t!p = t - t"f, with r! the residence time in the front given by the parameterized model. The remaining problem for the field variable Ct considered is to determine a (.if'P , tIp) of Ct. For this purpose another field variable Ctll! is introduced that contains the values at the nodes that are just filled during the past time step. In the same way, the velocity at front passage iijp{z) is stored. Given the position Zo where a particle leaves the front, the residence time tT f and the position Zj are known. A gridpoint between the folding point, and the wall can now be filled with a value taken from a position Zi On the front line calculated by solving

for t E [tIp, t]

(325)

The solution of this convection equation is carried out in the same way as equation (3.22).

3.2.6 Interface modelling

The description of a discontinuity in the material properties, e.g. 'I} Or p (thus the injection of more than one component), in a spatial discretization problem is only possible if the material interfaces coincide with element boundaries. This implies that a suitable rerneshing technique has to be implemented in order to follow interfaces between the two interfaces in time, Mavridis et at. (1987) used double nodes at the interfaces to model the discontinuity in the pressure field that arises when surface tension cannot be neglected or with viscosities bsing different. In that case two different values for the variables (iJ and p) are allowed 011 two nodes at the same spatial position. Such a method is relatively expensive since remeshing is needed at each interface,

In this study, three different type of interfaces can distinguished, First, there exists an interface between the polymer melt and the air during the filling In the simulation the possible influence of the air is neglect.ed and the front flow (interface) position is determined by local remeshing of the background IIIe~h, The second type of interface is the one between the solid and the liquid layer, occurring in non-isothermal filling problems. This solid/Iiquid interface is modelled using the above mentioned double nodes at the melt-liquid interfacefs). Iiemeshing has to be carried out. on ID meshes only (in the thickness direction), so not much computational effort is involved. The last type arc the interfaces between two different materials that occur in multi-component injection moulding (or in the multi-material piston driven flow problems (see Chapter 5), These are modelled via a discontinuity in the material properties that are approximated by continuous functions with a steep gradient. The maximum steepness is controlled by the local

40 Cbnpecr ;;

mesh ~i7,(~ and the order of the element- H if) assumed that the smoothing of discontinuities only has ,], local dff'd, on tho final solution. The main advantage of this approach compared to rcmcshing techniques is that. it can lw used easily in all kind of (complex) geometries in 2D or 3D EXl\lllpks of transient flow problems like) mould filling, whore tho f'n~(l surface is modelled by tho transition from il high (polymer) viscosity to a low (air) vi::;c')'>;it_y call be found in Thompson (lntl6); Xic et (.Ii. (1991); Akkerman (199:;) Other examples concern (,il(' irll,pr'faC(-l between a reactive fluid and the already p_(llkd material (Potcrs ct al., 199:3)_

In the latter interface modelling \',dllllqUC, UIP position of the moving int.erf;tc({~) can be ,ktcrllliued by using particle tracking. In accordanvo with tho method presentnd in Section 2_:-1 an initial ~ct of lal)('\~ ddillinp, UI(' pa.rtide identity it; transported through the flow dorna.in , thus roprescnt.ing the rtlaj,('rial dist.ribution. By doing so, the material properties becoJI)(1 also a function of the label values. The strategy is also I.(s(,,1 1,0 modi'l 1,],(' piston driven (contrnction] Ur.,w prn],I"'T1L Tho pistons are represented by an artificial high viscosity material which slips at the walls of the domaiTl. This so-called fixed domain approach call be (:ol)(lid(~rrd a,s an variation of the vr-rxion oj UUl pS('lldn.(:ollcNltratioll method as used by Thompson (1986)-

3.3 Model flow problems

The problems descibcd in this section refer to the piston ,hi vNI flows (with or without a contraction] that an' used t.o t('st (,Ill' pi\,rticl(~ tracking (s('c chapkr 5)- Since thes!' model flows are much simpler than the rnult.i-compoucut injection moulding problem (2D isothermal, rather than 2tD nnn-isof.htumai ), il. simplified and fast test code has been written based 011 the Iinitc ('i(,llWllt p<v.:kagc :':>l:;l'RAN (S('g<\l, 1(84)-

3.3.1 Discretization

The Stokes equations (2.38) are wived using ~, standard Galerkin finite olemunt IlIdho,L The ~p~I,t.ial di8(T('t.iza,t.i(Hl is performed with quadratic isoparametric triangular elements in IR~ (Segal, 1984). The velocity COTIlj)OlI('lIt.S are the unknowns in all nodal points of the clements whereas r.he pressuro and the pressure gradient <Ire only known ill the center node.

After spatial discretization of equation (2.38) ami applyillg the Gal(!rkill r()J-mlllat.ioll the Iollowing Hrt. non-linear differential equations is derived (Cu-

Numerical solution strategies 41

velisr et al., 1986)

By T

M at + S(yly + L p = f + b,

Lv= 0,

(3.26a) (:U6b)

with M the mass matrix, S( v) the stress matrix, LT'p the pressure gradient, and Lv denotes the divergence of the velocity. The right-hand side is formed by the body and boundary forces f and b. The unknowns u and p are columns containing the velocity and pressure in the interpolation points. The incompressibility constraint is diseretized with a penalty function method (Cuvelier ei al.. 1986). The balance of mass is perturbed with a small parameter e times the pressure:

(3.27)

The matrix M, is called the pressure mass matrix. The penalty parameter e must be chosen such that f.P = 0(10-6). By substituting equation (3.27) into equation (3.26a) the pressure is eliminated from the momentum balance by which means the momentum equations are no longer coupled with the continuity equation. Hence, the velocity is considered as the unknown and the pressure can be calculated afterwards. ThE' penalty function method reads:

Ov 1

M--;;;- + [Sty) + _LTM;lL]Y = f + b,

ul. e

1 1

p'-' -M- Lv. f P

(3.28a) (3.28h)

The time integration is performed with the two-step 8-method already described in section 3.2.3 . The first step is given by:

lIT _I 1

[O~tM -I Sly,,) + -;;L Mp L]Yt>+8 = BAtMv" + fn+8 + b,,+8,

1 -1

P,,+o = -Mp LV".j.o· e

(3.29a) (3.29b)

The extrapolation step is performed according to equation (3.21). It should be noted that ill the case of time-dependent boundary conditions, they have to be prescribed at. the time level t,,+o'

In the case of generalized Newtonian fluid behaviour, a linearization of the stress must be applied. For values of the power-law index of 0 < n < 1 Picard iteration (successive substitution) can be used.

For the solution of the particle tracking problern , the Streamline Upwind Pctrov Galerkin finite ('krnf'rl(, rII(>thod is applied using the samr mesh n~ ill the Stokes problem. The time integration is <:i1,II'[()d out with an Euler im pl] .. ,il, ~dl("'''' (0 I).

3.3.2 Piston driven flow

The W~Olll(~try for the piston driven How problem i~ giv'>r1 in liglll'(' :1.:1 The boundary condit.ions t.ha]. dt>firH,> the piston driven flow problr-rn (axi-

Ii

A

J r,

~--r---l" -.".-_- .. -_-_.-._-,-_-_-._-.-_-._-._-,~_~._-_.'-_-._-._---._-_--,I :' ..

1', .~ I

:l:

symmetric] arc:

'/I .- il on 1'1
(u',1) () on r,
'11 () on T':I
iJ iii) OIl [,1. (:uo)

Instead of moving the pistons, the cylinder is movorl with 'uo '"", l , ... Ii,,, or/' while the pistons (1re kept 11x('(I.

111 (.II(' point A, ~1r)d H Uti' boundary conditions am ill confiict. The points could 1)1' considered as a regular wall points or a~ belonging t.o !.lIe pi~I,ort. Following Vos ci ol. (1091), tIl<' point. at (.h(' driving piston (B) is (ollsid"rd t.o be part of the wall while at the driven piston (point. i\) SOltW I(~akai!,(' is lnl,n)dllc(,d hy a:>sigulnl,( t.hr: point to the piston. However, when applying proper mesh refinement ill this point, the kklk,lge will Bot inflmmcc the ovcral flow kinematics.

3.3.3 Piston driven contraction flow

This flow situation call be seen a~ a simplified version of tfl(~ problem of '.'!lIptying all accumulator into a mould, IiJ((~ ;-I,ppli(·>d in transfer moulding prou\ssns Tho flow domain is shown in figure ;t1 (only one half is c()f)sitkr(~d because of symmetry).

Numel-ical solution strategies 43

For this problem the flow domain changes in time. A possible way of solving this problem is to remesh the flow domain every time step with e.g. a compression-expansion scheme as advocated by Ding et al. (1990). In this scheme the mesh is compressed in front of the driving piston and expanded in the region in between the contraction and the driven piston. To circumvent. the problem of rsmeshing every time step, the fixed domain method is employed. As explained in this method the pistons arc represented by an artificial high viscosity material that slips at the walls. The shaded regions

fa

x

Figure 3,4, Piston driven flow through {I contraction: flow domain

represent the pistons which move synchronously with the velocity Vp. The viscosity of the piston material is chosen a factor 103 higher than t.he polymer viscosity. At the entrance boundary fl and the exit boundary r5 the material is supposed to flow continuously with the same flow rate to take care of mass conservation. The interface positions arc tracked by the label field initially defined as the axial coordinate .T. Material with .T ::; l:pl and x 2: Xp2 is defined as piston material (steel). Obviously, the mesh has to he large enough to represent the t.otal area that will be occupied by the melt during the experiment. The boundary conditions for this problem are given by:

V Vo on r.,
iJ Va on r2,
(v·n) () on 1'3,
v iio on r., (3.31)
11 110 on 1\,
i] - 'vo OIl f6,
11 Ei on 1'7,
v 'Vo on r; with i70 = [Vp Orr. Note that the boundaries f2, 1'3, 1'1, r6, 1'7, and 1'8 change every time step.

The interface conditions, at the wall as well as in the fluid, can be dealt with in several ways. Numerical studies (Oosterling, 1994) revealed that the viscosity jump can best be approximated by a linear interpolation within the biquadratic element. For the stick/slip transition at the wall the appoximation used is dependent of the position on the interface within the element.

44 Ch(l.pt('l" :I

Whml tho interface is situated between the first vertex m,de and thr- midsid« node of the clement, t.h(~ boundary condition at. these poi nts arc prescribed (~qlln,] to the piston velocity 'V() whereas a zero velocity is pn~~(:I'ib~d ilt the: other W'rLpx nodA at the wall, In the case that the piston has pa~'wd t.hr. midsidf) node, in all three nodes of tbe eI(~rnf)nt which arc situated al()llg the wall, a pII;(,(»1 v('locity is prescribed, To avoid a negative velocity in ('h(~ Intngrar.ion [Joints of the adjar:ell(, downstream clement, at (Jj(~ midside node of that element it velucity value is assigned such that UI(' interpolation polvuomial r(~nHl,ins positive within the' c:klIlI'll("

For this problem, two possiLI(' ways of tracking t.h(~ J[I\'I!['facc GUI he applied. F'irxl, t.he interface call be dcV'mli)l,~d by using particle idl'nt,it,i(~s as material markers, 1\ i;(~(\md, I('ss gr))cl'al, Wily is to caklllnU~ the exact positions of the int('rf;v:(' lIsirlg the (known) piston veiociLi()s. Tb(, latter Hlc(,))od ;j,ppP;\'I'(,d to givo the best results (~(,e Uosl,nriing, l~)~H) when COlllbiri!'d with it, label correction dllcidaL(~(1 in Ulr. following, In th« InOI'l' W'lI('ral approach, the intcriac«, ('sP('c!>lJly ,).1 the driving piston, starts to deform awl d,w I,ll the interaction between t1H~ hdwl field and the velocity li"ld, 1,11(· dd<'Jl'tnat,ioll grows with In,rw~r piston displacements. Applicatioll of a higher viscosity docs not give tho (k~in'd improvement and, moreover, n~.~lIlts in ii, d(,crcasing accuracy, since th8 penalty parameter (e < 0(10 1") reaches UI(' lIlachin(' accuracy. Since the identity labels ij,ri' employed for visualixat.ion 01' t.hr: deIormat.ion patterns in the fluid as well as the piston interfaces, a ('()IT(~di()Il or the label field at the driving piston proved to be llC'CC~NalY Tho 1,1[)(,1 v"hw~ in th(! f(~!l:iOllfi occupied by the driving piston are updated every (illl(~ step with their exact values. l n this way, the interface rnmainr-d str,l,ight and did not influence the dcformar.ion pattern in the fluid,

3.4 Conclusions

The cquat.ions dCrlv(,d in Chapter 2 arc cii1KTdized III both spaC(~ and t.iin« and the appropriate boundary conditions arc defined. Splittil)g or th(~ COIlv(~dioll operator from t.h« equations it> applied to he able of calculating thl' convor.tivo transport with a suitable solution method, The halallc(~ of mass b :-;at):,;fjpd loeally ill !'hi' 2~J) approach, by introdur.ing a special trTatrxl(~)I(, of flow splittings, The int(:l'faccs that, occur ill Ul(~ mould filling simulaf.ionx um lH' modelled with either a local relllr.~hing technique or with a. jump of the material properties at the interfaces. The latter techniqu(' requires tho material distributions to Ii" known in the flow dornain, which can be estublishnd with particle tracking, The piston driven contraction flow i" !lI()drllnd using the fixed domain approach.

Chapter 4

Material characterization

4.1 Introduction

In this chapter the material properties of the three polymers, polystyrene (PS, Styron 678E from DOW Chemical), acrylonitrile-butadiene-styrene (ABS, Ronfalin FX-50 from DSM), and polyamide (PA66, Zytal 101 L NC-I0 from DuPont), are presented and appropriate model parameters are determined, Since the rheological behaviour of a polymer melt has a significant. influence on the deformation patterns in the flow situations investigated, the emphasis is put on the determination of the rheological (model) parameters.

4.2 Rheological characterization

The rheological data were measured using small amplitude oscillatory shear experiments on a. Rheometries Dynamic Spectrometer RDS-II using the parallel plate geometry, In these type of experiments, a shear strain expressed by

( 4,1)

is imposed upon the sample, where 'Yo is the amplitude of the oscillation and i2 = -1 in the complex representation, The responding shear stress T will be sinusoidal with the frequency w, but. wit.h a different amplitude TO and phase shift 6:

( 42)

The amplitude of the stress TO can be written as 'YOGd where Gd is the amplitude of the complex shear modulus C·. The dynamic modulus Cd and the

45

4(; CJmpter 1

t.an S

(~~fI

.,

(4.:1)

where the storage modulus (;" and the 1()1;~ modllius Gil ,1]'(: the ITal alld imaginary part of C' respectively. Introdllcing a complex viscosit.y I/' dditwd by

'/.

'1/* .,- - -C' = r/ .. - 'I, 'II", w

(,II

, 7

'I) "" =:', w

If (7'

'I] =w'

(4.4)

t.hn dynamic: data can lw c:ollv(~d(~d to a. flow curve by ('lIlploying th« ompiral Cox-Mer» rule:

(,1.5 )

which sl"I(,('S t.hat thc steady state shear viscosity at a ccrl,il,ill xlu-ar raw equals the dynamic vi~(:()sity 7/,1 at an angular h'('(jlI!HICY equal to t he ,~ltl:nl' rate,

Using the RDS-IJ apparatus, th« frequency rarlW~ t.hM: can be imposed Oil the Hample is 10-1 :::; w ,~ GOO rod] s. Il(lw"vrl', the frcqucnrv ("UI b(·~ extended over many decades by ernploying the t.imc-tcrnporat.urc supcrposition principle" Assnming therrno-rhcologically simple material h<·haviour, tile lXJ.(·<~s\Ired isothermal data (Crt and 6) may be shifted onto a. singk maRtercurvc. Since the loss angle () is not influenced by t.h« temperature, al)d therefore does not need a vertical shift, the horizontal shift radon, (//1' can be dd('nnined properly using 6(w). Using the shift factors thus obtained to shift. the dynamic mod u Ius horizontally, an addi tional vertical shift. !'acto!' Ip/, Call t.hen be determined by shifting the modulus vertically to form cl. mastercurvc, III this way, both shift factors are d~terminr:d indcpendenr.ly. Tho mast.ercurves can b(~ expressed as:

8(waT, To) = o(w, T) I

C,,(W(1.,., To) = -G(J(w, T) liT

(4.6) (47)

with the horizontal shift factor defined ,,1:' aT = u.'/'(T, Iii) h b,I'U', 7~). For amorphous polymers the horizontal usually fitted with the Wl.Fvequation (FelTY, 1980):

and analogously shift. factors an'

III .. (T T. ) -al (7.' - 1u)

log (l'T • II =. _:_'---,-.,----"--'-

(/2 + J - To

CUi)

Material cilaractel'ization 47

and sometimes (e.g. close to and below Til) with an Arhenius t.ype of equation:

(1.9)

where E'A is the activation energy and He the universal gas constant. Note that. in the 7-constant Cross model the shift factor is modelled by

I . (" .. 1') __ -_cl--'--(T=-----=T.='-o)

n(LT ,0 -

c2+T-To

with (:1 = alln -I (10) and a2 = C2. The vertical shift. factors represent. the influence of the density on the modulus and might be modelled by a tern perat ure-densi ty correction:

(4.10)

I (1.' ,.,.,) _ peTe )7' ,~o - pT

where the index a represents the values at the reference temperature To and the density p call be evaluated from the pvT-data.

(4.11)

4.2.1 Polystyrene

The sample dimensions for the parallel plate geometry are; 025 x 1.5 mni. The data are measured as a function of the frequency (w E [10 1, ,500] rad] s) at different temperatures between 393 K and 538 K. The isothermal curves are shifted onto a single rnastercurve at a reference temperature of To = 462 K. The resulting curves for the 1088 angle and dynamic modulus of PS are shown in figure 4.1 and 1.2 respectively. The horizontal shift factors are plotted as a function of temperature in figure 4.3a which also shows the WLFfit and its parameters. The vertical shift. factors and the temperature-density correction are given in figure 4.3b.

From the dynamic data, the viscosity is calculated and fitted with a Cross model equation (2.7). The zero-shear rate viscosity '1)0 is chosen to be a function of the pressure and temperature:

-~I (T-:r::J lJO(p, T) = He <,+I-T- ,

The experimental data and fit are plotted in figure 4.4 for three different, temperatures, The parameters used in the 7-constant Cross model arc listed in table 4.1. The viscosity curve can roughly be divided into two separate regions: a plateau zone revealing Newtonian behaviour and a shear thinning zone. For high shear rates the slope of the curve is not constant but the

48 Chil.pt",· 4
flO
SO
70
GO
,)0
I
'"'0 40 -
30
20
10
0
10-> 393}f o 402 ic 113K 0 423J( »: ;133 J( d

443 J( • 45:3 J( "

1621{ + 483]{ 0 Goa [{ , 538}( d

L. __ ---'-- __ ---'-- __ ---'- _--'-- __ --'-- __ --'- ..... , ---'-------

1(1-1 10" 101 1O~ w:l lO" \0" III" lO7 I,J ['rail;',]

l;'i)l;lln~ 4.1: Loss angle as a function of angular frequency at different tcmp ernt.ures and the uiastercurvc ,\'1, 'l'u '" 4112 J( [()I' PS

l(f -- "l"
"
o o
lO6 ,,"
0°¢. ",+TcOD
09 ++ 0 x
10:; "~~~~~991lf1~~Xl:!""
oDD~x~a.6i
~ D:.:x=~~.~ ~ ~
~. XX"Q,.a !IIE¢"O "
>;: ~ )KI!I:-Q-
G " ~ c
101 .. • <>
• ",I: 6
• "<1 m ~
O,to+ 0 x
¢ ,/' CJ
,+ x
10:; + x
+
" + *~ ++.t

-*-W 3il.> K 6 ~O'2 K • 413 tc Cl 42.'IK x 133K '" 443K • 153 f{ " 462 J( ,.

483 K rt 503 tc x ::;38 J( ..

Figure 1.2: Dynamic modulus as a function of angular frequency ill (iilrerent tornperatures and the mastereurve at To = 462 K for I'S

lOl

_ _j_ __ -'------_ ... L __ ~ __ _j

102 10) 101 10" 106 107

w l{"(,dl.']

viscosity tends to reach it 8erond plateau. The Cross model ClIlplOVi,d docs not t.akc this behaviour into account.

Matcrlv,J cbaracterization 49

105
lO1
103
T 10'
'" 101
OJ
lOn 1.2

"',-~---.--------r-------. , - fI correction - experimental "

WLF-fit - experimental 0

L]

({j = 4.54 HI-I a~ '" 15(),4 }( 1"0 "" 162 j( 10-'

350 400 4~O

T[KJ

0.9 To = 10~ f(

fir} = 980,2 k_qm-~

0.8 L- _ _L__---'------_----'---- _ _j

350

500

550

(a) horizontal shift factors

(h) vertical shift factors

Figure 4.3; Shin factors against temperature for PS

500

550

For PS the glass transition temperature is considered to be pressure dependent:

( 4.13)

where Tqo = 373 I< and S" = 5.1 . 10-7 I</ Fa. In the numerical simulations the Tq is used as the no-flow temperature.

443 [K] c 473 [K] • 503 [KJ 0

10' 103 w [radl.;]

10~

Figure 4.4: Viscosity curve of PS at three different temperatures (solid lines: fit, symbols; measured data)

D's'''St~r<Jn 678E _J

;==.. ... -

n 0.2.")20

i-------t-- -.------:-

T'" Pa ~1.080· 10-1

II Pas 1.76.1010

r-------

'Ii! J( 373

f----- -.-----'1--------1

(:1 25.74

1--'"

r:~ K 61.06

1---,-'

51' K/Pa 5.1·10 7

.....-

Table 4.[: Parameters in 7-constant Cross model

4.2.2 Acrylontrile-butadiene-styrene

Th« (dynamic) shear dnth fOJ" A BS arc determined following the sam« proreduro as described before for PS. The mastercurves at a rdcl"l.")'''' tnmp('rclJ.lH·c ()j'T -: .. 473 1\ iln~ constructed usillg isothermal curves (:ovNing t.he temperature range T E [4.33, .")23] K, Figure 4.5 shows the determined horizontal shill fa.dol's and the fitted WLF-pa.l'arnllt.ers. The vert.iral shift. filct.()r.~ br :1.1'(, taken equal to 1. 1'11(' results for the loss I.tuglc and the dynamic nmdulus are shown in figllJ'(l ·1.G and 4.7 re:jpecL\v~ly.

10" ,--------.- 1()'1

10:1

11.12

WLF-f,~ - eXjJ8dm('nt.a[ 0

l()l_~

[0° (11 = 306

]() I il2 = 119.7 J( _ To = 4731<

10-2 10()

450

500

i:Jt;O

T [1\J

Figure 4.5: Horizontal ,,!tift factor n.gain~t. temperature for A BS

Material c11aracterizatioll 51

90
80
70
60
::;0
I
<-0 40
30 .. ..
20
10
PO-2 433 J{ ¢ 443 J( + 453 j( Cl 463K ~ 473K " 483 f{ • 493K c 503 J{ + 513K 0 523K x

Cl ,

10-1

101 102

W [J"tl.d/8]

Figure 4.6: Loss angle as a function of angular frequency at different temperatures and the mastercurve at To = 473 J{ for AllS

)06 ,--------,--------,-----,------,---,-------,--------,

lOr,
<l
~ 10-
"
C!l
10'1 433K " 443K • 453 j( o 163 J( x 473K .. 483]( • 493K o 503K ~ 5J3]( o 523K x

10~L-----J_-----L----~------L-----~-----L-----J

ID-~ 10-1 100 101 102 101 10· 10~

w [rad/s]

Figure 1.7: Dynamic modulus as Ii function of angular frequency at different tompcratures and the mastcrcurve at To = 473 K for AllS

The dynamic viscosity is plotted for three different temperatures in figure 4.8. From this figure it can be deduced that ABS does not expose a Newtonian plateau region, but the viscosity rises with a decreasing frequency.

52 ClJaptIC" 4

T------,-----,-' .. " data ,~t.'j, 1 d,lI,h s~t 2 44~ I< " 473 f{ 503 [{ "

'" 101
oj
~
'c
.,. 10:l '-- __ -'-- __ 1[_.. .. _j, __ --, __ __,c_

JOn lOl 102 10J

w [will,']

]()I 10-2

10 .. 1

FiR"]''' 4,~, Viscosity curve of A BS at three different temperatures (solid linos: fit, symbols: "",a."'Hl<1 'hl.a)

This observation p'.lipt.s to the existence of <I yiflld stress, Thr»«. linditll!:s arc in n,gr('(>tnl>nt. with those of Miinstsdt (1981) who xhowed that ill t.h« low shear rate region the viscosity can be influenced r.onsidcrably by illcn:asinp; the rubber ,,(llItI:))t of ABS, The presence of it yield ~t.n:c;s can be shown morn convincing when plot.ting the viscosity as a fllncLioll of the shear sl,["{'ss T (S(:C ligllrp 49). To model the speeifi(: shear properties of this AilS, tlw Cross model is not, suil.abl« unless it is extended to incorporate ;t Bingham-lih' bdmvj,,,H'. In il, paper OIl the existence and nature 01 ,"" yield stress EWI,IlS (j 5H12) proposed a g,'II(,rali'Ml,tion of the Cros» model:

,,'r.!.~ r/"", = I + /.::/("1

"10 - ")ex>

I n orcin!' to incorporate a, temperature independent yield stress hut. U:IlIlH'l'ature dependent. viscosity, equation (4,14) is extended alld rewritten in th,,: following form:

j (A' )-1

j ('I' ,,) -)' (T)' + '2"1

) ,~ -)0 1+ (AI'1)'" '

wit.h

(4,1:»

( 4.1G)

Material cilaracterization 53

10fi
10"
11)1
'"
<j
~
-e 101
.,.
102
T" = 760 PG
.101
102 r-----T"""T-----,,-----,------ .. "'1-~--------,

473 J( c 'dynamic' yidd stress - measured ywl(j stress -. --

10"

Figure 4.9: Viscosity as a function of shear stress for ABS at T = 473 K

For "r ---> 0 equation (4_15) reduces to it model for it Bingham liquid

(4.17}

Equation (4.17} reveals that the yield stress "v :::; H / fj~ is temperature independent. The results of the fit on the dynamic viscosity data are shown in figure 4.8. From this fit the yield stress value can be derived and equals Ty = 760 Pa. This value is considerably lower than the value reported by Miinstcdt (1981), who found that T11 = 3.103 - 101 Fa for ABS types with a rubber contents between 30% and 40% which is the case for the Ronfalin FX-50 grade, However, a material with a yield stress may not be expected to be thcrmo-rhcologically simple for low frequencies or shear rates. Therefore a yield stress determined from dynamic data is unreliable Additional measurements with a constant-stress rheometer at DSM R.esearch yielded a yield stress with a value 1875 ::; Tv ::; 3125 Fa. From these results the averaged value of Ty = 2500 Pa is taken for further investigation- The viscosity curve calculated with the latter value is also plotted in figure 4.8. All final parameter values for the generalized Cross model used for ARS art) given in table 4.2. The data set 1 is the result of the fit of the dynamic data, whereas data set. 2 incorporates the measured yield stress.

Rot 1 ~ct 2
711 0.79 0.79
Ii ru,,~ 9.;) . in:) ~),5 . 10:1
1() K 47,,) ,17.3
- .. ,.
rl 701 7.04
(:~ j( 110.7 IH).7
1----.--- ..
13, ~ 0.11 O.ll
D2 .~ 12,[, :3.8
'--_.-._ .. _. 4,2_3 Polyamide

The data I'or polyamide an: taken from (Peters ct al., 1994). TIt(: viscosity IS modelkxi using a Truncu«t j'()WOI' Law model:

(41 I))

The vi,~c()~it.Y ;'1." II. function of shear rate 1'01' three different 1''''''I'',J<It.l1n:s is shown ill figure 'l.lO.

10"

.-,------;---,-. . l" I I.

593 [1\J -., . . 573 [HI ' 55;ll!\'

L .. __ L___j___l__Ll.....LL_

Fip;llJ"n 4.10: Viscosity as a function of SIl(~lH rate and temperature f()r PA(;o

The model parameters are listed in table 4.3.

[ PA66, Zytd 101 L NC-lO I

n

0.61

1no Po, s :t455· 104

____ "."~.~ ~"m""" .... _

~ S (L147

----------::-----l !~ g-l 1.175. l(r'z

Table 4.3~ Parameters truncated power law model

4.3 pvT-data and thermal properties

The data for the specification of the density and the thermal properties are a.1I taken from literature. Since the AI3S material is [lot used in a simulation of the injection moulding proces, the puT-data need not to be modelled and a constant value of p = 945 kgm-3 (at T = 473 K) is taken instead.

4_3_1 Polystyrene

The pi/T -bchaviour for the rs material was measured by Flaman (1990). The data are fitted using the Tait equation (2.9). The model parameters are given in table 4.4.

PS, Styron 678E

.... _.
melt. glass
_ ..
ao m3/kg 9.72 . lO-4 ~L72 . 1O-~
al mO/(kg]{) 5.44 . 10-7 2.21.10-7
Eo Pa 2.53 . lOB ;'l. 5;~ . lOS
81 ](-1 4.08.10-3 3.00.10-3
'l~(0) tc 373
s Klra. 5.1 . 10-7
'-,,-- .. - .... ~. Table 4.4: Parameters in Tait-cquation

T}H: thermal conductivity is taken to be constant in the solid as well as ill the melt with a value of ,\ = 0.17 W/(rnJ{) (sec Flaman, 1990, chapter 4). The specific heat capacity (~p is modelled using constant values in t:'itller the molten or solid phase. For T ::; Tq the value is (:J' = 1785 .J I (kg I<) and (:" = 2289 J /(kg K) for temperatures above Tg.

4.3.2 Polyamide

The spccificat.ious of the nylon wer« again taken from (I'(~t(m; ct al., )994). Duo (0 lack of mow precise data of the dcnClity and the thermal propert.ies, ,,,,,)stant. values an~ assumed in either the molten or wild pha.s(l. The p:ll"a.ml'i.l'l" values are summarized in table 4.5

melt.

1------. ... . .... --,--+---+---

() kgm-' 950 1110

.----+----1

), W/(m]{) O):34f3 0.27

r-----.... ..-

r]' J/(kgK) 2897 2/132

Furthermore, a no-flow t('nlpmat\ln~ of T :.:: ~~07 f( was used ill I.h(· calculations.

4.4 Conclusions

Concluding thi" l1I;l.l,(~rin.1 characterization chapter we can sl.a,U~ t.h.u PS is t,I1(~ most extensively dpscriiJnd material (with st(~iI,dy stak ViS(".Ohity dati, ii1.('I'd by til!' 7-COIlStilllt Cross model and density (i;ti.a. llsinp; the (j para1l1d,('!" 'lait equation), folloWNI by ADS (7 par:I,I1I('t,Af gnncralizc:d eroso 111()(kl with inclusion or a yield stress] while I'A was, because of lack of OWli dal.a, IIl(H"C roughly .Ioscribcd (4 paramercr truncated pOWN law model alii! COil stant <ipnsity) Since ill this thesis we do not intend to perform visc()-l'lasj,ic cal«ulatious, t.he characterization has not been elaborated in this din'I·.I.iori St(' for a complete visco-elastic description of t.he PS used, e.g. Douvnn (I !ml) and/oJ" Zod.('lief" (Hl92).

Chapter 5

Particle tracking: two test problems

5.1 Introduction

In this chapter the numerical models, especially those for particle tracking, are tested using two well-defined problems. The first problem that. will be considered in section 0.2 is the How of a generil.li~ed Newtonian fluid between two pistons advancing in a tube. In the next section the geometrical complexity of the problem is increased, resulting in the piston driven contraction flow. In these problems, the numerically and experimentally obtained deformation patterns that develop during the flow, are compared.

5.2 Piston driven flow

In order to validate the particle tracking capability, the test problems have to meet some demands regarding their geometrical complexity and flow charactcristics. The piston driven flow problem proved to bA suited for this purpose, since its geometrical complexity is low awl the initial and boundary conditions aH.· well defined while the flow is complex. Moreover, the flow geometry is in accordance with the accumulator which is designed for future use in the multi-component moulding.

5.2.1 Material and methods Experimental

The experimental set-up used is drawn schematically in figure 5. [_ A sample (016 x 54 m.m) containing differently coloured slices is placed between two

07

58 Chapt,('I' iJ

J:li~llH~ S,1: Exporimcnral sot-up piston dr ivun Ilow

pistons in a heated cvlinder The lower piston is kept fixed, while 1,1i(' IlPP(~1' piston is able to move freely. A load P is put on top of the upper pist()11 to prevent the (j(,v,·\opm(mt. of shrinkage holes riming cooling "I' the sample and, moreover, t.o ensur« contact between (,It" piston and tlw polymer. After the cylinder has been heated to t.he required tcmperat.urc, the cylinder is driven downwards wit.l: a. constant rate 'Up- By doing so, til(' lower piston ad" like a driving piston and the upper piston can bp r'C'Wl,rdcd a~ dri von by the fluid. After moving the cylinder, the apparatus is cooled clown ,\,nd after solidification I.h.~ sample can be taken out. The dcforrnut.ion 1,,),t1',(~rns arc II1,l(k vi~i bl" by cutting the sample along the axial plan(·' of svmmr-try. The dnvclopment of the deformation patterns in time is visualized hy u~illK different cylinder displacements 6.s,

The experiments arc cA.rri(~d out with polystyreue (PS) and acrylonit.l'ikl)1Il,~vli"rl('"'kl.'yrnllc (ABS), These lIIatcria\~ differ in their flow behaviour at low shear rates a~ i~ r\",';(:ribed in chapter 4. PS shows Newtonian behaviour ill Lllio region whereas ABS tj'nds towards a more complex, Bingham-lib', How behaviour: TIl(' cxporimcnts arc carried out at different t;c·[JlI)(,l"ai.IlJ'('s and cylinder velocities ?Jr" The thermal soak-time was 20 minutes after the set-point of the temperature has b(·)on roached. Two different sets of t,,~ts >Irc, performed. Fir6C, t",sts with a. sample solely consisting of coloured slier,s of I'S or A IlS am discussed. In thc~e series also the influence (If the tmnpcratmc: or cylinder velocity Ull L1li! deformation patterns is examined. In the ~C(fllld

Particle trecking: two test problems 59

series, the samples are composed of a combination of the two materials: PS on top of ABS and the reversed configuration. The experimental conditions are summarized in table 5.1.

L::!.s[mm]

.,~ .. ..... "'- "-,
1 PS 1.0 473 0, 10, 20, 30, 40
...... ,.,._-
2 PS 1.0 443 40
.. .,.,-
3 PS 1.0 5()3 40
, ...... "" .......
4 PS 43.2 473 40
.... _,.' ........
5 ABS 1.0 473 0, 10, 20, 30, 40
- ----"~.- ..
6 ABS 1.0 443 40
_,-_ .. ","""'_ ._- "'-
7 ABS 1.0 503 40
~ .. ",.-".
8 ABS 43-2 473 40
~ ... ,---
9 PS/ABS 1.0 473 0, 10, 20, 30, 10
..... _ . .-------'----
10 ABS/PS 1.0 473 0, 10, 20, 30, 40
~- Table 5.1: Experimental conditions piston driven flow

Numerical

The numerical simulations are performed using a. finit.e element mesh consisting of 776 biquadratic triangular clements as is depicted in figure 5.2. The applied boundary conditions are already given in figure 3.3. Because of symmr-try, only one half of the geometry is modelled. The wall (f4) is moving

Figure 5.2: Finite clement mesh piston driven flow problem

with the velocity lip. The total time interval for the displacement of 40 rntri (= D..9jvp) is divided into 200 timesteps. For the simulations with the sam" ples consisting of one material only, one iteration for each time step proved to be sufficient, The combination of two materials in one sample introduce an extra non-linearity in the simulations, since the velocity is also depend

60 Clispter 5

of the distribution of those two materials which is defined by the hlwl lield. In these cases, the velocity field is calculated by a Picard iteration method (successive substitution) every time step. The particle tracking problem is solved using the SUPG finite element method with the time-dependent upwind parameter (see Appendix B), The tirne integration is carried 0111, with it Crank-Nicholson scheme (0 = 0.5). The axial coordinates z are taken as iuitial label field at t = 0. By making contour plots of that label field, in which the contours coincide with the tracer positions in the experiments, the deformation patterns arc visualized and can be compared with the experimentally obt.ained results,

5.2.2

Validation

The experimentally obtained and calculated development of the deformation patterns for PS are presented in figure 5.:1;1 and b respectively. The deformation pattern [or fl. displacement of Omrn shows that th« heating and S\\l.>~t>(j\l"nt cooling introduces considerable disturbance of til" i n itia.l ly s(.raight tracers. This effect can still he seen in the sample IHIr.I"rgoing a cylinder displacement (If' l O nun. From the displacements or 2() 111.111. or more, the fountain effect occurring at both pistons can be rccogHi:0(~d. Til 1.11(' fountain region ne(I,1" I.htl upper piston material elements move from t.hc C('U(.('I" towards the wall which '"(~Sll li.s in a strong elongation of" th()sp elemrnt.s close to the piston. When approaching the wall, the material acquires >1. reverse velocity causing continuously stretching Z-shapes. Note that these Z-~tw.pes correspond with the Vvshapcs that are found in mould filling (Sf''' (l.g. the discussion of Boris, 1987). At the driving piston similar behaviour can be observed, but hert' Ow material moves from the walls inwards to the C('ut.l'!" or L1w sample. The resulting Z-:>hapes now appear in the con.' in~t.('ad of close to the wall, These ()bSI~rVil.tions show that in piston driven now expcrirnents three different. rflgions can be distinguished: the region ncar the driving piston where it reverse fountai n now occurs, the fountain flow ]'egion ncar the driven piston and an intermediate region where an axi-syrnmctrical two-dimensional flow exists,

The c<~k\.ll('l(.,'d deformations reveal all the dctailx describod ahovc. The difference that attract the attention is that the deformation in t.lie l""giOll along the symmetry line or the sample is overprcdictcd for all displacements. However. the relative distance between two successive displacements 6.8 of the same tracer that. is situated in the 'main' flow, for example l.h(~ second 0110 from below, agree well with the expp.firrwnt.s. Also, the position of t11(' Z-:;hapc~ ill the upper part of the domain is predicted well. S(enmingly, the I'(~siden(;e time in the reverse fountain region is underpredicted. It is raised

a)

o rmn

b)

()mm

r~' ....

10 nuu

lOmm

Particle tracking: two test prohlflms 61

20m'!)),

20mm

30mm

:)Omm

40mm

40mm

Figure 5-3: Deformation of PS fur different piston displacements; a} cxpcrlments, b) calculated (vp = 1.0 mmf s, T = 473 K)

that visco-elastic effects in the fountain flow region (may) cause this deviation.

The results for ABS (see figure SAa and SAb for the experiments and simulations using the data set 1 (table 4_2) respectively) show a similar behaviour of forming the Z-shapes_ This was expected, since the the characteristic deformation patterns in the fountain regions are solely governed by the balance

62 Chapter 5

of mass irrespective of the rheological behaviour of the Iluid (8"'1' also l'lf,ris, 19S7). In contrast with the experiments with PS, the tracer lines ill the conter region reveal a plug flow rather than a parabolic shaped deformation of the tracers.

Comparing to t.hfl experimental results, the flattened I.ra.eer profiles ill the core of the sample are not that distinctly present. The width of the plllg flow region is too small and, like with PS, the delorrnar.ions along the symmetry line arc overpredicted. Vos et (11. (1991) showed that adjusting the ,mL(.f'r·ial data, i.c, changing the value of the yield stress for ABS, resulted in it improvsment of the results, Figure 5.5 shows Ole predictions wit.h tho matcnal data set 2 (see (,':1.ble 4.2). Now the resemblance between the exporimeut.al and calculated results is remarkably good, especially HI' to the 6.8 == ;{() "IIPI/.. Therefore, in the following only data. set 2 will be used,

Figur« 5.6<1 and 5.Gh show, respectively, the experimentally and numcrically obtained influence of the temperature and lh~ cylinder velocity 011 the deformation of the tracers for PS. For t.he I)S samples, the efl·ed. of tlu: ternperature disappear, almost. completely above T = 473 K, Til", ill(Il'I1.:c;ing temperature shifts the transition between the Newtonian plateau region and the ~he;l.r-(.)Iirllling region towards higher shear rates. In this way, t.h« vistosity gradient from the core toward« to wall that. is still present at T .; 1T.l J( levels off to zero. The difference that can be observed het.ween the deformations in the core of the sample is mainly governed by the sh,.".tr-.(.hIJIJllrlg behaviour of the PS. The experiment with the high (:y lindor velocity confirms this observation. In tho numerical results the discrepancies in the core ddnrmation are the same as described before, but the influence of the temperature and velocity show the same tendency as ill the experiments.

The same lest conditions are also used for ABS. The results CUI I)f' found in figurt:' .5.7. Compared wit.h rs, ABS shows a different response on t.h« variation of the temperature, Where the core deformations for I'S incrcas« with increasing temperature, for A BS t.he deformations decrease. This can b(~ explained by the effect of the yield stress in ADS. Sirw!:' (!rIC ShNU" SI.l"PSH for a c!:'rl .. ~.il! shear rr-[:d.(-' will increase with decreasing temperature. t.h« n:g;ioll of the sample whcr .. , th .. , sh .. .ar 8l!·e~~, gov(erBe([ by the local shear rate and viscosity values, docs not. exceed the yield ~t.r(:8~ will 1)(: smaller. All higher vvlocil.y (and UIU~ a higher shear rat!':) has the same effect upon 1.11(: width of t.he core region where the plug type deformation can be observed. J\~ a consequence, the resulting deformation patterns become similar to those of a sheil!" thinning liquid. This is clearly demonstrated wit.h iJH' sarnplr' t.hat has llndf'rgOlw the higher cylinder velocity. On the other hand, wit.h increasing temperature the yield 8tH':';:'; becomes more important and only I.hl': flow (:omli(lon~ in fl small layer of material ncar the wall is such that

a)

Omm

b)

Omm

lOmm

IOmrr!

Psrticle tmcking; two test problems 63

20mm

20 11).m

30mm

30mm.

40mm

40m111

Figure 5.4: Deformation of ABS for different piston displacements: 11) experiments, b) calculated (1!p = i.o tnrti] s, T = 473 K, material data set l)

Tv is exceeded. This explains the broadening of the plug flow region and the decreasing COre defonna.tion with increasing temperature. Again, the agreement between the experimental and calculated results is good.

The development of the deformation patterns ill case that two materials are combined in one sample is shown in the figures 5.8 and 5.9. The results of the two series experiments with the material combinations differ in many aspects. With the PS material situated OIl top of ABS, a second reverse

G4 Clinpter .}

On/,m

lOmm

2CJmm

3CJmrn

10mm

Figure 5,[" Calculated deformation of ABS for different piston displnccrnunts (Ill' = J.O mm/,', T = 47::1 1(, material data set 2)

fountain rflgiotl at the interface of the two materials can be observed, while in the reverse initial configuration the PS material protrudes in the ADS. II) figure 5.9a ii" Call be seen that the core deformation is 8VCII higher than was observed in figure 5.3a. For these series also, the resemblance between the predicted and measured tracer deformations is remarkably good, Even the onset of the reverse fountain region at the material interface is present in thfl calculations.

From the results we can conclude that the predictive quality of the simulation code (and the applied constitutive models with the measured parameters) it> satisfactory for the experimental conditions used, In order to attain ., better agreement between the calculated and experimentally obtained deformation patterns, visco-elastic modelling of the flow should b .. , employ,'d and thE' inOI.l('nce of shriok(lge should be eliminated or controlled and calculated,

a)

b)

443/{

473K

473 }(

503K

503 J(

43.2mm/~

Figure 5.6: Deformation of PS for different test conditions: a) experiments, h) calculated

(if> CbapteJ' 5

a)

143K

b)

44.1 K

473K

50JK

43.2 mm] s

Figure 5-7: Deformation of ABS for different test conditions: a) experiments, b) calculated

a)

b)

PS

Particle tracking' two test probJ(lm~ 67

20mm,

20mm

30m,m

JOmm

40m,rl

10mm

Figure 5.8: Deformation of the combination PSI ABS for different piston displace. rnents: a) experiments, b) calculated (vp = 1.0 msn]«, T = 473 K)

ABS 111 __ 1

Omm

PS

ABS

Omm

10mm

Wmm

GS C1Jil.ptcr s

a)

Omm

b)

ABS

PS

lOmm

20mm

30mm

30mm.

41) ill.m

407mn

Figure 5.9: Deformation of the combination ABS/PS for different piston displacernents: a) experiments, b) calculated (vp "" 1.0 mrn]», T = 17;) K)

20mm

Pa.rtic/e trecking: two test pr()hlem8 69

5.3 Piston driven contraction flow

III the basic design of an accumulator, the material is forced through a nozzle into the mould. The effect of such a contraction on the deformation patterns is studied with a piston driven contraction flow. These experiments arc also used for validation of the fixed domain approach.

5.3.1 Material and methods

The experimental set-up is shown schematically in figure 5.10. The master

Figure 5.10; Experimental set-up constricted tube experiments

cylinder contains a replaceable inner cylinder and a restriction with the dimensions 05 x 20m.m.. After the experiment; has been carried out, the sample and the inner cylinder and restriction can be taken out of the master cylinder for further examination. The plug (016 x 60mm) consisting of differently coloured slices is put between the lower piston and the restriction. After the apparatus has been heated to the required temperature, the lower piston is driven upwards with a constant rate by means of a hydraulic cylinder with a force Fp. On top of the driven piston a load Fl is placed to ensure contact between the piston and the polymer.

For the experiments, the same materials arc used as described in previous section. Also the same test series arc carried out: PS, ABS, PSI ABS, and

70 Clispter 5

AJ3S/PS. The tests are performed with a cylinder temperature of 47:'1 I{ and with a piston velocity of 5 rnmj.s. The thermal soak-time was ~() minutes after the set-point of the temperature has been reached. During heating and cooling, the hydraulic pressure was chosen such that the force Fp f'quaJs the load F). By doing so, the temperature change, which induces volumetric expansion or contraction, takes place under equally isobaric conditions at both sides of the restriction and therefore no material is transported from one side of the restriction 1.0 ths other.

In the calculations, the fixed domain approach is used (see chapter :1).

The mesh containing 2240 elements covers the complete flow domain, thus includes the region below thE' contraction as well as above the contraction (sf'e figure 5.11). The steel pistons are represented by a material wit.h a

'Vp -

l_:

Fig\Iff., S.ll, Mesh of piston driven contraction problem

high (Newtonian) viscosity of 107 Pa s which is about ,) orders in magnitude higher than the viscosity of the polymers. Moreover, the boundary condition at the wall is dependent on the material: a no-slip condition for polymer and a slip condition for the pistons. Every time step the velocity field is calculated using a Picard iteration. The positions of t}H> piston Zmaterial interfacf's are updated every time step t:..t by moving them wil.h all amount of 'vpt:..t. The viscosity is interpolated linearly per element to avoid the (Kcurrence of unrealistic values in the intergration points. This may occur at. the piston/material interfaces due to the quadratic shape functions of the elements used. As a consequence, the 'thickness' of the interface depends on the local mesh size.

The total time-span of 11.9 s (t:..s ::::: 60 mm) is divided into 595 timestcps.

The tracer deformations are made visible by plotting the colour contours at st:'veral values of the label field containing the initial axial coordinates. 'I'he t.ime integration is for stability reasons chosen to be fully implicit (0 c." ·1).

Particle tmckillI;' two tf-Bt problems 71

5.3.2 Validation

The deformation patterns for PS and ADS at different piston displacements can be found in the figures 5.12 to 5.15 respectively. These results show the same characteristics as is shown in the 'normal' piston driven flow: a reverse fountain flow at the driving piston and a fountain flow at the driven piston. It is remarkable, that the Z"shapes that arc developed at the driving piston also can be observed at the upper part of the restriction. However, these Z-shapcs originates from those that are developed at the driving piston and do not stem from a reversed fountain flow in upper part. of the geometry as is dearly shown in the calculations (see ego figure 5.13). The results for ABS differ not much from those of rs, only the deformation of the tracers is larger for rs. The plug flow observed for ABS in the previous section cannot be observed in these experiments, apparently because of the overruling influencp- of the (axial) deformation in the contraction region. For both materials, the calculated tracer deformations compare well with the experimental results. Only some slight deformation of the piston/polymer interface can be observed.

The tests with the combination or the two materials reveal that the differcnce ill flow behaviour of tho two materials has a large effect on the deforrnation patterns as is depicted in the figures 5.16 to 5.19. For the combination ABS/PS (figure 5.18) the tracer pattern in the lower part (between lower piston and the restriction) tends towards 1:1 plug like deformation with the black tracers at the wall surrounding the core material. A closer look shows that the PS material encapsulates the ABS. In figure 5.16 no such phenomenon can be observed, neither any difference can be found in the upper part between the restriction and the driven piston for both combinations when comparing the results to those depicted in the figures 5.81:1 and 5.9a. Again the resemblance between the experimentally and numerically obtained results is good.

5.4 Conclusions

The deformation patterns have proven to be sensitive to all aspects that may influence the flow kinematics and thus can be considered as good comparative experimental information for selective material testing. The material behaviour influences t.he deformation patterns considerably, Difference in fluid behaviour, e.g. the presence of a yield stress, can be revealed directly from the results,

The numerical simulations show a good resemblance with the experimental results, Even in the case with the piston driven contraction flow, where t.he piston material is modelled as a fluid with a high viscosity that slips at. the walls, the calculations compare well with the experiments.

72 Cbnpter 5

Omm

lOmm

20mm

JOmm

40mm

50mm

Figure 5.12: Experimental deformation of PS when using a restriction.

Ornm

10m",

Particle tracking; two test problems 73

20 1ft7n

30m",

UOmm

Figure 5.13: Calculated deformation of PS when using a restriction,

74 Chapter 5

Omm

10 mrn

20 mrn

30mm

40 rum

Figure G.14: Experimental deformation of AilS when using a restriction.

(,0 HI,'In

Omm

10mm

Particle ttecking: two test problems 75

20 rnm

30mm

40mm

Figure 5.15: Calculated deformation of ABS when using a restriction.

50mm

76 Chapter 5

PS

19 rnrn

38mm

57mm

Figure ,5. Hi: Experimental deformations of the materials combination PSI ADS

ABS

Ornrn

PS

Pnrticle trecking: two test problems 77

19mm

38mm

57m1n

)-.'ignr(' 5_17: Calculated deformation of tho material eombinations PSI AllS

ABS

Omm

78 Chapter 5

ABS

Figure 5-18: Expmirmmtal deformations of thr: t.wo materials cornbino.tion A lIS/PS

1-'8

Omm

ABS

Particle tracking: twO test PJ'I)/:lIll1llS 79

19mm

38mm

57mm

Figure 5.19: Calculated deformation of the material combination AI3SjPS

PS

Omm

80

Chapter 6

Multi-component moulding: some examples

6.1 Introduction to the case studies

Since modelling of injection moulding and particle tracking are combined in OTI(> simulation corle (VIp), filling with multi-colours or even multi-components can be studied, In this chapter, several characteristic", will be elucidated with the aid 0(' three different examples: ('i) co-injection of a strip, (ii) filling a product wit.h a bifurcation, and (-iii) inverse mapping in order to determine I,h(> inject.ion sequence for a desired colour or even material distribution.

6.2 Co-injected strip with ribs

Experimental results were obtained in a cooperation between Philips Centre for Manufacturing Technology, Eindhoven, The Netherlands and the Du Pont European Technical Centre where the products were moulded. The different test series arc described in Peters et al, (1994) and Hendriks (1992).

6.2.1 Experimental conditions

The product consist of a strip of 230 X 120 X 2.3 mm with 4 stiffener-ribs and a beam-like edge (see figure 6.1). The ribs are 12.7 mm in height and have different cross-sections, The beam-like edge (7 x 9 X 120 mm), acting as a flow divider, connects the strip with the runner section. The product is filled by sequential injection of two differently coloured nylons (PA66). The colour first injected was white-translucent, followed by the second which was blue. They are also referred to as skin and core material respectively.

81

82 Cnspter 6

120

230


1 1
1 I
I I
1 I
I I
I I
1 1
1 1
1 1
1 1
1 I
I I
1 1
1 1
1 1
1 1
1 I
I I
I I ~
---,---r--
V f 2_3

S S 5

7

gate?

Figure 6-1: Schematic drawing of test g()orrwtry

The strips WiJr€ injected in a mould with a. constant. temperature of 318 I< using different processing conditions. In particular the injection temperatu re of th~ two materials and their relative fractions were altered, see table 6.1. The core material is injected after the product has been filled up to the

------------"-_._-_._-,, -- - --,._---

series iir<j [K]

skin (1) core (2)

811Jp

PAl 571 556

67 %

PA2 571 556
PA3 571 556
PM 571 556
PA5 574 ;:)7:1
PA6 556 568 50 %

22 %

56 %

56 %

56 %

'I'abl(·~ 6_1 :f-'r()(~(lssinp; conditions test series PAl-PAil

volume percentage given by the switch point surp, The machine settiugs aimed at an average How rate of 1.33 -10-4 '!n:l/s_ The measured nlling t.irne proved to he about 1.9 ,9 for product including runners (with a total volume of 1. 16-10-,1 'If!"), thus the actual How rate was less: about 6_1-1O-s m3/s_ In t.he calculations the runner section is not taken into account, so the volume to hr filled is l.O() ,10 ~ m? The finite element mesh covering the midplane of the cavity i~ plotted in figure G,2, Since no details about the exact temperature

Multi-component moulding: f.;()I1Je examples 83

(distribution] are known, cooling is assumed to be symmetrical over the thickness direction and a Biot-type of boundary condition is chosen with a heat transfer coefficient of H = 30110 W/(m2K). For that reason, ami to save computing time, only one half of (,hEi thickness is considered. The half gapwise thickness is discretizcd using 25 gridpoints with 13 points (including the solid/Iiquid interface) equally divided over the fluid part and 12 points in the solid layer.

Figure 6.2; Finit« clement mesh coinjccted str-ip with ribs

The main points of interest in this example are:

• application to relatively complex geometries ~hol.dd be possible, since this is the aim of the simulations.

D the simulation code should he capable to predict. the influence of the main processing conditions On the material/colour distribution in the product .

• a proper prediction of break-through of the first injected material by the s('('onil injection material, because the occurence of undesired breakthrough is the most common problem to solve in practice of multicomponent moulding.

6.2.2 Validation of the numerical simulations

Identical to the test problems dealt with in chapter 5, the colour distribution is used for visual validation of the results. Since the skin material is translucent, the skin/core material distribution can be determined easily without cutting the sample in pieces, The same holds (but to a somewhat less extent.)

84 Cha.pter 6

for the areas where the blue core material breaks through the skin material and ends up at the product surface.

a)

.''111[> = 67%

b)

swp = 67%

st(lp = 50%

8W[i = 50%

,,'wp = 22%,

Figur" fl.~l; Experimental material distribution in the midplane (,'J and breakthrough areas (b). Influence of swith point (Ij".11,2 = t)71, 556 K).

<1)

swp = 67%

b)

swp = 67%

Multi-component lnimlding: some examples 85

s'WP = ,50%

8WI' = ~O%

.Qfll) = 22%

.~wp = 22%

Figure 6A~ Calculated roaterial d istr ihut.ion in the midplane (a) and breakthrough areas (b). Influence of swith point (Tinjl.2 = .571,556 K).

86 Clw.ptGr 6

a)

1"':.' '" 571,556 tc

b)

1', ,< = 571,556 j(

T,,~ = 574,573 g

Figure G,G: Experimental material distribution in the midplane «~) ,1.11<1 breakthrough a.reil.~ (h). Influence of melt temperature (S1IJyI = 1)G(Yc,),

a)

TI,2=571,55GK

b)

1'1,~ = 571,556 f{

Mult.i-componeut moulding: some exssuples 87

TI,2 =574,573 K

TL,2 = ,574,,573K

TL,2 = 556,568 J(

TL,2 = 556, ,568 J{

Figure 6.(;: Calculated material distribution in the midplane (a) and breakthrough areas (b). Influenee of melt. temperature (swp = 56%).

Since wit.h sequential injection the material or its colour only changes III time, the injection time labels provide the necessary information for vi-