You are on page 1of 17

Flow balancing of rectangular profile die

Version: 2 By Jiri Svabik (Svabik@CompuplastVEL.com) Compuplast International (www.CompuplastVEL.com)

This document shows problems encountered when designing simple profile die, problems mainly linked to flow balancing. For theory of the various flow balancing methods refer to document BalancingMethods.pdf.

Problem description
The profile to be manufactured is displayed on the following picture.

2.2

40

3 2.2 65 Due to symmetry we solve only half of the die. The material used will be rigid PVC.

The first considerations


The first consideration we have to make is about the processing condition range. This is driven by economics of the production we should make good quality product but run as fast as possible. Let us assume that the temperature used will be 200C. This is a common processing temperature for rigid PVC. Let as start with requirement to have at least 2m/min speed of production. As the final shape has a cross sectional area 464 mm2, we can choose initial mass flowrate from kg kg Q = vS = 1400 * .0 3 * 0.000464 = .02165 3 = 77.952 s hr where is solid material density.
We choose initial flow-rate to be 78kg/hr. We should at first estimate the desired length of the parallel die land. We will pick length 70mm. The parallel die land allows us to forget the flow history, orient the molecules and iron the shape. Typically people estimate the length by use of L/D ratio or they use experience. Often the larger the gap, the higher L/D ratio is used. As our profile is relatively wide and our gaps large, we choose parallel die land 70mm. After the polymer leaves the die it will be pulled down to the calibrator. So we assume initially that the die exit cross section is having the same shape as the final product, being only 10 % larger in size. This assumes that the draw down is isotropic the calibrator needs to be placed in sufficient distance form the die exit. If the die opening is made as described we get - for an infinitely long parallel die land the following velocity profile:

As it is symmetric, we can show only one half of it.

Also we can find, that the wall shear stress in the horizontal arm is around 150kPa/mm. Pressure drop per millimeter is 130kPa/mm. That for 700 mm long parallel die land would be around 9.1MPa. We should say that we constrain the speed of production by these two factors. The overall pressure should not be too large so that extruder is not overloaded. If we assume that maximum allowed pressure drop is around 15MPa and that about 80 percent will be consumed by parallel die land, we are still safe. Another limiting factor is the shear stress. If the shear stress is too high, we can over shear the polymer and have excessive energy dissipation. The typical limit for polymers is around 200kPa. So the chosen speed of production is safe again. As PVC slips and in the current design we ignore slip, we can afford to run at even higher production speed. If we try speed of 3m/min, we get mass flow-rate Q=116.92kg/hr. Wall shear stress in the horizontal arm is around 176kPa/mm. Pressure drop per millimeter is 149kPa/mm. That for 700 mm long parallel die land would be around 10MPa. So we are still safe and in fact we could be running even faster. Here we just want to show the methodology, so we stop increasing the line speed and use Q=116.92kg/hr for the total profile shape.

Flow balancing domain depiction


When we see the initial flow inside the parallel die land we can see that the material in the thicker section flows faster probably giving us more flow-rate then we need. Also the thinner vertical arm has insufficient speed and flow-rate. If the flow is unbalanced not having correct flow-rate in each arm, we will not get the profile shape we want (see BalancingMethods.pdf for discussion why) Looking at the picture allows us to depict the domain we need to control.

We certainly need to separate both the vertical arms from the horizontal one and control the flow-rate on these flow domains. The separation can be done in different ways. To keep the number of domains small and simplify the explanation we choose to have only 3 domains. One could also focus on the corners as special domains we do not do it here for sake of explanation clarity. II I III

So the real profile has 4 domains. As it is symmetric we get 3 domains. Then we can investigate the cross sectional area of the final shape.

II I III

Area [mm2] / [%] I II II total 56.5 / 24.35 137.67 / 59.34 37.84 / 16.31 232 / 100

Qreq [%] 24.35 59.34 16.31 58.46

Q [%] 45.58 45.10 9.33 100

Vavg [mm/s] 108.28 43.96 33.07 57.85

The table shows that if the parallel die land shape is kept similar to the required final profile shape without any flow balancing, we have very unbalanced flow. Domain I gets double the flow-rate, domain III gets half of the flow-rate required. It is also clear that the average domain velocities are very different! If we pull the profile with the speed 3m/min (50mm/s) the domain I will slow down to half speed! The domain III must accelerate. These post-extrusion effects will certainly change the shape. The profile domain I will be much thicker, domain III will shorten and become thinner. The calibrator will try to fix the shape but it will not be possible as it cannot redistribute the flow-rate in-between domains!!
(For explanation why for the flow to be balanced the required domain flow-rates must be in percentage equal to the domain area percentages - see document BalancingMethods.pdf.)

So we need to apply some balancing methodology.

Methodology
The various flow-balancing methods will be discussed and solved using the following assumptions: - the flow has dominant downstream velocity component - the cross sections should all be balanced they do not compensate for the good or poor performance of the previous sections. - when balancing by compensation section is performed, the assumption of no cross flows is adopted. This cross sectional solution is solved by use of VEL Profile die module. The assumptions above are of course too strong in many cases. Whenever we perform 3D solution we can see cross flows, compensating effects and much more. We certainly see different results from cross-sectional solution and 3D solution. Mainly in case of unbalanced flow case. However, the assumptions help us to balance the flow efficiently using gaps and for such a design the 3D solution matches the cross sectional solution quite well. The match between compensation section calculated and designed by cross sectional method and 3D will be investigated case by case. The 3D mesh is for simplicity made with little sharper corners then the mesh done in PD module. The new cross sectional area is by 4.51 percent smaller then original cross

sectional mesh. This does not affect the flow rates in each domain, however the average velocities reported would be different. Here to be able to compare we scale the average velocities by 4.51 percent in the 3D solution.

Unbalanced flow
At first we will show how the material flows on the geometry. Of course the unbalanced flow is more 3D flow by nature, so cross sectional method does only show it is unbalanced but does not quantify how much. The following pictures and tables show the unbalanced flow on the cross sections. We also indicate the 3D results wherever it is worth to show.

Z=(-260,-240)

Z=-165 I II III

Q required [%] 24.35 59.34 16.31

QPD unbalanced [%] 17.76 64.47 17.77

QP3D unbalanced [%] 31.81 54.75 13.44

Z=-155 I II III

Q required [%] 24.35 59.34 16.31

QPD unbalanced [%] 19.72 60.54 19.74

Q balanced [%] 33.18 53.4 13.42

Z=(-143,-116) I II III

Q required [%] 24.35 59.34 16.31

QPD unbalanced [%] 25.02 49.95 25.03

Q3D unbalanced [%] 35.62 50.16 14.22

Z=-107 I II III

Q required [%] 24.35 59.34 16.31

QPD unbalanced [%] 24.19 53.73 23.14

Q3D unbalanced [%] 35.94 51.18 12.87

Z=-105 I II III

Q required [%] 24.35 59.34 16.31

QPD unbalanced [%] 19.73 60.53 19.74

Q3D unbalanced [%] 36.09 51.3 12.61

Z=(-70,0) I II III

Q required [%] 24.35 59.34 16.31

QPD unbalanced [%] 45.60 45.60 9.34

Q3D unbalanced [%] 48.48 42.8 8.72

One can see that the unbalanced plates are mainly the last plate. The other plates are also not balanced but it is certainly less dramatic.

Also it can be found that the pressures are for the unbalanced case: z 0 -70 -105 -116 -143 -155 P P,z p 0 148.164 10.40 148.164 10.40 13.24 13.18 2.832 13.40 15.11 0.15 13.81 15.11 0.4 13.98 13.18 0.17

-240 14.29 3.427 0.3

-260 14.36 3.427 0.08

One can conclude that the only plate which really needs to be balanced is the last parallel die land. Minor effect will have also the flow balancing of the section with mandrel support. Here a picture showing flow passages for unbalanced flow is presented.

One can see the tendency of the material particles to redistribute prior the parallel die land. In the parallel die land the material flows straight.

Flow balancing by gaps


At first we will balance all the crosssections by opening and closing the gaps. Changing the gaps and creating balanced flow should be made mainly for the plates which have considerable flow resistance. That is the parallel die land and perhaps also the plate with mandrel support. Prior the mandrel support the mandrel starts to develop so there the balancing is not made. The way we do the flow balancing on the mandrel depends on the way we decide to manufacture the mandrel. It can be made as separate piece of steel which is later fixed to the main body, or the mandrel can be one piece with on of the plates. We choose here the second choice. Mandrel will be part of one of the plates. Of course the spider legs the mandrel supports around which the material flows generate two more problems. The material flowing around creates weld line and also the legs create extra flow resistance which should be taken into account while balancing the flow. That is, however, difficult to make as we need to modify surface inside the steel. So we do not balance the cross section where the mandrel support starts but we keep at opened as the mandrel support is. This allows easier manufacturing. Our choice of making the mandrel connected to plate can be relatively expensive to manufacture. User can choose a different method but it may need to modify slightly the flow balancing approach (use compensation section). The following pictures and tables show the PD and 3D flow solution on the cross sections.

Z=(-260,-240)

Z=-165

Z=-155 I II III

Q required [%] 24.35 59.34 16.31

QPD balanced [%] 18.29 69.36 12.36

Q3D balanced [%] 22.1 61.95 15.95

Z=-153 I II III

Q required [%] 24.35 59.34 16.31

QPD balanced [%] 22.24 62.92 14.83

Q3D balanced [%] 22.38 61.37 16.25

Z=(-143,-116) I

Q required [%] 24.35

QPD balanced [%] 24.29

Q3D balanced [%] 23.59

II III

59.34 16.31

59.62 16.09

59.34 17.07

Z=-107 I II III

Q required [%] 24.35 59.34 16.31

QPD balanced [%] 24.20 59.69 16.11

Q3D balanced [%] 22.89 61.31 15.80

Z=-105 I II III

Q required [%] 24.35 59.34 16.31

QPD balanced [%] 24.35 59.34 16.31

Q3D balanced [%] 22.89 61.48 15.64

Z=(-70,0) I II III

Q required [%] 24.35 59.34 16.31

QPD balanced [%] 24.35 59.34 16.31

Q3D balanced [%] 25.11 59.06 15.83

V3D[mm/s] 74.716 57.243 49.57

The pressures for balanced flow are slightly higher mainly as the flow resistance in parallel die land has been increased. z 0 -70 -105 -116 -143 -155 -240 -260 P P,z 0 170.05 11.9 170.05 13.89 13.04 14.03 12.36 14.39 13.39 14.55 11.23 14.89 3.427 14.96 3.427

The PD predict total pressure 14.96MPa, the 3D predicts 14.693 MPa.

Flow balancing by compensation section


Often the users insist on keeping the shape of the parallel die land as required final shape. They advocate by requirement of average velocities being the same over the exit cross section. Also they claim that the stress being more uniform can cause equal shrinkage across the profile shape. Here is a table showing the exit cross section average velocities for the various cases: Average Exit CS as required Exit CS as required Exit CS modified velocity shape (scaled by 10%) shape (scaled by 10%) by gap balancing [mm/s] unbalanced flow If flow balanced (Qs as required) I 108.28 57.85 73.30 II 43.96 57.87 58.18 III 33.07 57.79 54 It is obvious that the balancing by gaps brings the average velocities close to the desired equal for all domain averaged velocities, but there is still some difference. The flow balancing is in general based on the requirement to have flow-rates in each domain proportional to cross sectional domain area on the final shape. So if the shape of the cross section is different from final shape (having different area percentage on each domain then the domain has on final shape), the average velocities inside the die are not the same for each domain. Balancing by gaps while very simple and straightforward, cannot bring us to having also the same average velocities at exit cross section. Inside the die, the average velocities can be arbitrary they do not need to be equal for the domains. So we typically balance the cross sections by changing the gaps, only for the parallel die land we can use another method. If the parallel die land cross sectional shape is kept as required shape of final product, by definition the average velocities must be the same! In our case they differ only because we have scaled the shape by 10% what makes the wetted area to be scaled differently for different channels. So it can be concluded that keeping the shape as required may be in some cases (typically for non-calibrated profiles) a necessity. In such a case the flow balancing should not modify the last plate cross section. When the last plate shape is kept fixed, we need to do the flow balancing by compensation section. The approach of compensation sections is in detail described in BalancingMethods.pdf. We mark last plate by Y and the compensation plate by X.

Compensation section - keeping the shape of the plates fixed modifying lengths
Often, typically in case of simple dies, we feed the parallel die land plat from a plate with much higher cross sectional area. Then, designers tend to open the entrance part to each domains having less mass then required. What we are doing in reality is using balancing by lengts. In this case we need to identify if the flow resistances of the domains for compensation and unbalanced plate allow it or not. Typically the compensation plate has low flow resistance. As we want only shorten the length of unbalanced die plate, we choose as reference domain the one with the least pressure drop. PXr , z ( L Lm ) + PYr , z Lm = min[PXi , z ( L Lm ) + PYi , z Lm ]

Where Lm is the maximum length of unbalanced plate (in our case it is 70). Our total pressures on each domain are for different lengths plotted in the following table P[MPa] for Lm=50 For Lm=70 I 14.445 13.485 II 13.276 14.947 III 12.856 15.51 One can see that for shorter Lm we would need to pick as a reference domain the domain III. In our case when Lm=70 we pick as reference domain the domain I. From equations listed in BalancingMethods.pdf we can find out the new lengths (here for Lm=70): length X length Y I 35 70 II 52.21 52.49 III 50.27 54.73 One can easily check that the total pressures match for each domain. PXi , z LX i + PYi , z LYi = 13.485MPa Please note also that if we choose the reference domain using the above approach (the one with minimum total pressure), the overall pressure is minimized as we find new lengths such as to have the same total pressured drop on each domain as this reference domain.

Balancing by compensation sections opened shape balanced by gaps


If we choose that the compensation shape will be the shape which we have at the balanced cross section (z=-105), we get the flow resistances PX,z [kPa/mm] PY,z[kPa/mm] I 12.36 112.43 II 12.36 170.22 III 12.36 191.95 One can see the plate X is very open and also its flow resistances are identical (as it is balanced by gaps). If we assume that we need 5mm for the transition, we can find that the lengths should be X1/Y1 [mm] X2/Y2[mm] X3/Y3[mm] 20/80 49.3/50.7 55.42/44.58 When we solve the geometry suggested by Then we get at the Exit CS the following results: Qreq [%] Q3D [%] V average 3D [mm/s] I 24.35 38.64 78.973 II 59.34 49.68 47.895 II 16.31 12.85 45

Such a geometry was made by deciding that the Lm should be 80 mm. The other domains lengths in Y plate are, however, much shorter so one gets much lower pressure drop. One can see that the flow balancing produced a geometry, which does not balance the flow we have very unbalanced flow and also averaged velocities are different the method creates undesired cross flows!

Balancing by compensation sections closed shape balanced by gaps


In the previous chapter the shape of X plate was chosen to be very open. In case we choose very restrictive shape to be used as a compensation plate shape it may happen that we cannot find lengths which would balance the profile! The next table shows the flow resistance of each domain for case when the very final shape balanced by gaps was chosen as compensation plate: PX,z [kPa/mm] PY,z[kPa/mm] I 170.05 112.43 II 170.05 170.22 III 170.05 191.95 It can be easily found that we cannot find the lengths which would balance the die! So if we use shape balanced by gap as a shape of compensation plate, it cannot have higher pressure drop per millimeter then the unbalanced section! However, as any cross section in-between the two balanced part (open and closed) must also be balanced - we can pick arbitrary position. We picked a section at z=-75.25mm. Then we get the flow resistances: PX,z [kPa/mm] PY,z[kPa/mm] I 77.28 112.43 II 77.28 170.22 III 77.28 191.95

And we can find that the lengths should be X1/Y1 [mm] X2/Y2[mm]

X3/Y3[mm]

5.25/70

30.88/44.37

36.24/39.01 V average 3D [mm/s] 95.371 46.883 41.707

Then we get at the Exit CS the following results: Qreq [%] Q3D [%] I 24.35 40.63 II 59.34 48.63 II 16.31 11.91

The flow indicates that the cross flows occur prior the parallel die land.

Compensation section balancing by both the lengths and gaps


Alternatively the user can have a reason why to use a compensation section of some other shape then the shape balanced by gaps. The shape can be almost arbitrary, but we need to check if the balancing can be reached. Such approach can be also used inside the die to compensate for some plate which is difficult to be balanced by gaps.

Compensation section considerations


To be able to balance the flow reasonably well we can choose a compensation plate which has lower flow resistance on a domain, which was getting not enough flow-rate, and vice versa. The pressure drop per millimeter for each domain assuming correct mass flow rates is listed on the following table. p,z[kPa/mm] X balanced gaps Y I 160.44 112.43 II 86.63 170.22 III 59.25 191.95

The previous pictures shows how the material would flow on the compensation plate only and on the unbalanced plate only. Of course in reality, they will influence themselves and ideally we will have the correct flow rates at the die exit. In this case the shape of the compensation section is chosen arbitrarily to have bigger gap where the unbalanced plate has smaller gap and to have smaller gap where the unbalanced plate has bigger gap. The next table shows the lengths which the theory suggests for certain reference length they always result in getting the same pressure drop for all domains: X1/Y1 [mm] X2/Y2[mm] X3/Y3[mm] 10/60 42.65/27.35 39.33/31.67 20/50 36.91/33.09 34.71/35.29 30/40 31.16/38.84 31.09/38.91 40/30 25.42/44.58 27.48/42.52 50/20 19.68/50.32 23.86/46.14 60/10 13.93/56.07 20.24/49.76 From the table it is clear that we can minimize the change of length by choosing lengths of unbalanced domains (39.3, 39.238, 39.16). This is almost negligible change of lengths !!! That would be easier to manufacture! That means we have found an optimal length for our arbitrarily chosen domain. What is, however, easier is to define lengths to be the same for all cross sections and find the cross sectional shape of the compensation plate. This has been done in profile die module. The new predicted length should be 40 for both compensation and unbalanced plates. The pressured drops per millimeter are for the particular domains: PX,z [kPa/mm] PY,z[kPa/mm] I 228.33 112.41 II 170.50 170.21 III 148.80 191.94 Exit CS I II II Qreq [%] 24.35 59.34 16.31 Q3D [%] 32.98 54.08 14.11 V average 3D [mm/s] 77.4 52.14 49.42

The previous picture shows 3D domain painted with pressure color field. The 2 cuts indicate different speed on each plate. The pathlines indicate flow patters with cross flows. One can see the material redistributes in the region where compensation plate changes to unbalanced plate the cross pressure gradient would be too high so the material cross-flows.

Compensation section - Conclusions


None of the attempts to compensate the flow were successful. All the design domain produce heavily unbalanced flow. The product is not going to be as required. One can see that there is some balancing effect but as the unbalanced parallel die land is still too long, the material redistributes. Making the unbalanced parallel die land shorter can help to get more balanced flow. In general this method is certainly not recommended as the die would be very sensitive to flowrate and material variations. The only possible change how we can get correct flow balancing by using length (compensation sections) is the flow balancing with flow separation.

Compensation section with flow separation


The material cross flows as if it would not, the cross pressure gradient would be too high. The difference in pressure gradient is the largest at the place where the two plates meet. So we need to separate them there. The separation does not necessarily go from beg of plate X to end of plate Y, eventhough this would be the ideal case. We are in how the separation is made restricted by the requirement not to have visible marks on the final product. The pressured drops per millimeter are for the particular domains: PX,z [kPa/mm] PY,z[kPa/mm] I 146.38 115.29 II 88.73 172.92 III 63.40 198.24 Exit CS if separation gap is kept open Qreq [%] Q3D [%] I 24.35 32.4 II 59.34 54.5 II 16.31 14.23 Exit CS if separation gap is more closed Qreq [%] Q3D [%] I 24.35 27.37 II 59.34 58.52 II 16.31 15.28 V average 3D [mm/s] 72.77 50.31 49.84 V average 3D [mm/s] 64.243 56.418 53.515

From the 3DFem simulations made it is clear that the partial separation technique is efficient but only when the separation gap is left very small so that material cannot flow across.

You might also like