Simulated pressures far from reality

[PascalGosset]

I'm not a novice. I think since many years the pressure predicted are often false.

So I made a series of comparisons between simulated and real pressures.

I not only compared the pressures at switchover because they can change rapidly.

But I compared the pressures curves to see all their increasing during the filling.

For my simulations, I took into account :

- the part and its runners

- the material in the press nozzle

- the real molding settings

- the exact material used.

I found pressures far from reality except in one case.

The error comes always because of an underestimation of the pressure drop in the feedings.

See 8 samples below.

Then my questions are :

Do you also find big differences between simulated and real pressures ?

Are the errors due to the data of the material ?

Its compressibility ? Its juncture loss ? Something else ?

Find by simulations the real pressures encountered is it an utopia ?

 

[Anonymous]

Hello Pascal,

you present very nice overview and indeed that's when discussion becomes interesting, I mean when people show real data and compare them with simulation.

may I ask you these quaestions:

- how did you measure these red lines showing real pressure? Is it the pressure sensor located on nozzle of injection molding machine or somewhere else to sense pressure of molten plastic material? Could you be more specific?

- regarding each material you mention: what about these materials' status in Moldflow database? I mean are these materials "golden" or "bronze" regarding rheological and thermal properties in Moldflow database? Or are some perhaps "red" meaning, these would be just generic data?

- when thinking what "from geometry side" might have biggest impact...: you know really well actual dimensions of gates on these real molds? I mean, when I was apprentice, after some first trials, easily we adjusted (opened) gate, for example, of course that this change never got back to designer, or something like that, so I'm just trying to think what has "obvious" effect of mainly the "offset" of shapes of red (real measurmenet) and black (simulated) pressure curvers from your posted info.

- when you write "real molding settings", would you maybe on one example attached (at this forum) really all details of all (all you know...) injection molding parameters used in real molding?

If I remember well from youtube, there a video showing Moldflow material lab where is shown what they measure and how they meaure and at the end of video they say "last test is injection molding of sample to measure that predicted max. filling pressure is matched with simuated one - as proof of how well particular plastic material was in this lab characterized".  So I'm wondering if somebody from Autodesk can answer your problem.

Best

Rob

[PascalGosset]

To answer your questions :

Red curves are the pressure displayed by the molding machines.

They are measured with a sensor on the hydraulic unit.

They are turned to the plastic pressure by the machine software considering the screw diameter.

The parts studded were chosen perchance.

The materials also. Their data are +/- reliable as the materials I use for my simulations generally.

Some details of the materials tested :

Material                 Manufacturer        source         last        Viscosity correction              source PVT          Quality

                                                         visco         modif     C1          C2           D3                                     indicator

Plexiglas 8N            Roehm              Moldflow       2001      -              -             0             Moldflow               S-S-S

Apec 1695               Bayer                 Bayer         2011      -              -              0            Generic                 S-B-B

Bayblend T85 XF      Bayer                 Bayer         2011      -              -              0            Generic                 S-B-B

Ultradur B4300 G10  BASF                 BASF        2002      -              -              0             Generic                 S-B-B

Terluran HH106        BASF                 BASF        2002      +             +             0             BASF                   S-S-S

Technyl A218 V30    Rhodia              Moldflow      2007      +             +             0             Moldflow               G-G-G

PP PHC31-81          Sabic                Moldflow      2005      +             +             0             Moldflow               G-G-G

C1, C2 are the Juncture Loss Coefficients : + when they are filled (sometimes measured : 20% of the database).

D3 is the term of the viscosity pressure dependence (rarely measured : 1% of the database).

About the geometry, I took into account the updated 3D part, checked its thicknesses, measured the cold runners, the gates, modelized the hot runners and the press nozzles using the drawings.

Concerning the machine settings, I took into account the filling time, the screw speed, the barrel temperatures.

I estimated the cavity temperature from the water of the coolers.

Temperatures are difficult to measure.

But I made simulations varying the material and cavities temperatures and found only little changes.

At last. I don't know the video you are talking about.

I'm interested in.

So please, leave me the link if you have it.

[Anonymous]

Hello,

little video is this one (they talk about correlation between simulated vs measured pressure at 7m 20 s of video):

http://www.youtube.com/watch?v=A279vqI48Ow

Regards

R.

[Anonymous]

Hi

very nice results. But there are some details you should think of:

- quality of data files is important, especially when simulating in 3D

- I would advise setting up the runner system also in 3D for juncture loss, if material data has juncture loss data

- it would be better to compare pressure data aof a pressure sensor in the mould, there are so many pressure losses of the machine

Also think of how many layers you used over thickness for 3D. 8 would be minimum.

Greetings

Philipp

[raalteh]

Dear Pascal,

Could you please provide the files to our support team. We would like to have someone take a close look at these. Please indicate that Hanno van Raalte asked for these files.

Best regards,

Hanno

[PascalGosset]

To Rob :

Thanks for this video.

They effectively say few words about the correlation between simulated and real pressures.

But they don't show anything.

And I realize now I never saw anything from Moldflow about this subject.

On the other hand, I read several messages on this forum showing big differences between real and simulated pressures.

To Philipp :

I also think the trouble can come from the materials characterizations.

I tried to mesh the runners in 3D but this didn't increased really the pressures.

A sensor in the mould could be interesting.

Except my problem is to predict the maximum pressure necessary to fill the parts and their feedings.

To Hanno van Raalte

Thank you for your intervention.

In fact I already sent reports on this subject to my local support.

They replied me the 8th sample demonstrates the results can be good if the material is properly characterized.

But, how to know if a material is properly characterized ?

The Technyl A218 V30 seems to be (REP and Moldflow tested. Classified Gold) but it gives bad predictions.

And what about the other materials ?

Do you find it normal to have such gaps between simulations and reality ?

I did not show my 7th sample to my local support until now.

I think I'm going to do this with your recommendation.

This because it is a material I often need, there are 2 files for this grade in the database, one seems complete.

[raalteh]

Hi Pascal,

I will connect with support in France about this as well. I don't want to speculate on things or give you a generic response of things to look out for. I'd much rather have some peopel investigate the cases so we have a better understanding of what's going on. Some of hte discrepancies in pressure predictios are certainly not what you would hope them to be.

Having said that; there are things that happen on the machine and the shopfloor that can cause discrepancies between what's seen on a machine and in the simulation.Some examples (which may or may not apply here); material drying conditions being different, the use fo regrind, material degratdation in the barrel, leaking check valves, inclusion of colorants can affect material behavior, faulty temperature sensors on the machine, 'last minute' changes to the mold (hat were not included in the  model). These are a few of the things we have even run into ourselves when we generate data for research and validation. There are many possible sources of error in both the simulation and 'reality'.

I'm not suggesting the experimental results are wrong, but it's one source of discrepancies that should not automatically be discounted either. The simulation provides a simplified and idealized representation of reality; sometimes reality is more complex and less ideal.

Again, I will try to follow up with support in France and if you could provide your 7th case as well, that would be helpfull.

Best regards and have a great weekend.

Hanno

[kiamtm]

Hello Pascal,

From a long time I have similar doubts than the ones you showed. I've seen you even modeled the barrels (IM nozzle) into your models.

However, my feeling is that there is one big challenge in all this story - it  is hard to be 'really fair' to the design of both:

(1) IM nozzle and

(2) hot runner systems: the actual designs generally  have 3 'branches' at the end and some 'flow restriction features' that simply could not be modeled by beam elements (I think you also tried to take this into account in your models). Some hot runner systems (like the moldmasters E-type torpedo) depends on the mold plate tooling process precision. I've seen cases that a difference (error) of ~10^-2 mm make all the difference in terms of a mold 'work or not work'.

These 'extra' restrictions generate 'viscoelastic' phenomena during filling phase.

IMPORTANT: I never seen any trainning model or material in this 'level of detail'; explaining that we should take care with these features (i.e. all hot runner models are very simplified and no barrel modeled).

One doubt: have you performd 'air shot test' in any of the cases you are showing (i.e. tried to isolate the barrel pressure drop) and even compared to the pressure drop of your modeled barrel?

In the case you are running cold runners mold, at least the runners design became far more simple. And there is an alternative test to the air  shot (that would be not feasible in some machines): by producing short-shot starting from the runer system (cold sprue, runners, gate) and to take notes of the machine Peak pressure. As long as the cold sprue P drop tends to be very small drop, we would assume that most of 'sprue' value is the barrel P drop. Furthermore, you can generate more detailed information about P drop into feed system:

A final 'tip': in the past MF provided some 'generic juncture loss' into help: In the case one like to have this taken into account  (i.e. for poor characterized materials).

Please, let me know your comments.

Happy new year!

Best Regards,

T Kiam

[phammer]

HI 

interesting table. From what literature is this tabel from?

greetings

[kiamtm]

Pascal,

I was thinking on your post; and I have two additional remarks:

(1) If you are measuring the pressure drop by measuring the machine hydraulic pressure system: remind that there is a 'gap' between these values and the actuall nozzle pressure due to pressure losses in hydraulic system (compressibility and resistance).

(2) Finally, It would be really valuable if you perform the 'air shot' test by purging material thru the hot runner system: then you'd have the TOTAL pressure drop for machine nozzle + hot runner system (by performing an aditional air shot test just with machine nozzle, you can decouple the values). Some machines allows this test. You just should be care and cover the moving mold half by using cardboard paper (I perform this kind of test to asses the hot runner flow balance).

Notes:

1) The pressure drop plot is from MF document "to match an actual molding process (REV1)" - attached

2) PHammer: The table on the previous post is from MPI50 theory reference - page attached (in the case you have interest on the full document, contact me)

T Kiam

[phammer]

Hello Kiam,

great documents. Especially the match mold process.

In my last trainings I didn't see anything like this.

Greeting Philipp 

[Anonymous]

Thank you for a good resource.

How can I make this graph?

Please tell me....

[Anonymous]

Hello Philip,

Good day.

Did you found out a way to match the simulated pressure to actual molding pressure ?

I too face same sought of problem to match the actual injection pressure with simulation pressure.

Presently i work with a 1.5 meter wide part with 2.5mm thick.

In this part we have 12 direct hot valve nozzles & the material used is Sabic X2500UV (PC+PET).

During the trials we tried to fill the part sequentially by opening four nozzles to the side wall first. But pressure is too high (1830bar) when only 40% of part is filled. I tried to simulate the same and at 40% part filled the pressure what moldflow shows is only 830 bar.

Do you have a solution / suggestion for this on how to match the actual pressure with simulation.

One more question, how reliable is the clamp tonnage result from moldflow, if the pressure is matching with real situation?

Waiting for your reply.

Best regards,

Anoop.