Discussion Groups

Simulation Moldflow Insight

Posts: 5
Registered: ‎09-02-2012

How can we know the simulation results is good

449 Views, 2 Replies
09-14-2012 12:57 AM



I have one question please. I have design Part Model using Pro Engineer. Also, I have design three types of cooling channels on the part. These cooling channels are: Conformal Cooling Channels, Conventional Cooling Channels , Conformal cooling channel for cavity and Baffle cooling channel for core.


What I want to know is:


How can we know the simulation results is good and is acceptable? Could you please give to me your answer in details how and why


Thanks and Regards

Please use plain text.
Board Manager
Posts: 266
Registered: ‎04-14-2009

Re: How can we know the simulation results is good

09-20-2012 08:11 AM in reply to: 007330757468

This is a very open question! :smileyhappy: and there is no 'catch all' answer to this.


There are many many factors coming to play when it comes to 'accuracy'. The categories below are all equally important to get to the highest degree of accuracy:

  • 'Accuracy of model and process'. This may sound like lame, but the reality is that very often what was modelled and what is on the injection molding machine is ofen not the same (tool changes are still commonly made after a simulation is done). Another thing is how well the simulation imput and assumptions reflect what is actually happening on the machine. The reality is that when the simulation is done, many things are still unknown, and the process that will be run on the machine in reality is often not a 100% match with the simulation input or assumptions you are making in the simulation. Did you include everything that is relevant?

In fact, 'accuracy' can be a stretchable concept in injeciton molding. If you see a large variation in what's happening on the injection molding machine on a shot-to-shot basis, than there is an 'accuarcy problem' on the injection molding machine, and 'simulation accuracy' may be a meaningless idea (the simulation will produce 'one' result, while the machine will produce a wide variaty of 'results' so ... what do you compare. There was a good presentation on this topic at AU two years ago (link here, you may have to log in first)

  • 'Accuracy of the material'. There are a few important points to make here: 1) the material you use may or may not be fully tested and generic data (representative of the family of mateirals, but NOT for the mateiral in specific) may be added. To get a 'high level' degree of confidence in the material data, look at the material quality indicators (bronse, silver, gold) in the stury tasks, material section (in red in the image below). the three medals indicate 'confidence for filling', 'confidence for packing' and 'confidence for gold'. material_quality.JPG

If you dive into the material data itseld you can find more details on whether the material data was measured or supplemented. If the material data is written in Black, it means that there is actual measured data, if the line was written in red letters, it indicates that the data is generic (see below).


The second part of the material story is related to the 'state of the art' in material data. Polymers are very complex materials. We capture the behavior through standard and propriatery testing methods to have the best understanding of the 'nature of the material'. Although most testing methods are established and have a long history, there are some limitations in the ability to capture the material behavior at injeciton molding conditions. In collaboration with research institutes we are working on narrowing some of the gaps through new testing methods and new algorythms.


  • 'Actual solver accuracy': This breaks down again in two areas; alogrythms, and numerical accuracy. We are using the latest and most advanced algorythms and are continuing to refine them release after release. We also do frequent numerical tests to see that althorithms converge to the correct answer. Although I have no way to 'represent' this for your individual model case, our solvers do provide important clues while the analysis is running in the form or errors ane warning messages in the log file. The log may contain important clues like the examples below:

** WARNING 98686 ** The specified flow rate is greater than the maximum

machine injection rate at 0 %fill. The maximum

machine injection rate will be used instead.


** WARNING 304920 ** There is insufficient refinement of the tetrahedral mesh in some areas, which

may affect solution accuracy. Inspect with the "Node layer number" plot and

consider improving the mesh. [For the 6 tetrahedral refinement layers requested,

the node layer number is expected to reach 4 at the part centreline.]


** WARNING 304930 ** The node layer number on the part centreline is 3 or less for 45.8 percent of the part.


Warnings are indications that something may not be 'optimal', an ERROR will actually stop the analysis (because the results would clarly be meaninless).


Admittedly, quite a few things may be unknown at the moment when a simulation is done, so you have to make the best possible desictions to get the highest degree of confidence in the results. All of the three categories mentioned above are important; there is no point having an numerically exact simulation if the model and input is not correct. I hope this is somewhat helpful ...


Another very good AU presentation on this topic can be found here (again, you may have to create an AU login to access this content).




Hanno van Raalte,

Product Manager for Autodesk Simulation Moldflow products
Please use plain text.
Posts: 5
Registered: ‎09-02-2012

Re: How can we know the simulation results is good

09-23-2012 07:39 PM in reply to: raalteh

Thank you so much for your answer


Please how can we decide that Injection Machine is suitable for my part? Because before doing analysis I need to select the type of machine



Please use plain text.