Optimization of Injection Molding Process Settings Using Iliad and Moldflow

SHUBHAMKAR KULKARNI: Hello, everybody. I'm Shubhamkar Kulkarni. I work as a research and development engineer at OmniQuest. Today I'm going to talk about how the Iliad and Autodesk holds insight can enable the optimization of injection molding process settings. So the learning objectives for this session are an introduction to the design exploration and multidisciplinary optimization capabilities of Iliad, demonstrating the application of numerical optimization to injection molding simulation, and the use of meta-modeling to identify optimal injection molding settings, reducing the need for large computational times.

Here's a brief outline of my presentation. I'll talk about our company OmniQuest and more about our product Iliad. I'll summarize the current capabilities for design and solution exploration in Autodesk modular insight, followed by an instructional demo where we look at a few examples about how this collaborative Iliad workflow coupling enables optimization in injection molding. And then we'll conclude with the summary of the capabilities presented here today.

So starting with who we are, we are OmniQuest, also known as Vanderplaats research and development. Our company is over 4 decades old. It was founded by Doctor Garrett Vanderplaats. Our vision is to see optimisation unbound. We are one of the pioneers in the field of applied optimization. And our mission is to create these optimal technology and optimization enabling software, which would make design optimization commonplace.

So optimization is something which happens every day, but we talk about delivering optimization beyond the human intuition. As part of that, we have developed a lot of optimization engines over the years, which have proven their worth in the industry for over three decades. These are listed as Dot family and that's the name that they're popular with. And over the years they have shown to be really robust in terms of the number of variables they can solve and also the different domains they cover.

So Iliad, the formal name Iliad is Design Exploration and Automation Studio, or our ideas. And the objective of this tool has been to enable design engineers or process engineers to simplify their solutions and to help bring in automation and automatic improvements in their solution process. So, if you think about how an engineer develops a solution there is going to be a series of pseudo disconnected steps, each involving multiple analyses and manual inspection of certain things and that's how the entire solution is developed.

So Iliad fills certain gaps in this particular process. One, it helps to bring all these individual analyses onto a single platform. And this is done through some integration capabilities, and you have the entire workflow represented inside Iliad. Once that's done, an engineer can control everything from Iliad.

And this can also be used for design exploration, which means seeing how these individual input parameters can be changed and what their effect is on certain key outputs. We can start executing some automation. If there are some repetitive tasks we can automate the individual processing or analysis calls for each of those tasks. And then we can also bring in numerical optimization through our Dot family engines which are built into Iliad.

Taking a deep dive into the capabilities Iliad provides, so design of experiments and optimizations is something that I've mentioned before but Iliad can also do sensitivity analysis, meta modeling, and response surface based approximation. And we'll see some examples today on those two topics. We also have a probabilistic analysis tool for doing reliability evaluations associated with uncertainty in the data.

So really, quite a variety of different studies can be done through Iliad. And Iliad has been programmed to make things easier for the user, and as part of that we have integrated battle running and remote node integration for Iliad. There is also a lot of utilities available to do things in real time, such as depending upon certain decisions need to be made on certain critical parameters.

So we have some looping conditions, some conditional statements, and Iliad has been built to be robust. So in case, the underlying analyses fail, Iliad is able to flag those failure points and move forward continuing the design exploration and optimization study. There are also some simulation monitors available, as you can see in the snapshot in the middle.

The graphs that you see are something which update in real time and show how the objective is improving with each design iteration. Iliad, since I spoke about how Iliad integrates all the underlying analyses, we have generic components to do that. But over the years we have also created certain dedicated interfaces to other CA software, and that makes this process of integration really simple and more reliable.

So we cover both CA analysis software as well as scripts. And scripting software include Matlab, Excel Python. But we also have a dedicated interface to Moldflow through our collaboration with Autodesk. So that's the technology that I'll be demonstrating today. Iliad also has a variety of post-processing tools. And these tools have been developed with the idea of giving maximum information possible to the user and minimizing the need for rerunning, or reducing the number of analysis required to get to those conclusions.

So some of these tools into text reports, PowerPoint reports, table summaries, some pie charts. But we also have a proprietary what-if tool, which is a very practical tool. Once a study has been done, you may have certain changes introduced into the ranges or the input variables, and you may wonder what the new optimum point should be. So in that case, instead start rerunning the analysis, we have a way to get to the optimal point, approximately, without extensive analysis.

So let's look at how a coupled Iliad and Moldflow integration helps to bring optimization to the field of injection molding. So when you talk about optimization in injection molding, it's mostly ad hoc and dependent upon heuristic tricks and guidelines that the expert designer process engineers have developed over the years through their experience.

And there are a number of reasons for this. So if you were to go about design exploration or solution space exploration, using some of the techniques that are popularly used, such as experimental investigation, you have certain challenges to that because the cost of tooling modification and equipment modification is really high. So that's not economical. Talking about simulation, we have very powerful tools now. But for industrial cases running a single analysis typically means that it's going to take a few hours or days.

Keeping these things in mind, Autodesk Moldflow insight, and since it's a very powerful and robust software, has developed certain tools to explore this solution space and also to make things a little bit easier for the designer to improve their solution. And we will see how Iliad helps or adds to this particular capability to help provide an even more powerful performance in this particular field.

So here's a summary of the capabilities available inside Moldflow Insight, and what are coupling as in the Iliad reward for coupling adds. So talking about optimization, Moldflow Insight has two tools. The first one is the design of experiments. And this is applicable to certain key input parameters such as the mold and melt temperature, the injection and packing time, the thickness multiplier, injection and packing profile modifier, et cetera.

And if you look at the key critical outputs that are available for studies, they include temperature, shear stress, shrinkage, sink depth, et cetera. So if you look at the tool inside Moldflow, it can run certain combinations of these input factors and show the effect you have it has on the outputs in the form of some Visual graphics.

And a user can go in and through their interactive plots look at how just changing one or more of these parameters affect the output. So in addition to that capability, Iliad can automatically run the analysis using optimum settings. So that reduces one manual step where the user has to go in and identify, playing with those sliders, what's the best possible combination.

Instead, Iliad would feed that information to a numerical optimizer, and then the optimizer would change those analysis settings to give you the final answer. We also have a lot of additional GOE design options such as Latin hypercube, which adds the Taguchi and factorial options available inside Moldflow.

The equations are displayed to the user. So they can take those equations out of Moldflow or Iliad and keep that in the repository or for some hand calculations. So that's a plus point. There's also an option to use a dynamically evolving response service model. I'll talk about it more. But this is very helpful when you are unsure of how many design points to begin with.

We also have the biometric study option. Now, this is more extensive in terms of the number of analysis it requires. And it's mainly used for geometry modification, but it can also be used for process settings. So things such as identifying the optimal grip thickness of, Iliad does not currently support this capability. So that's something that we don't have right now.

There's also process optimization in terms of the ram speed and backing pressure profile. So Moldflow has some inbuilt algorithms that generates optimal RAM speed and packing pressure profiles, and that's also something we don't have right now. Talking about the gate location, there are two distinct features in Moldflow insight.

They include the gate region locator algorithm, which uses, the algorithm uses the geometrical characteristics of the pod and then calculates the molding feasibility and accordingly places the gates. The other option is to use the advanced gate locator algorithm. And this actually is based on no resistance.

Again, the recommendation is that after these tools have been used the user should rerun the analysis to make sure that the results are valid. So Iliad-- what Iliad can add to this particular capability is that it, since the optimizer takes into account certain key outputs, that verification is done sort of intrinsically as part of the process.

So how does the coupling between Iliad and Moldflow work? So Iliad and Moldflow interface happens mainly through the three utilities developed by Moldflow. And they are Studymod, Runstudy, and Studyrlt. So Studymod is used for modifying the study file or the analysis file. They Runstudy utility is used for running this modified study file. And then the study result is used for exporting the log file containing the results.

Now, for the study more to work an example file needs to be generated with the changed settings. And Iliad, based upon certain user inputs can automatically generate that. That means this whole loop of running this for each modification is automated. And the user, once they set up the study, can take their hands off and wait while the software runs this and gives them the results.

So I'm going to switch to the demo now. We are going to look at this technology through some use cases. First example I have here is a response service driven optimization for reducing warpage. So let me pause here and talk about why we're doing response services. So, in typical optimization you have direct evaluation. That means the optimizer would evaluate a bunch of different points, and this process is done iteratively until a certain objective is improved.

In problems such as injection molding simulations, because each analysis is so expensive in terms of the computation time and resources, an easier way to solve it would be instead of solving the regular problem, you solve an approximate problem. One way to do that is to construct a response surface and then use the equations of the response surface to drag drive to the optimum.

So this is made very simple through Iliad's interface and the different components available. So let's look at that. The example under consideration is the mouse cover, as you can see in the left hand side. And this is an in-built example in Moldflow insight. So we're only considering a single cavity in this particular case, just for simplicity of explanation. And this is a cool fill packed with warp analysis. We have two coolant lines in there and one gate.

And the problem, as I said before, is to reduce the maximum warpage using static response services. So if I were to formulate this in terms of an optimization problem, the objective is to minimize the differential warpage particularly the maximum value. And the design variables that I can play with are the coolant in the temperature and flow rate.

And I've set certain bounds. So the coolant in temperature can vary between 20 degrees Celsius and 60 degrees Celsius. And the flow rate is between 1 to 10 times by 10 to the bottom lines by 5 meters per second. So the basic parts that we need for this particular study are the Moldflow study file, which has the extension.sdy and ASCII design file with the extension.udm for reading the inputs.

So this file can be exported through Moldflow's user interface. And it's essentially a compilation in terms of T codes and T sets of all the configuration settings involved in that particular analysis file. So that's the input file, or it contains input parameters that are available for editing. And then the third file is the ASCII log file, which can be generated using study. So all the results are essentially summarized in a text format.

And when we set up the workflow, here are the steps that we are going to implement. First one is creating a representation of that workflow inside Iliad. So that consists of a number of sub steps. First one is to set up the Moldflow and Iliad integration so that all Moldflow analysis can be automated. Then we also have a macro that I recorded for exporting certain resulting plots.

So we are also going to automate the execution of that macro for each iteration. Then we will also create back up. We want to save the results and the analysis files for each of those executions. We'll construct a response service, then move that into the optimization algorithm.

So if you look at the workflow on the right hand side, you can see that the design of experiments the DOE component is housed inside the optimization component. And inside of that component you have all the components which need to be executed for each iteration, or each design point. Once that's done, we'll validate and then execute and then we can the post-process.

All right, so our starting point consist of these three files, the UDM and study. These are the three input files that we need. And additionally I have this macro, which is a Visual Basic Script. If I open this and show it to you, you can see that essentially records all the actions involved in opening up a certain Moldflow project and then exporting the results.

OK so these are the extra things that we have added in this particular study. And the plots that would be generated are going to be the deflection indicated with the warpage. And we have-- each component of the warpage is exported separately. So the x, y, and z values are separate.

So to begin this I'm going to launch Iliad from here. OK, and this is the graphical user interface. I'm going to create a new project. So I'm going to browse the location and where I want to create this, I'm going to say RSM demo.

And this colored screen that you see here is the Canvas. This is where you would recreate all the steps in your design process. So we are going to run an optimization study based on DOE component. So I'm going to the drag and place this DOE component, which is going to construct the response surface. And then I'm going to drag the more fluid component for running the Moldflow analysis, inviting component for running the macro and then a back for people.

OK, and to finish up my workflow I'm going to add a stop sign. So I go ahead and connect all these. We have certain tools to automatically connect and rearrange this so that it's more systematic and pleasing to the eyes. There we go. We have the workflow created. Now we'll go in and set up all of the components.

So let's start with the flow component, the most important. So once you switch to that, you would be transferred to the component editor tab. And in here you have all the options for running Moldflow. So do you want to run the analysis locally, remotely, or in parallel. You can choose the appropriate radio button, and depending upon that you would need to feed in more information.

So for simplicity, I'm going to stick with the local analysis. Then you need to browse the location of Runstudy.exc. Typically this is located in program files Autodesk, and there would be a folder Moldflow insight 2021. So since I have already saved this information once, Iliad would remember it and it automatically fills it using one step tool.

Now the interface works in two separate modes, whether it's process settings and gate locations or process settings and other boundary conditions. So depending upon the type of boundary conditions you use, you may need to choose one or the other option. So in this particular case, I'm not interested in the gate locations. I'm only going to look at boundary conditions. So I'm going to click on that radio button.

And then here are the three files that I mentioned before. So the study file, and automatically when you click on this button, it's going to show you a window with only those extensions filtered out. So for example, the certified can only have .sty format. So only the study file are shown in this particular box.

And similarly you would repeat the process for the UDM file, which is the design file. And this may take a few seconds because typically the UDM files are quite large, so parsing that takes some time. And then the last thing is the log file, which contains the results. In this case, I had generated using the study mold utility a file called warpresults.tsd.

OK so that's about it. And then I'm going to hit the load data button. And now you can see that the input and output parameters are all important in here and you can go ahead and choose whichever ones that you want. So in this particular case, I'm going to select the flow rate and coolant in the temperature. The flow rate and coolant in the temperature for the second line.

The important thing to note here is that the T set, the list of T set and T codes is quite huge. So we only have, by default, only a certain number of T sets and T codes are imported. But you have the flexibility to change that by modifying this particular file.

So inside the location where you created this file, Iliad generated and included T codes on that. So you could add more T codes in here and then reload the data. And then the new settings, or you could take out certain goods if you want, and the updated settings would be shown in these tables.

OK, so we got these two particular settings and you saw how easy it is to import the input and output settings. Let's go ahead and create some responses as well. So for simplicity, I'm going to just choose the z directional warpage. Now if you look at there's two values here, the min value and max value.

And it doesn't matter to me, the negative sign is not that important as the magnitude. So the key here is to reduce the magnitude. So for that particular case I need to create a synthetic variable that will take into account just the magnitude. So I'm going to create a synthetic variable called unit z. This is an added variable. I'm going to set this to be a synthetic variable.

And then I'm going to hit the Edit dependency equation. So this opens up a Python component, which is in-built inside of Iliad. And you can create your own variables. So the inputs to this particular variable are the min chance additional different warpage, and the max warpage in z direction. And then you type in conversion.

So if negative min is greater than the max value, that means the magnitude is more than your variables should be. This particular one, the negative effect, or else it will be the max [INAUDIBLE]. And you can see that for every iteration now it's going to check between those two values. And the value with the maximum magnitude would be assigned the differential warpage in the z direction.

OK, so similarly you can go ahead and create more variables for y and x. And since this will take some time I'm going to switch to something that I already completed beforehand. Let's go through this. OK, so this is something similar. And what I did, in this particular case, I created diff x and y as well.

So these were the input variables I had. And you can look at the initial value for each one of those has already been parsed. Then I have configured my macro. So in this particular case, it's going to read the macro file and I can switch to this. So I have appended the itieration number at the back of each image file so that I can differentiate between every iteration.

Now I want this value to change depending upon the iteration number. So for the first iteration, it should be 1. And then second for the second iteration, and so on. So to do that I can again automate that using a Python script. That's what I've done here. It's going to parse that file, and then change that number and then execute the particular macro. And then the third component I have here is just a backup component showing all the files that I can backup. So the three analysis files.

Then I have-- I'm going to switch to the DOE component, and through here I can control the order of my surface. So you could-- the response of this model could go from a linear polynomial to a full quadratic. Forward stepwise regression is helpful in the regard that it will take out terms which are not contributing significantly. But you can also choose a reading model.

And then you need to define the initial seed of the design of experiments that could be used to create the response service. So as I mentioned before, we have a lot of options. Using, adding, additionally we have a user defined option. So if there is past data available and you don't want to run the analysis you can directly import that to a CSV file.

In this particular case, I'm going to stick with Latin hypercube because it uniformly distributes the points and I have specified the number of points I want, and that's about it. So that's the DOE component setup. Then we'll go one level up to the optimization component. And in here you can define the variables and the objective.

So what I've done is I've taken those magnitudes of x, y, and z directions and I've added that together creating another variable all the total differential warpage, Which is essentially an addition of all these things. And I've said that to be my objective. Then for each of those variables I have set up the lower and upper bounds, and I've set up a goal for the objective.

OK, and that's about it. So if you look at the view monitors you can see that there is a reduction in the total differential warpage. Initial value in this particular case was 2.35185. Now, let's look at how much improvement can be achieved. So if I go to post-processing and select optimization, look at a table of the design points.

Let's see, we got to arrange this. You can see that compared to that particular value, the optimal point shows a value, which is about 50% less. That's a significant reduction. And this is a predicted value made by the response service model. Let's look at the response service model as well. So I have a way to interact with that.

So here we see the relationships that these input factors have with the output. So for example, if I look at how the flow rate and the temperature affect the differential in the x direction, differential warpage, you can see that it's almost like a quadratic or a second order. So the optimum volume is not actually at the lowest temperature but somewhat about 318 or 319 Kelvin.

And these are the plots and that surface is essentially fitted using a regression analysis. And again, through here you can look at how changing each of these affects the output. So you can see this red point right here moving along as I slide these, and also the changes in the output. So this equation is what is being used by the optimizer. And this is very powerful. Now you have a simple deterministic relationship looking at the output without the need for running so many additional analysis.

This is the advantage of the response of this model. OK so similar to this, we also have a dynamic response service to an optimization. So in this particular case, the response evolves. And so in this particular case, you can see that I'm just going to use the response surface model. And you can choose how many points you want to start with. Again you have control over the type or the order of the model, and this may require additional analysis. Or sometimes this may require [INAUDIBLE] anaylsis.

If I look at the results with this particular case, you can see that within four iterations I already got to a lower differential value using this combination of settings. So this is pretty advantageous. Now quickly, I also want to show you the-- so by running the macro, you can see I generated those three resulting plots for each of the deflections.

And this is one step further because instead of a used to going and changing those, Iliad is going to automatically do that. And for some results, which cannot be quantified easily, such as air traps or vent lines, that's very important. Because the Visual inspection is more important than just looking at an absolute number. Similarly, you can also do gate location optimization. And I have, again, the workflow is similar but now instead of choosing the second mode you use the first mode where you have the gate coordinates.

And again I'm going to quickly switch just to show you the results. Once again, in this similarly, I have similar to the first two problems I showed you, we have six variables. And in this particular case, I have some constraints. So I want the mass to have at least minimal value to ensure structural integrity, but I've also said limits on the maximum clamping force that can be exerted.

So similarly, we run the analysis. And if you look at the response service model, you can see we have simple deterministic relationships connecting these key input parameters to the volumetric shrinkage, in this particular case. And we saw a reduction of about 25%. Yeah 25%, in the volumetric shrinkage value.

OK, so to conclude, Iliad enhances the design exploration capabilities inside Moldflow. And this is taking those capabilities one step further and automate [INAUDIBLE] in the design. And this is done through the application of numerical optimization, and that involves using gradient and non gradient based algorithms. We have a fleet of different optimization engines, and that are robust enough to deal with different situations.

Iliad supports the integration of macros and other analysis components. So if you have, let's say, after you do the injection molding you want to do a structural analysis using ASCII Since you have an ASCII interface, everything can be controlled through a single workflow. That's a big advantage.

And the last point is, we enable quicker optimization by formulating and solving response service models instead of direct evaluation. With that, I want to say thank you. I look forward to questions and discussions with all of you. Please visit our website. You can scan the QR code shown on our screen to look at all information on Iliad. Or if you'd like to get in touch with us, please email us at Iliad.support@omniquest.org data.world.