Moldflow vs. Mold Floor: When Simulation Doesn’t Match Reality
By: Jennifer Schmidt, Sr. Simulation Instructor, AIM
It can be a struggle to accurately correlate simulation results to the molding floor. Many variables can contribute to discrepancies between the two. This article highlights some of the possible points of discrepancy.
We’ll start with modeling accuracy, as this issue can be two-fold. First, is the steel actually machined identically to what is modeled in simulation? If we recall the Hagen–Poiseuille equation for pressure drop shown below:
Small variations, even within a tolerance in dimensions such as: runner diameters, gate diameters, and wall stock, can make a large difference in actual pressure losses due to the thickness and/or radius being applied exponentially. Additional issues with the mold like runner mismatch, core or plate deflection, and mold wear can also cause dimensional variation between the mold and the model.
Secondly, is the geometry being accurately represented in the software: Are said thicknesses assigned correctly? Is the part meshed at a high enough quality? Too coarse of a mesh can cause errors in the filling progression, overlooking defects like air traps, weld lines, hesitation, and race-tracking. There are software-specific guidelines specifying the minimum number of elements required through features like thin wall sections, living hinges, across gates, and where weld lines might form, to name a few. Even the type of mesh used: 3D tetrahedrals vs. a Dual Domain surface mesh, or beam runners vs. 3D tetrahedrals runners, can potentially change the results like pressure and warpage predictions.
Shear induced imbalances are often not accounted for in simulation. If the feed system is modeled using beam geometries, the flow about the axis is symmetric and shear induced imbalances won’t be predicted. Shear effects can be picked up in 3D Tetrahedral elements, but extra steps must be taken beyond standard procedure to ensure the simulation can pick up this effect. Even then, the simulation programs tend to pick up on the correct trending, but not necessarily the correct magnitude.
Modeling Processing Conditions
Another potential source for error is accounting what was modeled and run in the simulation compared to the molding floor. More often than not, the machine barrel details are not modeled into a simulation. You will see machine and screw conveyance losses on the molding floor that likely weren’t accounted for in the simulation.
Take caution on default statements like “typical machine pressure losses of 4,000 psi” or some variation thereof. We have seen examples that are less than that, and then examples that exceed 10,000 psi. The only way to know for sure is to purge through the nozzle and record the pressure. Caution should also be exercised when trying to break down pressure losses beyond the nozzle purge (ie. through sections of the feed system and parts). That is too much to unpack in this article, so I’ll kindly reference the ANTEC Paper, “Evaluation of Methodologies Utilized to Determine the Pressure Drop Throughout an Injection Mold”, by Eric Bowersox, Dave Hoffman, and Ben Ellis.
Obviously, any differences in processing conditions will impact the results. Do the fill times, temperatures, packing pressures, etc match? Is the part being molded with a profiled injection velocity vs. a constant fill speed in simulation? Is the packing profiled? Do the cycle times match? Additionally, where was the fill time applied in the simulation? Was the simulation for just the cavity and not accounting for the flow and/or pressure through the feed system? Was the cooling modeled, or was an isothermal condition used for warpage predictions? Depending on the cooling design, both warpage magnitude and trending has the potential to change. If the simulation did have cooling modeled, are the lines plumbed the same as on the molding floor? Is the cooling pressure or flow rate limited on the floor versus full turbulence modeled in the simulation?
Measuring Melt Temperature
How we measure melt temperatures on the molding floor presents another problem. The set barrel temperatures are probably not going to equal the melt temperature in the simulation. If we measure melt temperature with four different methods, we will likely get four different results. Add to that multiple people taking the measurement, and we have even more variation. The graphs below show data gathered during a Plastics Technology and Engineering (PTE) class here at AIM. The graph on the left shows the average temperature over multiple shots taken by each method. The graph on the right shows the range between the multiple students taking the measurement.
As you can see, there is a large variability from equipment to equipment, and also person to person. Both the thermal mass (shot size) and color can also influence the readings. Not matter the method you chose, it’s important to be consistent in the method as melt temperature can have a large effect on the viscosity and pressure predictions in the simulation.
Material characterization, or lack thereof, presents more opportunities for inconsistencies. Is the actual material characterized, or is a substitute resin being used. How was the substitute chosen? How close is it to the specified material? It would be difficult to answer this last question without having it characterized.
The level of characterization will also affect the end results- whether the thermal properties are single-point or multi-point data. Supplemental (family averaged) vs. measured (measured per specific grade) data for pvT and mechanicals will drive and influence warpage calculations. Does the material characterization include other information like juncture loss, D3 coefficients, and CRIMS data?
The experience level of the analyst can also play an important role. They need to have knowledge of the key inputs and meshing requirements, but also a full understanding of both the capabilities and the limitations of the software. The analyst needs to know how to interpret and apply the results, ideally coupled with practical experience in rheology and injection molding.
Simulation programs are starting to be able to predict certain molding defects, such as tiger striping and weld line strength, but these features are still quite new to the software. In general, simulation can easily pick up on probable air trap and weld line locations but tends to struggle more with true cosmetic issues like gate blush and surface finish defects.
Bringing it All Together
Simulation shows us one perfect snapshot in time. It can’t account for inherent variation in the mold and molding process, such as: poor venting, mineral buildup in the cooling channels, faulty heaters/ thermocouples, seasonal variations, moisture content, resin lot to lot variation, regrind, contamination, worn check rings, and any other items that would generally cause shot-to-shot inconsistency.
Whether discrepancies between simulation and the molding process are caused by variations in modeling or dimensions, rheology, processing, or multiple factors, it is important to remember that simulation is a tool. When used properly, in the correct context, and with proper expectations, simulation can help us make better informed decisions. To get the most out of your simulation software, consider creating internal best practices for its use coupled with a commitment to reviewing feedback comparing the analysis results to actual molding. By recording and tracking these comparisons, you gain knowledge and confidence in the data, making you smarter for your next project.