Thermally Active Mold Volume
This Development Note highlights a new feature for casting users to be released in FLOW-3D v9.3.
When Version 9.2 was released in early 2007, many casting users were ecstatic over the speed-up in simulation times they were seeing due to the new Unstructured Memory Allocation (UMA) feature introduced in the release. At the same time, they were disappointed that they could not include heat transfer when doing filling simulations and get the same speed-up. In the upcoming release of Version 9.3, Flow Science hopes to reverse some of that disappointment.
Defining the Thermal Layer
Figure 1: Typical heat penetration depth as a function of time
for a steel die (solid line) and a sand mold (dashed line).
Recognizing that only part of the die needs to have active mesh cells in order to properly treat heat transfer, especially, for example, during the very fast filling of a high pressure die casting simulation, Flow Science's development team has come up with the concept of a "thermally active mold volume." With this new concept, users will be able to limit the number of cells in the active mesh for a filling simulation with heat transfer by defining the thickness of the thermal boundary layer for the mold component. As Figure 1 shows, this thickness can be determined as a function of filling time for different types of dies and molds.
When the user defines the thickness of the thermal layer, only the computational cells in this layer will be included in active mesh; all others will be excluded in the same fashion as other blocked cells with UMA. If the user does not define a thickness, the default will be to include the entire component. Figure 2 shows the preprocessor results showing a 2D slice of the thermally-active layer for a die casting problem.
Figure 2: Pre-processor output showing thermally active layer (in gray)
surrounding the cavity (blue). Red color highlights passive cells in the die.
Speeding It Up
Figure 3 below shows how the new thermally active mold volume concept can speed up filling simulations where heat transfer is turned on. Based on Figure 1, the thickness of the thermally active volume was set a 2 mm for the filling time of 100 ms. A visual comparison (see Figure 3) shows temperature distributions during filling that reveal virtually no difference in the filling or the temperature pattern.
Figure 3: Temperature distribution in metal after 80 ms of filling. On the left, the full die was used for heat
transfer, while only a 2 mm thick layer was used in the simulation on the right. The simulation on the right ran 2.5 times faster with no loss of accuracy.
But the big difference appears in comparing runtimes. For regular meshing, which required 4 million active cells, the CPU time was 26 hours. For the unstructured mesh with the new thermally active mold volume algorithm, only 652,000 active cells were required (comprised of 354,000 open cells and 284,000 cells for a 0.5 cm thick shell) and the problem was run to completion in 10 hours.
At the present time, as mentioned above, the user must define the thermally active layer. Future plans will be to make this an automatic feature.