Casting Simulation System ADSTEFAN
ADSTEFAN 2021 enhances the function Advanced Defect Prediction Tool (ADPT) which uses the machine learning (*1) to generate the defect prediction model.
Results by Solidification and Thermal Stress Analyses can be adopted as the target Training Data of ADPT, wehereas only Flow Analysis results was the target in 2020.
ADPT performs machine learning using modified ADSTEFAN projects with actual defect information, and generates new prediction model (called Defect DB) for the defects specified in the training data.
ADPT also performs a defect prediction using the Defect DB to output the probability of the registered defects against new prediction targets which includes not only under different casting condition such as casting temperature, but new casting design and products as well.
The prediction model is directly linked to the actual defect information. It allows us to reduce troublesome "fitting" effort, and more actual shop floor-wise defect prediction.
Schematics of the ADPT
Example of defect prediction (Cold Shut)
New calculation module "Particle Method" (*2) Analysis is added in ADSTEFAN 2021.
It enables to calculate multiple and complexly related physical phenomena occured in the casting process simultaneously.
Calculation example : Die casting
Calculation example : Gravity casting
ADSTEFAN 2021 adopts GPGPU (*3) mode to accelerate the Solidification Analysis. The GPU mode provides considerable acceleration in the execution of Solidificaiton Analysis compred with CPU by utilizing many cores in GPUs.
*1 Technology to generate a data processing model by automatic learning from Training Data as a reference.
*2 One of the numerical simulation method to solve continuum body behavior based on the Lagrangian method.
*3 Technology to apply high parallel processing ability of GPU (Graphics Processing Unit) to general purposes not only for the graphical processing.