PREDICTION OF OPTIMIZED DESIGN UNDER DYNAMIC LOADS USING ML
2022-01-0949
03/29/2022
- Content
- The stamped components play important roles in supporting many sub-systems within engine and transmission assembly. In few cases, the stamped components will not meet the design criteria whereas, due to overdesign in few other components, there will be opportunities to reduce the mass. Hence in CAE simulations, it would require to perform multiple iterations to enhance the component design by varying multiple design parameters (such as thickness, bend radius, material, etc.,) as per allowable limit. The conventional process of design enhancements will follow unidirectional parametric changes, though it helps in meeting the design criteria, it would be very difficult to produce the best optimized design within the limited time span. With the aid of Altair-HyperMorph technics, multiple design parameters can be controlled simultaneously, and DOE analyses are performed using Altair-HyperStudy to extract simulated results corresponding to the pre-defined design parameters. These DOE results are analyzed thoroughly using various statistical tools to optimize the design. Also, the DOE results can be used in machine learning (ML) methodologies which would help in predicting the optimized results without performing the corresponding iteration. This paper describes about the usage of ML process to avoid repetitive CAE simulations and optimize the stamped components using DOE data. It illuminates the process of interpreting the results in order to optimize the stamped components and eliminate repetitive iterations which in turn helps in getting the best optimized design within the available simulation cycle time.
- Citation
- S, S., Nayak, S., and Freiman, D., "PREDICTION OF OPTIMIZED DESIGN UNDER DYNAMIC LOADS USING ML," SAE Technical Paper 2022-01-0949, 2022, .