Deep learning methods have had a significant impact on design process in the recent past. SimAI is a deep learning-based AI platform that has shown to be very effective in approximating the behavior of fluid flow applications, especially fully developed steady state flows simulated by CFD solvers. The underlying neural networks in SimAI are very versatile and can be easily extended to structural applications as well. This study aims at demonstrating the applicability of SimAI for non-linear transient structural simulations like pedestrian protection. We start with a simple Tube Crush model to demonstrate the use of SimAI to predict the deformed shape of the Tube at any time instance. We then train a model on different Tube shapes to show SimAI’s ability to learn from non-parametric geometry. Finaly, we demonstrate how SimAI can be used to accelerate Ped-pro evaluations. The NCAC Accord model is used to generate 96 training points. This dataset is used to train a SimAI model and the resulting trained model can predict the full field hood deformation as well as the HIC value for the corresponding hit location within 10% relative error on any point on the vehicle hood. SimAI is many orders of magnitude faster in predicting the HIC than direct numerical simulation and hence can be very effective in evaluating designs upfront in the vehicle development process.
Topology optimization plays a crucial role in generating initial design concepts during the early stages of vehicle development. It is a Finite Element Analysis (FEA) based technique that helps to optimize the shape and distribution of the material in a desired packaging space. This paper explores the application of LS-TaSC and LS-DYNA for topology optimization of giga casting in a vehicle’s underbody. This paper touches on the importance of topology optimization in the design development process and elaborates how the LS-TaSC tool can be used to get directional guidance before initiating detailed design (CAD) work. BIW (Body-in-White) global static bending and torsional stiffness load cases were considered while setting up the optimization model. Various optimization setting parameters, constraints, and post-processing tools available in LS-TaSC were explored and have been elaborated in the paper. The use of LS-TaSC and LS-DYNA in this project enabled the generation of an initial giga casting design concept, indicating the critical areas where material is needed or can be removed. These design concepts were further refined by the design team using CAD tools, considering more realistic manufacturing and performance constraints.
The purpose of the Crashworthiness analysis is to assess how well a vehicle's structure can protect its occupants during a collision. This study involves transforming the vehicle Crash Model Partially into a Lumped Parameter representation using DEP MeshWorks’ - Reduced Order Modelling (ROM) technology. The ROM model shows an impressive 85-95% correlation with the original Detailed Finite Element model. The complex process of converting the Detailed Finite element model into lumped parameter representation is automated through the ROM approach. The ROM model is further parametrized with a broad ranging category of parameters involving a) shape, b) gage, c) material, d) spot welds, e) adhesives, f) seam welds, g) crush-initiators, h) reinforcements, i) darts, j) bulk-heads, k) slots/holes, l) laser welded blanking, m) ribbing and o) composite lay-up to convert it to an intelligent ‘Parametric ROM Dyna model’ using DEP MeshWorks. This parametric ROM model is integrated with a DOE (Design of Experiments) based Optimization scheme to obtain an optimized design that maximizes Crashworthiness performance and minimizes weight & hence cost. Thanks to the significantly reduced number of nodes/elements in the parametric ROM model, the entire optimization process can be completed in less than 50% of the time that detailed models would require.
Reaching and fulfilling several design and crash criteria during the development process is what makes the engineer adapt and redesign the simulation model over and over again. Ideally resulting in new simulation runs with in best case improved performance, matching the intention of the applied changes. For the more demanding case of unforeseen results which do not necessarily fit to the expectations of the actual changes, methods and a workflow are being introduced here, which allow to identify the root cause of this behavior.
Accelerate Ped-Pro assessments using SimAI