Discover the newly implemented machine learning method that is built on top of Abaqus/Explicit in this SIMULIA Tech Talk.
Today, automotive crash finite element simulations are one or two orders of magnitude slower than stiffness, NVH, and ride & handling simulations.
A reasonably accurate full car crash simulation with 10 million elements will take more than three hours of simulation time, even when hundreds of cores are applied, which renders crash simulations the bottleneck for multi-disciplinary concept design optimizations.
To address this challenge, we present a newly implemented machine learning method that is built on top of Abaqus/Explicit, which has been widely adopted for crash simulations for aerospace, automotive and diversified industries and has proven to provide robust and realistic results. The machine learning method accelerates the finite element explicit crash simulations by one to two orders of magnitude and enables multi-disciplinary concept and optimization studies.
Using machine learning on the results of an extensive design of experiment (DOE) study, we have created prior corrections to the material and element formulations that enabled us to use coarser meshes in the simulations. The additionally employed Adaptive DOE helped us to bring down the required simulation time for a frontal crash optimization process for our test case from a month to less than an hour on the 3DEXPERIENCE platform.