Data-driven machine learning enhanced optimisation of vehicle crashworthiness design

Prof. Dr.-Ing., apl. Professor Marcus Stoffel (Aachen)

Objective

  • Vehicle design: Interdependent complex systems
  • Importance of crashworthiness: Average 300k road accidents in Germany (Source: UNECE Publication)
  • Strict regulations standards: Euro NCAP, NHTSA for US
  • Increasing complexity- Light weight structure with complex materials
  • Increase in computational cost
  • Need for powerful yet optimized computational strategies

Data Generation

  • RL based approach needs data
  • Data: Plastic deformation, energy absorption, impact load, displacements
  • Data collection methods:
    • Automated Finite Elements (FE) simulations
    • Deep Convolution Generative Adversarial Network (DCGAN): Synthetic data

Predictive Machine Learning Framework

  • Intricated ML network:
    • FEM output data and DCGAN1 > Synthetic data
    • Synthetic data > SLNN > DCGAN2
  • DCGAN2: To predict the overall vehicle response
  • Optimized methods with improved efficiency

Contact

 

Prof. Dr.-Ing., apl. Professor Marcus Stoffel

Eilfschornsteinstraße 18, 52062 Aachen

To the top of the page