Construction of Surrogate Model for Safety Assessment of Pedestrian Crashes

Abstract

Evaluation methods instead of CAE is needed for fast design process of industrial products. Machine learning methodology is one of the most anticipated way to accelerate evaluation process in the design process. The authors have applied convolutional neural network (CNN) to regression problems. CNN has a strong ability to interpolate a nonlinear physical phenomenon, however training data should be well considered to construct a good predictor with a small number of training data sets. In this paper, we apply CNN to a prediction of lower limb impact generated by dynamic explicit FE analyses.