Multi Layer Perceptron Integrated Network Driven Automotive Door Impact Load Reconstruction Method

Impact load reconstruction serves as a crucial validation method for automobile side impact tests. Aiming at the problem of load reconstruction model pathology caused by sparse sensor arrays, a load reconstruction method with integrated network model of multilayer perceptron is proposed. Due to the poor performance of traditional modeling methods in identifying load techniques, a machine-learning based inverse model relationship between sensing arrays and impact loads was constructed, inspired by deep learning and migration learning. The method consists of three parts, the first part is a one-dimensional fully convolutional neural network for extracting local features, the second part is an attention mechanism that weights the feature map and highlights the important features, and the third part is a global feature extraction module that performs a nonlinear transformation on the output to generate the final prediction results. In order to verify the effectiveness of the methodology in this paper, an automotive door impact load test system was designed. The experimental results show that the traditional regularization method has an average relative error of 20% and a peak relative error of 1.00%, while the integrated multilayer perceptron method has an average relative error of 5.47% and a peak relative error of 0.15%. This indicates that the method of this paper has a better reconstruction effect on the reconstruction of the automotive door impact load.