Nonlinear dimensionality reduction method for mannequin time-sequenced injury curves

As the key detection equipment of vehicle crash safety, dummy collects damage curve as the key index to evaluate occupant safety. Furthermore, damage curves, being a complex data form, frequently exhibit nonlinear relationships. In this study, we propose a nonlinear dimensionality reduction method utilizing an autoencoder to reduce and reconstruct the damage curves gathered from various components of the dummy. The results indicate that the nonlinear dimensionality reduction effectively captures the synthetic acceleration of the dummy head's center of gravity, achieving a high degree of reconstruction accuracy. However, the reconstruction quality for the rib displacement viscosity, displacement, and abdominal force remains unsatisfactory, indicating a need for further enhancement.