Fatigue life prediction and optimization of two-stage stiffness leaf spring with ANN

Fatigue life prediction of leaf springs is critical during the vehicle design stage. Fatigue life calculation based on linear superposition cannot guarantee accuracy, while fatigue life calculation based on direct transient analysis is extremely time-consuming. This paper aims to assess the fatigue life of two-stage stiffness leaf springs under random road loads with high accuracy and efficiency. The key to predicting the fatigue life of two-stage stiffness leaf springs is to calculate the full-field stress-time history based on artificial neural networks (ANN) and finite element analysis (FEA). Road loads were measured using a road test, and multi-body simulation (MBS) was performed to calculate the leaf spring force-time history. Statistical analysis was conducted on the force-time history to generate input for FEA to acquire ANN training data. The ANN was built and trained to calculate the leaf spring stress-time history using the leaf spring force-time history as input. Fatigue life calculations were performed using the stress-time history. The results show that the calculation efficiency of the stress-time history based on ANN is improved by orders of magnitude compared to transient dynamics analysis. Meanwhile, the calculated fatigue life correlates well with the test results, and different loads have different contributions to fatigue damage.