FATIGUE LIFE PREDICTION AND PERFORMANCE EVALUATION OF SMALL MECHANICAL TRANSMISSION SHAFTS BASED ON 3D SIMULATION AND LSTM MODELING

Abstract

This study proposes a fatigue life prediction and performance evaluation method for small mechanical transmission shafts by integrating 3D simulation with LSTM modeling. First, a three-dimensional finite element model of the transmission shaft is established, and structural and fatigue simulations are performed under different loading conditions to obtain key response features, including equivalent stress, deformation, fatigue life, and safety factor. The simulation results indicate that the shoulder fillet and keyway root are the main fatigue-critical regions, where local stress concentration significantly reduces fatigue life. Then, the operating parameters and structural response features obtained from finite element simulation are constructed into a time-series dataset, and an LSTM model is developed to learn the nonlinear dynamic relationship among loading history, stress response, and fatigue life. The results show that the LSTM model achieves high prediction accuracy and outperforms BP neural network, SVR, Random Forest, and GRU models. Feature sensitivity analysis further demonstrates that simulation-derived stress, strain, and damage features can substantially improve prediction performance. This study suggests that the integration of 3D simulation and LSTM modeling provides an effective framework for fatigue life assessment, structural optimization, and maintenance decision-making of small mechanical transmission shafts.