A Hybrid Genetic Algorithm–Particle Filter for Fatigue Crack Propagation Prediction

Fatigue crack propagation prediction plays a critical role in structural health monitoring and remaining useful life (RUL) assessment of engineering structures. However, conventional particle filter (PF) algorithms may suffer from particle impoverishment and insufficient particle diversity, which can adversely affect prediction accuracy and stability. To address these limitations, a hybrid genetic algorithm–particle filter (GA-PF) is developed for fatigue crack propagation prediction by incorporating genetic operations, including selection, crossover, and mutation, into the PF framework to optimize particle distribution and enhance global search capability. The proposed method is evaluated using fatigue crack growth experimental data, and its performance is compared with that of the conventional PF algorithm. The results show that the GA-PF method achieves improved prediction performance for fatigue crack propagation and remaining useful life estimation. At 255,000 cycles, the GA-PF algorithm predicted a median RUL of 22,500 cycles, with a relative RUL error of 9.04%, whereas the conventional PF algorithm resulted in a relative RUL error of 45.41%. These results indicate the potential benefit of introducing genetic optimization into the particle filter framework for fatigue crack propagation prediction. The findings further suggest that the hybrid GA-PF method can improve predictive performance compared with conventional PF on the tested dataset.