Multi-Task Time-Series Modeling for Fatigue–Creep Damage Prediction in Proton Exchange Membranes

Fatigue–creep damage-induced mechanical degradation has been a significant failure mode of proton exchange membrane (PEM) for fuel cells, but the damage behavior and intelligent assessment method are not well understood. A multitask time-series model is proposed to address this, which predicts fatigue–creep crack growth by incorporating in situ fatigue testing with mean stress effects. The results show that both fatigue and creep damage influence crack growth, with higher stress ratios accelerating it. The plastic zone size correlates with crack growth rate, highlighting the coupled fatigue–creep mechanisms. A damage assessment model is developed to quantify contributions from both factors. The approach, validated with K-fold cross-validation, demonstrates high predictive accuracy. SHAP analysis identifies stress ratio and stress intensity factor as key factors for creep and fatigue damage. This method improves prediction accuracy and enhances understanding of PEM degradation, offering insights for predictive maintenance and durable PEM design in fuel cells.