Fast Fatigue Life Prediction of Polymers Through Combined Constitutive Mathematical and AI-Based Modeling

Abstract

The prediction of fatigue life is critical in the design process, and current models offer a viable alternative to costly and time-consuming experimental fatigue testing. The constitutive fatigue model used integrates low-cycle and high-cycle fatigue behavior. This model is grounded on the concept of fatigue damage evolution and incorporates a moving endurance surface within the stress space, eliminating the need for ambiguous cycle-counting methods. An interesting observation is that many polymers exhibit macroscopic fatigue characteristics, specifically, the form of the (Formula presented.) curve similar to those observed in metals. Consequently, all fatigue model parameters were expressed in terms of the well-established Coffin–Manson–Basquin model parameters. However, the constitutive mathematical modeling itself is computationally time-consuming, particularly when applied to predict high-cycle fatigue across large design spaces. Therefore, the proposed model was utilized exclusively to generate high-quality data for training machine learning models that offer significantly improved computational efficiency. The high-cycle fatigue design of polymers and other ductile materials, traditionally dependent on expensive and time-consuming experimental methods, is now expedited through an advanced modeling framework that combines constitutive mathematical modeling with AI-based approaches.