Ensemble learning-based artificial intelligence approaches for predicting low-cycle fatigue life for aluminum alloys

Abstract

This paper comprehensively analyzes the artificial intelligence (AI) models that predict low-cycle fatigue life for an aluminum alloy, considering tension–compression and torsion–torsion experiments. The study highlights the crucial need for precise fatigue life predictions to ensure the reliability and safety of structures subjected to different cyclic loading conditions. By incorporating diverse loading scenarios, the aim is to develop AI models that surpass traditional approaches’ limitations, offering enhanced predictive capabilities. This work explores the potential of AI models for fatigue life prediction, focusing on bridging the gap between traditional learning and ensemble learning predictive techniques. Its analysis improves the understanding of fatigue and contributes to developing accurate and versatile predictive tools. The ensemble learning techniques that combine the strengths of different algorithms are targeted to assess their performance. Based on the methodology, the results provide a thorough explanation of the proposed AI models for normal and shear strain-controlled cyclic loading conditions. The impact of ensemble learning methods, LightGBM, XGBoost, CatBoost, extra trees, and nearest neighbors are investigated on the accuracy of fatigue life prediction. A comparative evaluation of these AI models in terms of their predictive capabilities in different strain scenarios is presented. This approach enhances the overall efficacy of the learning system by considering a host of perspectives and drawing on the strengths of diverse models.