Modeling the strength of high-speed train wheels with finite element analysis supported artificial intelligence methods

The Finite Element Method (FEM) is a computer-aided engineering (CAE) technique that provides effective solutions for simulating various physical phenomena in numerous engineering fields. Train wheels are subject to fatigue due to extreme and variable loads encountered throughout their operational life, which is a critical concern in wheel design due to its impact on lifespan and safety. However, modeling complex geometries such as train wheels using FEM software (e.g., ANSYS) presents significant challenges in terms of computational requirements, including processor power, memory capacity, and extended calculation times. These limitations constrain the applicability of FEM, particularly in scenarios requiring rapid decision-making. This thesis introduces an Artificial Intelligence (AI)-based Surrogate Modeling (SM) approach to overcome these limitations. AI models, trained on datasets generated from FEM simulations, offer an efficient alternative by predicting fatigue life in a fraction of the time required by traditional computational methods. The key findings of our study are as follows: 1.The developed AI-based surrogate model performed fatigue analysis of train wheels approximately 90% faster than the traditional FEM approach. 2.When compared to FEM results, the surrogate model demonstrated an average deviation of only 3%, validating its high accuracy. 3.Among various AI algorithms tested, Convolutional Neural Networks (CNN) exhibited the best performance, achieving the lowest Relative Mean Absolute Error (RMAE) of 3.0%. 4.The developed model successfully captured the effects of parameters such as wheel radius, axle radius, axle width, and applied force on fatigue behavior. These results demonstrate that the AI-SM approach offers significant advantages in terms of both speed and accuracy for the fatigue analysis of train wheels. This methodology has the potential to contribute to the development of safer and more optimized wheel designs in the railway industry while substantially accelerating design and analysis processes. The comparative analysis between FEM and AI-SM approaches reveals that the latter provides a solution approximately 90% faster, highlighting its efficiency and potential for practical applications in train wheel design and fatigue analysis