Abstract: For the aluminum alloy welded joints in the carbody of electric multiple units (EMUs), their mechanical properties gradually degrade during long-term service due to the combined effects of natural aging and cumulative fatigue damage. To address this issue, fatigue evolution analysis of aluminum alloy welded joints was conducted based on artificial intelligence methods. High-cycle fatigue tests were performed on welded joints of aluminum alloy skirt plates of EMUs with different operating mileages and service durations, and their fatigue strengths were determined using the staircase method. Four mathematical models including polynomial regression, exponential decay, logarithmic function, and multiple linear regression, were employed to analyze the evolution law of fatigue strength of the aluminum alloy welded joints. Furthermore, a fatigue strength prediction model for aluminum alloy welded joints was proposed by integrating the generative adversarial network (GAN), whale optimization algorithm (WOA), and support vector regression (SVR). The results show that the fatigue strength of the aluminum alloy welded joints exhibits a significant nonlinear decay trend with increasing service duration and operating mileage. Among the models, the polynomial regression model most accurately describes the fatigue performance evolution under multi-factor coupling conditions. The proposed GAN-WOA-SVR model effectively accounts for multiple influencing factors and uncertainties in fatigue strength, demonstrating high prediction accuracy.