Abstract: To address the issue of gear fatigue failure in the electric drive axle transmission system of 6×4 electric heavy-duty trucks under high torque and frequent shifting conditions, this study establishes a load input model for representative operating conditions. Based on the S-N curve of 20CrMnTi steel and Miner’s linear damage accumulation theory, a multi-condition fatigue life evaluation method is developed. Finite element simulations are employed to identify stress concentration zones at the gear root, pinpointing regions most susceptible to fatigue failure. Subsequently, a multi-objective structural optimization approach combining response surface methodology and genetic algorithm is implemented to refine gear module, tooth number ratio, and tip relief coefficient, thereby reducing local stress and enhancing structural strength. Bench tests and real-road trials confirm that the optimized design extends gear fatigue life by 3.2 times, reduces gearbox noise by 4.6 dB, and lowers energy consumption per kilometre by 3.4%. The results demonstrate that the proposed optimization method offers significant engineering effectiveness and practical value in improving gear fatigue resistance and overall drivetrain performance in commercial electric trucks.