Based on Archard wear model and BP neural network-genetic algorithm, the influence laws of process parameters on die wear depth and maximum stress during the forging process of pump body parts were studied by finite element simulation. Firstly, the orthogonal test of five factors and five levels was conducted by selecting forging temperature, preheating temperature, friction coefficient, hardness and pressing speed, and the key influencing factors,namely forging temperature and friction factor were determined by grey correlation analysis. Then, the key influencing factors were randomly sampled by Latin hypercube sampling method, and the simulation was conducted by software DEFORM. Furthermore, taking forging temperature and friction coefficient as the input and wear depth and maximum stress as the output, a 2×12×2 neural network-genetic algorithm optimization model was constructed, and the optimum process parameters were determined as the forging temperature of 481.6 ℃, the friction coefficient of 0.38, the minimum wear depth of 1.44 ×10-5 mm and the maximum stress of 1362 MPa. Finally, through actual production verification, the simulated values of wear depth and maximum stress were consistent with the predicted values, with the relative errors of only 0.7% and 4.0%, and the main wear area and the stress distribution were basically consistent with the actual production, verifying the feasibility of the optimized model.