Abstract: Taking the die preheating temperature, initial forging temperature, final forging temperature, cooling rate and forging speed as the input layer, and the tensile strength and yield strength as the output layer, the neural network optimization model for forging process of 6A02 aluminum alloy connecting plate was established by using 5×30×10×2 four-layer topology structure, and the model was trained, predicted and applied. The results show that the neural network optimization model of the forging process of aluminum alloy connecting plates has a relative error of tensile strength prediction between 1.54%and 3.29%, and an average prediction error is 2.03%; the predicted relative error of output yield strength is between 0.88% and 2.98%, the average predicted relative error is 1.43%, and the overall relative error is small; the mechanical properties of the forgings are increased by 15% after optimization, the optimal forging process parameters are die preheating temperature of 340℃, initial forging temperature of 460℃, and final forging temperature of 340℃, cooling rate of 50℃/s, and forging speed of 12 mm/s.