Abstract:  Based on neural network and genetic algorithm, an optimized neural network model for engineering machinery steel forging process was constructed with steel grade, heating method, forging pressure, initial forging temperature, final forging temperature, temperature of preheating, speed of forging and temperature of heating as 8 input layer parameters, hidden layer calling tansig function, output layer calling purelin function, and tensile strength, yield strength, and elongation after fracture as 3 output layer parameters. After training, prediction, and verification, it was found that the model converged after 9113 iterations, and the relative prediction errors of the model's output tensile strength, yield strength, and elongation after fracture were all less than 5%. The model has high prediction accuracy. Better predictive ability,; Compared with the current forging process of 40Cr steel used by enterprises, the use of neural network models to optimize the forging process resulted in an increase of 82MPa in tensile strength, 88MPa in yield strength, and a slight increase in elongation after fracture within a small range. The neural network and genetic algorithm optimization model also has better practical performance.