以耐磨钢牌号、淬火温度、淬火冷却方式、回火温度和回火冷却方式作为输入层参数,以硬度作为输出层参数,采用BP神经网络算法构建了耐磨钢热处理工艺优化的BP神经网络模型,并进行了模型的预测和应用验证。结果表明,该模型的输出参数平均相对预测误差为2.2%,具有较好的预测能力和较高的预测精度。与生产线现用工艺相比,采用BP神经网络模型优化工艺热处理后的NM360、NM400、NM500耐磨钢的磨损体积分别减小26%、26%、28%。 Taking wear-resistant steel grade, quenching temperature, quenching cooling method, tempering temperature and tempering cooling method as input layer parameters and hardness as output layer parameters, BP neural network algorithm was used to build BP neural network Model, and model prediction and application verification. The results show that the average relative prediction error of the output parameters of this model is 2.2%, which has good prediction ability and high prediction accuracy. The wear volume of NM360, NM400 and NM500 wear-resistant steel reduced by 26%, 26% and 28% respectively compared with the existing process in the production line after the heat treatment with BP neural network model.