基于BP神经网络的蒸汽发生器用镍基合金热处理工艺的优化

Heat Treatment Process Optimization of Nickel-based Alloy for Steam Generator Based on BP Neural Network

采用4-20-2三层拓扑结构,以合金牌号、退火温度、淬火温度和回火温度作为神经单元输入层,以抗拉强度和伸长率作为输出层神经单元,构建了蒸汽发生器用镍基合金热处理BP神经网络优化模型。并对该模型进行了预测验证和应用验证。结果表明:该模型的输出层神经单元的相对预测误差均小于3.5%,具有较好的预测能力、较高的预测精度和较强的工程应用价值。与生产线现用工艺相比,采用该模型优化工艺热处理后NS315镍基合金的抗拉强度增加53 MPa、伸长率增加8%。

Taking the alloy grades, the annealing temperature, the quenching temperature and the tempering temperature as the input layer of the neural unit, and taking tensile strength and elongation as the nerve unit of the output layer, the BP neural network optimization model of heat treatment of nickel-based alloy for steam generator was built by using 4-20-2 three layers topology structure. And predictive validation and application verification of the model were carried out. The results show that the relative prediction error of the nerve unit of the output layer is below 3.5%, which has good prediction ability, high precision and strong engineering application value. Compared with the process on production line, the tensile strength and elongation of nickel-based alloy NS315 after heat treatment by the optimization based on the BP neural network model increase by 53 MPa and 8%, respectively.