【摘要】 采用等离子熔覆法在Mn13高锰钢上制备了低碳Fe–Ni合金层。以熔覆电流、喷头移动速率、离子气流量和热处理温度作为输入参数,以冲击韧性作为输出参数,建立了BP(误差反向传播)神经网络模型和粒子群算法优化(PSO)BP神经网络模型,并跟冲击韧性与热处理温度之间的线性回归模型进行对比。结果表明,线性回归模型、BP神经网络模型和PSO-BP模型的平均相对误差分别为7.06%、6.12%和3.03%。PSO-BP模型的预测结果与实测值的误差较小。更多还原

【Abstract】 Low-carbon Fe–Ni alloy coatings were prepared on high-manganese steel Mn13 by plasma cladding. A BP(back propagation) neural network model and a particle swarm optimization(PSO) based BP neural network model were established with cladding current, sprinkler scanning velocity, ionic gas flow rate, and heat treatment temperature as input parameters, and the impact toughness as the output parameter, and compared with the linear regression model representing the relationship between impact toughness and heat treatment temperature. The results showed that the average relative error was 7.06% for the linear regression model, 6.12% for the BP neural network model, and 3.03% for the PSO-BP model which had the smallest error between the predicted result and the measured value.更多还原