热处理工艺对电阻炉寿命影响的神经网络分析

Impact Analysis of Heat Treatment Process on Electric Resistance Furnace Worklife by Neural Network

以固溶处理温度、时间、冷却方式和时效处理温度、时间为输入层节点参数,以质量增加百分比和疲劳循环次数为输出层节点参数,构建了5×15×12×2的四层神经网络结构模型,用于分析热处理工艺对电阻炉寿命的影响,并进行了试验验证和生产线应用。结果表明,该神经网络预测模型各输出参数的相对训练误差均小于5%,相对预测误差均小于6%;与生产线上的热处理工艺相比,神经网络预测的最优热处理工艺参数可使其在600℃×72 h高温氧化后的质量增加从4.66%减少至0.97%,在600~25℃时的疲劳循环次数增加93.16%,从而明显延长电阻炉的寿命。

To analyze the effect of heat treatment process on electric resistance furnace worklife, the neural network model with 5×15×12×2 four layers was built with solution treatment temperature, solution treatment time, solution treatment cooling method, aging treated temperature and aging treated time as the parameters of input layer, and with mass increase percentage and fatigue cycle times as the parameters of output layer. The test validation and the applications of production line were carried out. The results show that the relative training error of output parameters by the neural network forecasting model is less than 5%, and the relative prediction error is less than 6%. Compared with the online heat treatment parameters, the predicted parameters based on the neural network can make mass increment after high temperature oxidation at the temperature of 600 ℃ for 72 h decrease from 4.66% to 0.97%, and make fatigue cycle times at 600 ℃-25 ℃ increase by93.16%, and prolong the worklife of electric resistance furnace life.