基于正交试验和BP网络的2524铝合金热处理工艺优化及预测  

Optimization and Prediction of Heat Treatment Process for 2524Aluminum Alloy Based on Orthogonal Experiment-Artificial Neural Network Model

作　　者：宋新华[1,2] 陈胜迁[1] 陈立[1] 宋斌[1] 袁江[1,2] 周金鑫[1] 

机构地区：[1]张家界航空工业职业技术学院航空制造工程系 [2]湖南大学机械与运载工程学院

出　　处：《特种铸造及有色合金》2014年第5期549-551,共3页Special Casting & Nonferrous Alloys

摘　　要：对2524铝合金进行热处理正交试验,结合电导率测试,对正交试验结果进行极差分析和方差分析,并且通过人工BP神经网络模型优化,得到2524铝合金的最佳热处理制度:固溶温度为500℃,固溶时间为11h,时效温度为185℃,时效时间为5h。结果表明,该模型能较好的反映热处理工艺参数与铝合金电导率之间的内在规律,BP网络预测最大相对误差为2.19%,表明该神经网络模型具有较高的预测精度。The orthogonal experiment of heat treatment process was designed for 2524 aluminum alloy. Combined with the conductivity testing, range analysis and variance analysis of orthogonal experiment were carried out. The process was optimized by artificial BP neural network model, and the desirable heat treatment parameters of aluminum alloy were presented as follows： solid solution treatment at 515 ℃ for 11 h, and aging treatment at 185 ℃ for 5 h. The results indicate that, this model can better reflect the relationship between heat treatment process parameters and electric conductivity of aluminum alloy, and the maximum relative error of BP network forecasting is 2. 19%, showing a higher prediction accuracy.

关 键 词：正交试验 BP网络 铝合金 电导率 

分 类 号：TG166.3[金属学及工艺—热处理]