基于响应面法和BP神经网络的7050铝合金腐蚀疲劳寿命预测及对比  

Prediction and Comparison of Corrosion Fatigue Life of 7050 Aluminum Alloy Based on Response Surface Method and BP Neural Network

作　　者：汲高飞 李志鹏[1] 宋贤海 JI Gaofei;LI Zhipeng;SONG Xianhai(School of Materials Science and Engineering,Nanchang Hangkong University,Nanchang 330000,China)

机构地区：[1]南昌航空大学材料科学与工程学院,南昌330000

出　　处：《机械工程材料》2024年第11期128-134,共7页Materials For Mechanical Engineering

摘　　要：对7050铝合金依次进行腐蚀和疲劳试验,获得了腐蚀后合金的疲劳寿命;分别利用响应面法和BP神经网络得到腐蚀时间、NaCl溶液浓度、加载频率、最大应力与腐蚀疲劳寿命之间的映射关系,并对该合金进行腐蚀疲劳寿命预测,比较了两种模型的预测误差。结果表明:在不同载荷条件下,两种模型的可靠性均较好,响应面模型和BP神经网络模型预测得到的腐蚀后合金的对数疲劳寿命与试验值的均方根误差分别为0.0710,0.0683,决定系数分别为0.9519,0.9980;BP神经网络模型的预测精度优于响应面模型。Corrosion and fatigue tests were conducted on 7050 aluminum alloy successively,and the fatigue life of the corroded alloy was obtained.Response surface method and BP neural network were used to obtain the mapping relationship between corrosion time,NaCl solution concentration,loading frequency,maximum stress and corrosion fatigue life.The corrosion fatigue life of the alloy was predicted,and the prediction errors of the two models were compared.The results show that the reliability of the two models was good under different load conditions.The root mean square errors of the predicted values by response surface model and BP neural network model and test values of the logarithmic fatigue life of the corroded alloy were 0.0710 and 0.0683,and the coefficients of determination were 0.9519 and 0.9980,respectively.The prediction accuracy of BP neural network model was better than that of response surface model.

关 键 词：BP神经网络 响应面法 7050铝合金 疲劳寿命预测 腐蚀疲劳 

分 类 号：TG13[一般工业技术—材料科学与工程]