基于多层线性模型的铝合金大气腐蚀规律研究  

Atmospheric Corrosion Rule of Aluminium Alloy Based on Hierarchical Linear Model

作　　者：周俊炎 王竟成 杨小奎[1] 王津梅[1] 周堃[1] 舒畅[1] ZHOU Jun-yan;WANG Jing-cheng;YANG Xiao-kui;WANG Jin-mei;ZHOU Kun;SHU Chang(Southwest Institute of Technology and Engineering,Chongqing 400039,China)

机构地区：[1]西南技术工程研究所,重庆400039

出　　处：《装备环境工程》2023年第6期147-154,共8页Equipment Environmental Engineering

摘　　要：目的针对不同地区铝合金大气腐蚀差异性和样本数据利用不充分的问题,构建精度更高的铝合金大气腐蚀模型,研究铝合金在不同环境中的大气腐蚀规律。方法基于多层线性模型,构建具备层次结构的腐蚀率模型。以某型号铝合金腐蚀数据为研究对象,逐步建立零模型、随机系数回归模型、完整模型探究大气腐蚀规律,并进行预测评估。结果通过交叉验证进行模型评估,多层线性模型(MSE=0.0013)优于幂函数回归(MSE=0.0055),远优于线性回归(MSE=0.0316),模型预测精度提升。多层线性模型能有效分解总方差,增强了模型的可解释性。结论多层线性模型有效结合铝合金腐蚀数据区域差异性特征,能表征大气腐蚀规律,具有一定的实用价值。The work aims to establish a more accurate atmospheric corrosion model of aluminum alloy aiming at the differ-ence of atmospheric corrosion of aluminum alloy in different regions and the insufficient use of sample data to study the atmos-pheric corrosion rule of aluminum alloy in different environments.Based on hierarchical linear model,the corrosion rate model with hierarchical structure was constructed.With corrosion data of a certain type of aluminum alloy as the research object,zero model,random coefficient regression model and complete model were established step by step to explore the atmospheric corro-sion rule and make prediction and evaluation.Model evaluation was performed through cross validation and HLM(MSE=0.0013)was superior to power function regression(MSE=0.0055)and far superior to linear regression(MSE=0.0316),im-proving the prediction accuracy.HLM could effectively decompose the total variance and enhance the interpretability of the model.The HLM combined with the regional difference characteristics of aluminum alloy corrosion data can effectively char-acterize the atmospheric corrosion rule,which has certain practical value.

关 键 词：多层线性模型 铝合金 大气腐蚀 腐蚀模型 区域差异 可解释性 

分 类 号：TG172.3[金属学及工艺—金属表面处理]