基于多层线性模型的铝合金大气腐蚀规律研究

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=0.001 3)优于幂函数回归(MSE=0.005 5)，远优于线性回归(MSE=0.031 6)，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。

Atmospheric Corrosion Rule of Aluminium Alloy Based on Hierarchical Linear Model

The work aims to establish a more accurate atmospheric corrosion model of aluminum alloy aiming at the difference of atmospheric corrosion of aluminum alloy in different regions and the insufficient use of sample data to study the atmospheric 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 corrosion rule and make prediction and evaluation. Model evaluation was performed through cross validation and HLM (MSE= 0.001 3) was superior to power function regression (MSE=0.005 5) and far superior to linear regression (MSE=0.031 6), improving 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 characterize the atmospheric corrosion rule, which has certain practical value.