基于OPLS方法的混凝土坝小样本安全监控模型研究

OPLS method-based small sample safety monitoring model for concrete dam

在已建立的混凝土坝小样本安全监控模型基础上,采用正交偏最小二乘回归分析方法对自变量共线性问题进行研究:通过正交信号修正,去除水压、温度和时效等因素中与监测效应量正交的信息,以增强混凝土坝小样本安全监控模型的分析和解释能力。工程实例分析表明,相比于普通的偏最小二乘监控模型,所建模型在拟合和预测监测效应量变化规律和趋势、以及模型的总体解释能力等方面均有所改善,证实了正交偏最小二乘回归分析方法在混凝土坝小样本安全监控模型应用上的合理性,从而为完善混凝土坝安全监控模型理论提供参考。

On the basis of the existed small sample safety monitoring model for concrete dam, a study is made on the collinearity problem of the independent variables with the method of orthogonal partial least squares（OPLS） regression; i.e. eliminating the information among water pressure, temperature and time-dependent components those are orthogonal to the monitoring effect size through the correction of the orthogonal singles, so as to enhance the analysis and interpretation capacity of the small sample safe- ty monitoring model. The analysis on the engineering case shows that compared with the conventional partial least squares regres- sion model, the model established herein is improved to some extent in the aspects of the changing law and tendency of fitting and predicting the monitoring effect size as well as the integral interpretation capacity etc. , and then verifies the reasonability of the application of the method of orthogonal partial least squares（OPLS） regression to the small sample safety monitoring model for concrete dam, thus provides a reference for improving the safety monitoring model for concrete dam.