OSC-PLS法在大坝安全监控模型中的应用  

Application of OSC-PLS method in dam safety monitoring

作　　者：李波[1] 顾冲时[1] 李智录[2] 柳利利[2] 

机构地区：[1]河海大学水利水电工程学院 [2]西安理工大学水利水电学院,陕西西安710048

出　　处：《水利水电科技进展》2008年第4期4-7,共4页Advances in Science and Technology of Water Resources

基　　金：国家自然科学重点基金(50539010);国家自然科学基金(50539110);雅砻江水电开发联合研究基金(50539030-1-3);国家科技支撑计划项目(2006BAC14B03);国家自然科学基金(50579010);中国水电工程顾问集团公司科技项目(CHC-KJ-2007-02)

摘　　要：为了提高PLS大坝监控模型的可解释性,引入一种数据预处理方法——正交信号修正(OSC)法,对原始自变量进行预处理,去除和因变量无关的信号,再建立PLS监控模型。实例分析表明,OSC-PLS法只需要1个主成分就可以建立监控模型,与PLS模型相比,该模型的回归系数具有明确的物理意义,因子的重要性指标更合理。因此,OSC-PLS大坝安全监控模型结构更简单,解释性更好,具有一定的推广价值。The PLS （partial least squares） model for dam safety monitoring has excellent stability and interpretability. However, when excossive noise exists in the original data, the interpretability of the model is lessened. In order to improve the interpretability of the PLS model, the orthogonal signal correction （OSC） data preprocessing method is introduced, then the OSC- PLS model for dam safety monitoring, which can preproeess original independent variables and remove signals unrelated to dependent variables, is proposed. A case study shows that the OSC-PLS model needs only a single principal component to be established. Compared with the regression coefficients of the PLS model, those of the propesed model have clear physical meanings, and the importance index of factors is more reasonable. Therefore, the OSC-PLS model for dam safety monitoring, with a simpler structure and better interpretability, has a certain popularized value.

关 键 词：正交信号修正 偏最小二乘回归法 大坝安全 监控模型 

分 类 号：TV698.1[水利工程—水利水电工程]