CAJ | 학술논문

来自工业现场的数据往往具有非线性特性且包含离群点,利用非线性偏最小二乘法(partial least squares,PLS)建模易受离群点的影响.针对这一问题,结合径向基函数(radial basis function,RBF)网络,本文提出了一种基于投影寻踪的非线性鲁棒PLS方法.该方法首先利用RBF变换将自变量与因变间的非线性关系转化为线性关系;然后利用投影寻踪算法提取变换后自变量的鲁棒偏最小二乘法成分:最后建立鲁棒PLS成分与因变量之间的鲁棒线性回归模型.将该方法应用于湿法冶金萃余液pH值软测量建模问题,结果验证了其有效性.
래자공업현장적수거왕왕구유비선성특성차포함리군점,이용비선성편최소이승법(partial least squares,PLS)건모역수리군점적영향.침대저일문제,결합경향기함수(radial basis function,RBF)망락,본문제출료일충기우투영심종적비선성로봉PLS방법.해방법수선이용RBF변환장자변량여인변간적비선성관계전화위선성관계;연후이용투영심종산법제취변환후자변량적로봉편최소이승법성분:최후건립로봉PLS성분여인변량지간적로봉선성회귀모형.장해방법응용우습법야금췌여액pH치연측량건모문제,결과험증료기유효성.
Data from industrial field usually possess nonlinear feature and contain outliers; modeling with nonlin-ear partial least squares(PLS) method may suffer from these outliers. For this case, combining with radial basis func-tion(RBF) networks, we present a nonlinear robust PLS method based on the projection pursuit. First, the nonlinear relationship between independent and dependent variables is changed into a linear one by RBF transformation. Then, pro-jection pursuit algorithm is employed to extract the robust PLS components of transformed independent variables. Finally a robust linear regression model is established between robust PLS components and the dependent variable. Applying the method to the soft-sensor modeling for pH value of raffmate solution in hydrometallurgy, we validate the effectiveness by the results.