Soft sensor modeling method based on secondary variables KNN analysis

A soft sensor modeling method based on the K-nearest neighbors method (KNN) was proposed. This method applied KNN to secondary variables classification and used the classified result, principal component analysis (KPCA) and support vector machine (SVR) to establish a model for soft measurement. KNN analysis was independent of the correlated regression model, but directly affected the model structure. Via KPCA as a middle layer, under the instruction of assorted result of the kernel function, the method was able to capture the high-ordered principal components among the secondary variables, and use SVR to establish a correlated regression model between the featured principal components and the primary variable. The proposed KKS method was used in soft sensor modeling for the end point of crude gasoline. Compared with the models of linear PLS, RBF-SVR and KPCA-SVM, the result obtained by the KNN-KPCA-SVR (KKS) approach showed better estimation accuracy and was more extendable.