Soft-Sensing Model Based on Multiple K-Nearest Neighbour Regression Algorithm

A soft-sensing modeling method is proposed based on multiple K-nearest neighbor(MKNN) regression algorithm to solve the problem that a single model has lower prediction precision.The method adopts Gaussian process to choose secondary variable for soft sensing model.Then,an adaptive affinity propagation clustering method is adopted to divide the input samples data into several groups,and sub-models are built by KNN in each group.The predictive outputs of sub-models are combined by principal components regression(PCR).The proposed MKNN method is used in soft sensing modeling of the end point of crude gasoline.Compared with single KNN modeling,the simulation results show that the algorithm has better prediction precision and generalization performance.