Soft sensing and optimization of pesticide waste incinerator

Abstract

Three soft sensor models [radial basis function (RBF), support vector machine (SVM), and independent component analysis–support vector machine (ICA–SVM)] are developed to infer the Chemical Oxygen Demand (COD) of the quench water produced from the pesticide waste incinerator, respectively. An optimization model of COD is further proposed based on the aforementioned soft sensor models. Furthermore, chaos genetic algorithm is introduced to solve the optimization model. The procedure is demonstrated and discussed with the practical industrial cases, where the mean relative error of the proposed ICA–SVM model for COD prediction is 0.16%, and the mean COD of the practical factory is decreased from original 1140 mg/L to current 393 mg/L, decreasing 65.53%, with the proposed optimal soft sensing approach. © 2012 Curtin University of Technology and John Wiley & Sons, Ltd.