Research Of Multi-model Soft-sensor Based On Adaptive Fuzzy Kernel Clustering

In modem complicated industrial process, some variables are very hard to be measured or even cannot be measured on-line by existing instruments and sensors. Soft sensor is an effective means of implementing the on-line evaluation of these variables. The ethylene cracking process is complex, characterized of nonlinearity and time-variant properties.In order to improve the utilization of raw materials, change raw material composition according to operating parameters in real time and make the ethylene and propylene yield maximization, the cracking severity must be controlled on better values.This research is based on the current state of ethylene cracking severity online measurement and analysising the disadvantages of traditional soft sensor modeling methods. To address the problem with the complexity and volatility of Naphtha feedstock components, adaptive fuzzy kernel clustering method was developed to divide the naphtha database optimally. After establishing multiple models of least squares support vector machine, in order to improve the model accuracy and generalization ability, differential evolution algorithm was used to determine the proper parameters of LSSVM model.In this paper, to address the issue of the initial clustering centers and the number of clustering, we proposed an adaptived fuzzy kernel clustering algorithm. Finally, through the experiment of Iris data set and naphtha attribute data, we qualify the effectiveness of our new algorithm. To solve the problem of parameters selecting of LSSVM, differential evolution algorithm is proposed. We also propose specific optimization strategy to apply this algorithm. Simulation on polyester esterification rate results shows that the model has stronger ability to generalization. To address the problem of ethylene cracking severity soft-sensor modeling, we introduced the background and process of ethylene cracking process and proposed a multi-model modeling method-LSSVM After selecting auxiliary variables and leading variables through mechanism analysis soft sensor, we established each sub-model based on the condition of sub-condition in chemical process. also the switching strategy is based on weighted value. Simulation results show that the proposed model is precise.