Soft Sensor Modeling And Parameter Optimization For On-Line Estimation Of Gelatin Concentration

The concentration of gelatin is a very important control parameters of gelatin production,but there are some problems of it.At first,the import equipment for monitoring concentration of gelatin on-line is so expensive.Then,some are difficult to maintain that it is not suitable for application.At present, the off-line sampling and monitoring method with low accuracy is applied in general. Aiming to provide practical and affordable industrial-scale control technology of extraction, we developed a soft sensor to estimate concentration of gelatin on-line based on temperature and time in the process.This thesis focuses on soft sensor modeling method based on LSSVM, by analysing mechanism of the extraction phase in gelatin production, the soft sensor method was advanced for the problem of on-line measurement of gelatin concentrations. Temperature and time are chosen as the auxiliary variables of soft-sensor modeling.The data set in this paper are collected in the work site of the Qinghai Gelatin Company and input datas were pre-processed to reduce the effect of noise and measurement delays. Based on relative literatures about soft-sensor technology, a soft sensor model based on LSSVM Was established.By comparing with soft sensor model which based on RBF, the soft-sensor model based on LSSVM was effective to measure gelatin concentration. Model parameter is a key factor which affects the model performance, so a parameter selection method based on PSO was studied to select LS-SVM parameters. Due to the local optimality problem, parameter selection method based on PSO was improved by Kmeans. Compared the experimental results between the two genetic algorithm, the improved PSO gave the better performance. Because of difference of every material, a local LSSVM modeling was applied instand of traditional global modeling method. The simulation results show that the model has more effective generalization performance and higher precision.To the request of real-time measuring, the commuIlication between PLC control system is established. In an attempt to resolve the communication between MATLAB and WinCC, in an effort to data communication based on OPC Technology.