Soft Sensing Method Based On Support Vector Machine And Its Application

Soft sensor technology is one of the most important research field in the area of process control,and it has been widely used in industrial process control.Soft sensor technology is able to measure important process control variable which can't be measured directly using measuring equipment ,and it makes an important role to improve the quality of product and assure safety in production.The core problem of soft sensor is to construct appropriate mathematic model.Machine learning based on data mining technology is an important topic of modern artificial intelligent techniques.Statistical Learning Theory is in possession of a perfect and solid theory,and it is a small-sample statistics and concerns mainly the statistic principles when sample are limited.Support Vector Machine based on Statistical Learning Theory is a new framework for the general learning problem.Compared to other learning methods,Support Vector Machine has many advantages ,for example overcoming the local minimum ,over-fitting and better model generalization ability.This paper concentrated on the research work listed below and achieved some results.1.Based on good understanding of the Support Vector Machine theory and the Support Vector Machine algorithm,Least Square Support Vector Machine algorithm which has perfect control performance is picked out.Least Square Support Vector Machine algorithm whose constraints are nonlinear to linear not only possesses classics Support Vector Machine algorithm with overcoming the local minimum and over-fitting,but also possesses the simple solution and solving curse of dimensionality.2.Atfer the analysis and comprehension of industrial process soft sensing,we introduce the support vector machine method into the soft sensing of important treated water indicator of wastewater treatment.Aiming at the problem of predicting important treated water indicator of wastewater treatment that test is complex and takes a long time and the secondary pollution is serious,least square support vector machine method was introduced.The basic theory and algorithm of the method were presented and application of the method to predict important treated water indicator was conducted.By analyzing the simulation results,the method was faster in computation and had a good generalization ability and got the anticipated result to predict important treated water indicator.3.Through the study of parameter optimization using standard grid method,Genetic Algorithm and Particle swarm algorithm in intelligent algorithm are embedded into parameter optimization.The simulation results show that intelligent algorithm can seek parameter optimization efficiently and fast.And those three methods all can get perfect model parameter. 4.Based on standard off-line soft sensor technology,we introduce the method of real-time and on-line to soft sensor technology,and it is realistic that soft sensor model was updated in time.In conclusion,this paper applied its main content to research of support vector machine and its application on soft sensor technology.In some respects,such as support vector machine algorithm,soft sensing,parameter optimization,updating soft sensor model real-time and on-line and so on,some helpful results are achieved after the reseach work.