Research On Complex Industrial Process Model Of Generalized Knowledge Based On Multi-information Fusion

In recent years multi-sensor information fusion theory and the artificial intelligence technology are widely applied in the complex industrial process,which become an important research field.This paper has done the research to the Support Vector Machine(SVM) multi-classifications class into the complex industrial process multi-information fusion's application,and studies monitoring data of the daily wastewater treatment plant.This paper introduces the relevant knowledge on information fusion,which has been studied the relevant knowledge on information fusion,the principle and conventional methods of fusion,and then discusses the complex industrial process generalized knowledge model,finally has studied the SVM classifica-tion algorithm,and deeply studies the SVM multi-class classification algori-thm,proposed the improvement decision tree SVM multi-class classifica-tion method.In the process of pattern classification on large-scale datasets, many training samples will lead to low classification speed and require high performan-ce of computer memory.The SVM's training time is a very essential performan-ce index.When the sample quantity is very large,it will take a long time during the quadratic programming process.The decision function of SVM is ultimately determined by only one set of support vectors which the classification hyperplane merely has the relationship with.Thus,the one set of support vector can fully describe the data characteristic of whole training sample set.This paper proposes that the support vector that most possibly become the samples could be found through some strategy before the training of SVM,then these samples could take place of the whole training set to carry out the training,which will surely shorten the training time and has small influence to the precision.For the SVM multi-class classification method,decision tree classification via SVM based on inter-class separability of probability distribution of vector projection is proposed.In the daily monitoring data test of wastewater treatment plants,this paper introduces the theory of SVM into information fusion , then uses three algorithms such as"one versus rest","one versus rest"and decision tree SVM algorithm improved . At last , it is successful to bring SVM into multi-information fusion,by the analyses on experimental results.In a word,the feasibility and efficiency of the proposed decision tree SVM method have been validated,the classification efficiency and accuracy has achieved a relatively ideal state.