The fuzzy neural network(FNN)is the combination product of neural network and fuzzy theory,which has both the capability of calculation ability of the neural network and the fuzzy reasoning capability of fuzzy system.It has great advantages in dealing with nonlinearity and ambiguity,and has been widely used in artificial intelligence fields such as automatic control,pattern recognition,and data mining.Similar to the modeling method of conventional artificial neural network(ANN),the FNN constructs the nonlinear mapping relationship by adjusting the parameters and structure of the network through training continuously.However,since the FNN is the combination product of traditional neural network and fuzzy theory,which makes it have more complex connection structure.How to find appropriate network parameters and structure has become a problem widely studied by scholars.Therefore,this thesis mainly focuses on these problems,and the main work is divided into the following points:(1)A weight initialization method for fuzzy neural network based on rule partition.Aiming at the problem that the connection weights of the FNN are complex and the rules are prone to be redundant in the initialization process,this thesis proposes a weight initialization method of FNN based on rule partition.Firstly,a concept of rule partition is introduced,and each fuzzy rule is partitioned.Then,the selection range of the weight is determined based on the criterion of making the neurons have the maximum activity.Finally,the fuzzy rules are divided within this range,and different fuzzy rules perform initial weight selection in different subspaces.Different subspace weight selections make the fuzzy rules have greater differences,so that the network has better performance in the training.The simulation results show that the weight initialization method based on regular partition can make FNN obtain better generalization performance.(2)Design of self-organizing Fuzzy Neural Network modeling method based on effectiveness analysis.Aiming at the problems of difficulty in determining the size of FNN and redundant rules,a self-organizing fuzzy neural network modeling method based on effectiveness analysis(SOEFNN)is proposed.Firstly,through analyzing the mapping relationship between the input layer and the membership function layer,a fuzzy rule effectiveness evaluation index is proposed to judge the importance of fuzzy rules.Then,We realize the dynamic adjustment of the FNN by splitting the fuzzy rules that contain more useful information and merging the unimportant fuzzy rules.The experimental results show that the SOEFNN model can obtain a more concise network structure and further improve the prediction accuracy of network.(3)Design of an effluent BOD soft measurement model based on SOEFNN.To solve the problem that it is difficult to predict the effluent biochemical oxygen demand concentration(BOD)in the wastewater treatment process accurately and real-time,this thesis constructs a soft measurement model based on the proposed SOEFNN.Firstly,the feature correlation and inter-feature redundancy are analyzed to select feature variables based on Minimal Redundancy Maximal Relevance(MRMR).Then a soft measurement model based on SOEFNN is constructed and applied to the prediction problem of wastewater treatment effluent.The results show that the soft measurement model based on SOEFNN has better generalization ability and can predict the effluent BOD accurately.(4)Design of the intelligent software for effluent BOD smart prediction in wastewater treatment process To solve the problem that the traditional measurement for effluent BOD takes a long time and is easily interfered with by the environment,we designed an intelligent prediction software for effluent BOD prediction based on the MATLAB GUI.The design of the software includes demand analysis and development plan,software function development and design,and the main contents include user registration and login modules,user work interface,effluent BOD measurement display.The software can achieve high BOD measurement accuracy with less time and lower cost.It realizes the rapid and accurate measurement of BOD in wastewater treatment,and has a practical significance and application value for ensuring stable and efficient operation of wastewater treatment plants. |