Waste Water Fault Diagnosis Research And Software Development Based On Unbalanced Waste Water Treatment Data

Sewage treatment system is a complex biochemical reaction system with nonlinearity,large hysteresis and strong interference,which causes Sewage treatment plants sometimes have various causes of failures.If the failure can not be diagnosed in time,it will lead to a series of serious problems.Machine learning and data mining algorithms are of great significance for waste water treatment plant.However,when the classical machine learning algorithms are diagnosing the fault of the sewage treatment data,they often have difficulty identifying the sample of the fault state because of the unbalance of sewage treatment data collected by the waste water treatment plant.Therefore,there is an urgent need to research the fault diagnosis method to timely diagnose the sewage treatment system for subsequent processing.This paper takes waste water treatment as the application background.The main research contents are as follows:1.A clustering-based unbalanced data classification method,which based on the research of common clustering algorithm theory,is proposed for the unbalanced data classification problem.The experimental results show that this method can effectively improve the classification accuracy and the G-mean value of most classifiers,and it has a good effect in the fault diagnosis of waste water treatment process.2.By researching the single hidden layer feed for ward neural network,the SMOTE algorithm for sample oversampling and the Bagging ensemble framework,a SMOTE_Bagging ensemble waste water treatment fault diagnosis method based on weighted extreme learning machine is proposed for the imbalance of waste water data.The result of contrast experiment show that the method has higher classification accuracy rate for fault samples in the waste water treatment fault diagnosis,and the method also obtains higher G-mean value,which is suitable for actual sewage treatment fault diagnosis.3.The training of SMOTE_Bagging_WELM takes a long time.To address this problem,this paper further improves the ensemble framework and proposes a rapidly ensemble wastewater treatment fault diagnosis method SP-EWELM based on weighted extreme learning machine.The result of contrast experiment show that SP-EWELM has higher classification accuracy rate for fault samples in the waste water treatment fault diagnosis,and the training time required for modeling is less,which is suitable for actual sewage treatment fault diagnosis.4.In order to make realistic use of waste water treatment fault diagnosis algorithm convenient for staffs of waste water treatment plant,this paper designs and implements awaste water biochemical treatment fault diagnosis and early warning system software using Python,which has the function of water output index prediction,sewage fault diagnosis and sewage data management.The software has been proved to be reliable and easy to use,which can greatly facilitate the staff of waste water treatment plant to diagnosis the fault of waste water treatment system.