Research On Fault Diagnosisand And Prediction Of Monitoring Equipment Based On Data Mining

The automatic water quality monitor is an intelligent equipment used for surface water quality monitoring.If equipment located in various places can realize automatic fault diagnosis and prediction functions,it will greatly reduce the maintenance cost of the equipment manufacturer,improve the manufacturer’s after-sales quality,and even help the equipment Upgrade and improvement.This article aims to realize the fault diagnosis and prediction of the water quality monitor by means of data mining.The main work contents are as follows:(1)Research on fault diagnosis method based on random forest and XGBoost.First,through experimental comparison and analysis,I found the fault classification algorithm XGBoost suitable for the data set in this paper,and then combined with random forest to perform feature selection on this basis to reduce the complexity of the final model.Aiming at the parameter optimization problem of XGBoost,the particle swarm optimization(PSO)algorithm is used for XGBoost.Aiming at the problem that the PSO algorithm is easy to fall into the local optimum,an adaptive particle swarm optimization(APSO)algorithm is proposed.This algorithm defines the evolution degree and the aggregation degree of the particle swarm,and uses them to realize the adaptive update of the inertial weight.After that,multiple test functions were used to verify the optimization ability of the APSO algorithm.The results show that compared to the standard PSO algorithm and the PSO algorithm with decreasing inertia weight,the APSO algorithm is better in terms of convergence speed and optimization results.Finally,the APSO algorithm is applied to the hyperparameter optimization of the fault diagnosis model of XGBoost,and good results have been achieved.(2)Research on fault prediction methods based on a combination of multiple models.Aiming at the prediction problem of the running state of the automatic water quality monitor,through theoretical and experimental analysis,a combined prediction method based on ARIMA model,random forest,XGBoost and LSTM neural network is studied.This method adjusts the combined weight of each sub-model according to their goodness of fit.In order to allow models with better prediction effects to occupy a greater combined weight and thus play a greater role in combined prediction,the method also uses the deformed tangent function to amplify the difference in the goodness of fit of each model.The final result shows that the combined model combines the characteristics of the above four forecasting models and shows a better forecasting effect than a single model.(3)Design of remote diagnosis and prediction platform for water quality automatic monitor.After completing the research on fault diagnosis and prediction methods,this article has carried out the architecture design,database design and back-end software design of the intelligent diagnosis platform of the automatic water quality monitor.According to the flow of data,an intelligent information system with integrating data acquisition,analysis,storage and display has been formed.