Research On Blockage Fault Diagnosis Of Silage Harvester Based On RFNB

With the gradual improvement of the living Level of Rural Residents,the demand for milk,eggs,and meat products has also increased significantly.The development of animal husbandry and the cultivation and production of forage crops have become particularly important.Silage corn is the main feed source for cattle,sheep and other livestock,so it has a great demand for silage corn each year.Silage harvester is the harvesting and preparation machinery of silage crops,and its efficient and intelligent operation is the premise of improving productivity.The structure of silage harvester is complex.Due to repeated operation and complex and changeable crop density and terrain,the blocking failure rate is high.Traditional fault diagnosis mainly relies on the operator’s experience,the diagnosis accuracy and efficiency are relatively low and the operation status of the silage harvester cannot be accurately judged,so that operation suggestions cannot be given in time.Therefore,it is an urgent problem to diagnose the working status of the silage harvester quickly and accurately.At present,the research of silage harvester mainly focuses on the optimization of machinery and less on its fault diagnosis.The project takes the silage harvester as the research object,constructs and optimizes the fault diagnosis model for the blockage failure,and provides early warning of the blockage failure,so as to improve the work efficiency of the silage harvester.(1)Determine the input parameters of the fault diagnosis model according to the characteristics of the blockage of the silage harvester.Obtain the parameters associated with the blockage fault through the data acquisition equipment;use the Fisher Score algorithm to analyze the parameter correlation,and select the parameter with the largest correlation with the blockage fault as the input parameter of the model.The input parameters of the model are the forward speed,cutting speed,grain crusher speed and fan speed.(2)Construct a fault diagnosis model.Aiming at the problem of low classification accuracy of traditional decision trees in random forest,a multi-parameter fault diagnosis method based on improved random forest algorithm(RFNB algorithm)is proposed.The idea: when constructing a decision tree,random forest constructs a naive Bayes classifier for each node of the decision tree,and uses naive Bayes tree to replace the traditional decision tree in random forest to establish a fault diagnosis model.The parameters of the diagnosis model are optimized by grid search method to determine the best parameters of the diagnosis model.(3)Model performance evaluation.Compare and analyze the diagnosis results of improved random forest(RFNB)algorithm and TCRF algorithm,random forest(RF)algorithm,support vector machine(SVM)algorithm and logistic regression(LR)algorithm.The confusion matrix,ROC curve and generalization error are used to evaluate the performance of the RFNB algorithm.The results show that the improved random forest algorithm improves the accuracy of fault classification,has strong generalization ability,and the diagnosis method is more reliable.The model established in the experiment can accurately identify the failure state of the silage harvester,has good robustness,and provides a theoretical reference for the failure monitoring and early warning of large agricultural machinery.