Study On Fault Diagnosis Of Hydraulic System For Shield Machine's Cutterhead Drive System

Since the shield machine is always under poor working conditions and heavy impact,and its load varies randomly,the hydraulic system of it becomes the part with high fault rate.Once there is a fault and it cannot be cleared timely and effectively,the construction schedule will be delayed,thus lead to irreparable economic losses or even major casualties.Currently,the fault diagnosis of the hydraulic system for shield machine is mainly based on single sensor detection and heavily relied on manual work,which leads to low efficiency and accuracy of diagnosis.Therefore,a new fault diagnosis method combined with PCA and SVM is proposed.Focused on the hydraulic system of cutterhead drive,applied with theoretical research,simulation modeling and experimental analysis,the study is aimed to reduce the dependence on maintenance personnel and realize the automation and intelligence of the hydraulic system for shield machine.Major works accomplished in this study are listed as follows.(1)Based on the analysis on principles of cutterhead drive hydraulic system,the mechanism and characteristics of its common fault are summarized,the mathematical model of the key components is established,and the relationship between the model parameters and the fault characteristics is revealed,which provides the theoretical basis for fault simulation.(2)The AMESim simulation model of the cutterhead drive hydraulic system is established.Based on the actual hydraulic components and system,detailed parameters of the simulation model are set,and the data points are determined.Besides,simulation on typical faults is carried out,and the corresponding values are collected to provide sample data for fault diagnosis.(3)By the improved PCA,the fault detection is carried out on the sample data of the simulation model.During this process,the feature parameters are reduced,the redundant information is removed,and the correlation between the features is relieved.Then,the sample is extracted as the input of SVM to train after the PCA.In the end,the SVM classifier is used to identify the fault type.(4)A trial for this method is carried out.It is proved that the fault diagnosis method proposed by this study is feasible,and the accuracy rate of the proposed method can reach above 95%,which is of promising application prospect.