Research On Fault Identification And Location Of Hydraulic Pipeline Based On Phase Feature Analysis

The hydraulic pipelines normally work in high-pressure and heavy-vibration environment,in which a diversity of faults such as looseness,cracks,as well as leaks may occur during a long-term operation.Different countermeasures are necessarily required for distinct types of faults.Therefore,it is significantly vital to monitor the operational state of the hydraulic pipelines through different sensors and to extract effective features for the fault identification in time.At present,the researches on fault feature extraction have mainly given a great emphasis on time domain,frequency domain and time-frequency domain so that most of them focus on the detection of single fault,rather than the identification of fault types.In this paper,for the clamp looseness and crack on hydraulic pipelines,the valid phase characteristics are extracted from the signals originating from the vibration meter as fault features to distinguish the different characteristic manifestations of the two types of faults.Finally,the identification of fault degrees and types and the location of faults are both achieved.The main researches involve:(1)Aiming at the problem that the conservative fault feature extraction methods are insufficiently effective in the fault identification of hydraulic pipelines,a novel feature extraction method based on the instantaneous phase is proposed,which utilizes the characteristics of the instantaneous phase,with a fixed waveform and a delicate sensitivity to the changes of signals,to amplify the differences caused by faults through the Autocorrelation Scatterplot of Instantaneous Phase(ASIP).Then the diversity of ASIP is calculated as an effective index of fault feature.The method is applied to the three-dimensional vibration signal to identify different degrees of looseness and crack respectively,and each fault type can be identified effectively.Results indicate that this method provides a higher accuracy of fault recognition compared with other feature extraction methods commonly used.(2)To solve the problem of fault location on hydraulic pipelines,a method based on Multivariate Experience Modal Decomposition(MEMD)and instantaneous phase is proposed on the basis of distributed Fiber Bragg Grating strain(FBGs)signals.Firstly,MEMD is performed on the distributed FBGs signals,and the diversity of ASIP is calculated at different frequency scales for each FBGs signal.Then the corresponding weight is allocated for each decomposed component according to its energy,and all extracted phase features are weighted and summed.Finally,the amplitude information of the signal is fused into the weighted phase features to solve the problem that the phase feature is independent of the signal amplitude.The results illustrate that the pipeline faults can be identified and located effectively according to the presented weighted fusion feature vector.(3)In order to enhance the identified accuracy of fault type,a method based on phase texture image and Bag of Word(BoW)model is proposed.The phase unwrapping technique is employed to expand the wrapped phase ranging from [-?,?).The phase texture image based on the linear unwrapped phase of distributed FBGs signals is established,and a feature dictionary is constructed based on the gray gradient co-occurrence matrix.Then the constructed texture image can be represented by a BoW vector,which is seem as the input to Support Vector Machine(SVM)for training and prediction.Finally,the identification of fault type between clamp looseness and crack is completed by the presented method.