Intelligent Fault Diagnosis Of Internal Leakage In Hydraulic Cylinders Using Acoustic Emission Technology

Hydraulic system has been widely used in various industrial fields,and hydraulic cylinder is a common actuators in hydraulic system.Monitoring the working state of hydraulic cylinders and making early fault diagnosis are conducive to the operation and maintenance of hydraulic cylinders,which can avoid major economic losses and personnel safety accidents caused by the failure of hydraulic cylinders.As a common fault of hydraulic cylinders,internal leakage occurs inside hydraulic cylinders,which is difficult to observe and detect.At the same time,the working condition of the hydraulic cylinder is complex,the hydraulic cylinder moves back and forth and on-site noise is strong.Especially in the case of small leakage,the signal is weak and vulnerable to interference from external noise,resulting in fault information being submerged in the noise,which brings great challenges and difficulties to the fault diagnosis of hydraulic cylinder internal leakage.Therefore,intelligent fault diagnosis of internal leakage in hydraulic cylinders has great practical significance for reliable operation and maintenance of hydraulic equipment.In this paper,the research on intelligent fault diagnosis of hydraulic cylinder leakage using acoustic emission(AE)technology under complex working conditions is carried out.The main contents of this thesis includes:(1)Considering the weak,non-stationary and nonlinear acoustic emission signals of internal leakage in hydraulic cylinders under different working conditions,as well as the problems of strong noise and small fault samples on site,this paper proposes a threshold noise reduction method based on Complete EEMD with Adaptive Noise(CEEMDAN)and determines the optimal threshold with the highest accuracy.Multivariate domain features are extracted from noise reduction signals,which include time domain features,frequency domain features and entropy features.In order to improve algorithm efficiency and realize data enhancement,Compensation Distance Evaluation Technology(CDET)is used to select sensitive features.Support Vector Machine(SVM)is trained with small sample data to intelligently identify fault types.Compared with other methods,this model has higher classification accuracy under different working conditions,the experimental result verifies the effectiveness of this method.(2)As internal leakage signals have the characteristics of complexity,diversity and coupling,and different working conditions also reduce the difference between classes.Moreover,the large amount of sensor data collected on site makes it difficult for the traditional shallow neural network to classify accurately.Meanwhile,considering that Convolutional Neural Network(CNN)only conducts convolution to reduce the number of parameters,CNN has insufficient feature extraction ability for signals under complex working conditions.An intelligent fault diagnosis method based on time-frequency maps and convolutional neural network(CNN)is proposed.CEEMDAN and Hilbert Huang Transform(HHT)are used to obtain time-frequency grayscale maps,which not only preserve the characteristics of the original acoustic emission signal in time and frequency domains,but also convert one-dimensional data into two-dimensional data.Apply convolutional neural network(CNN)to mine the fault characteristics of hydraulic cylinder internal leakage from time-frequency maps and output the diagnosis results.Experimental results show that compared with other intelligent fault diagnosis methods,this method can obtain higher accuracy and faster convergence speed,which shows that the model has strong robustness to load changes.(3)Aiming at the on-line diagnosis model for on-site hydraulic cylinder leakage requires good real-time performance,an intelligent fault diagnosis method based on optimized Deep Belief Network(DBN)is proposed to achieve end-to-end hydraulic cylinder internal leakage diagnosis,which extracts features from the original signal adaptively and outputs the diagnosis results.The hybrid Particle Swarm Optimization-Simulated Annealing algorithm(PSOSA)is used to obtain the optimization parameters of DBN model to avoid the influence of hyper-parameters on recognition accuracy.Experimental results show that this method can effectively improve the diagnostic accuracy and model adaptability,which shows good real-time performance.(4)In view of the fact that single sensor information is insufficient to fully reflect the fault characteristics of internal leakage,and that multi-sensor data fusion rules depend on manual design,a diagnosis method of hydraulic cylinder internal leakage based on deep learning and multi-sensor feature fusion is proposed;The unsupervised DBN is used to extract the adaptive features of multi-sensor;The Bi-directional Long Short Term Memory(Bi LSTM)model is used for multi-sensor feature fusion and internal leakage fault recognition.This model can better mine the fault features inherent in the time series under time-varying working conditions;The experimental results show that the method can more accurately diagnose the hydraulic cylinder leakage under time-varying working conditions.(5)In view of the problem the single sensor information is insufficient to reflect the fault characteristics and the multi-sensor data fusion rules depend on manual design,a method based on deep learning and multi-sensor feature fusion for internal leakage diagnosis in hydraulic cylinders is proposed.The unsupervised DBN is used to realize adaptive feature extraction.The Bi-directional Long-term Short-Term Memory(Bi LSTM)model is proposed for multi-sensor feature fusion and internal leakage levels identification,which can mine the inherent characteristics in the time series under time-varying working conditions.Experimental results show that this method can effectively realize multi-sensor feature fusion and fault identification under complex variable loads.Experimental results show that this method can more accurately diagnose hydraulic cylinder internal leakage under time-varying working conditions.