Research On Fault Diagnosis Technology Of Inter-shaft Bearing Based On Multi-source Heterogeneous Information Fusion

Inter-shaft bearing is a key component in the supporting transmission system of modern aeroengine.It works between the high and low pressure rotors of aero-engine.It has the characteristics of high speed,high load,high temperature and difficult lubrication,so it is easy to break down.Once the fault occurs,it will bring catastrophic damage to the aero-engine.The fault signal transmission path of the inter-shaft bearing is long and the fault characteristics are weak.It is difficult to identify the fault accurately by traditional fault diagnosis methods.The information fusion technology synthesizes the fault information of multi-point and multi-type sensors,which provides an effective way for fault diagnosis of inter-shaft bearings.Therefore,starting with theoretical analysis,this paper combines numerical simulation and experimental research to develop multi-source heterogeneous information fusion technology based on acoustic emission and vibration signals,so as to achieve accurate diagnosis of the fault of intermediate bearings.This research has important significance and engineering value for improving the safety and reliability of aeroengine structure.In this paper,the experimental research and numerical simulation of the four typical faults of the inter-shaft bearing are carried out,the fault mechanism of the inter-shaft bearing is further clarified,the experimental and numerical simulation data of the typical fault are obtained.Secondly,aiming at the characteristics of acoustic emission signals of inter-shaft bearings,the wavelet energy spectrum algorithm is established to extract the fault feature of acoustic emission signals.Thirdly,the hierarchical permutation entropy of fault vibration signal is extracted,and the extracted hierarchical permutation entropy is fused by local linear embedding algorithm.Finally,the improved random forest algorithm performs multi-source heterogeneous information fusion on the fault feature of acoustic emission signals and vibration signals,and realizes the fault diagnosis of the inter-shaft bearing.The specific research work and achievements in this paper are as follows:(1)A fault simulation test rig for inter-shaft bearings is built,and simulation experiments are carried out for normal state,inner ring fault,outer ring fault and rolling element fault of intershaft bearings.The vibration and acoustic emission signals of multi-speed and multi-measuring points are collected.Aiming at the low signal-to-noise ratio of vibration signal,a spatial correlation denoising algorithm is proposed to reduce the noise of vibration signal.A shortcoming of the large amount of data of the acoustic emission signal is proposed.The SKPHDS method is proposed for the acoustic emission signal.The experimental results show that the method can reduce the data dimension while retaining the critical fault information,and can be reduced to 1/2000 of the original data.(2)Based on Hertz's non-linear contact theory,a local defect dynamic model of inter-shaft bearing considering time-varying displacement excitation and elastohydrodynamic lubrication is proposed.Based on the established model,the time-frequency characteristics of typical fault signals of inter-shaft bearings are simulated and analyzed.The model is validated by experimental research and theoretical analysis.The results show that the model can accurately simulate the time-frequency characteristics of the inter-shaft bearing fault signal,and the calculation error of fault frequency is less than 1%.The established model is used to analyze the evolution law of the weak fault of the inter-shaft bearing.The variation law of peak value,absolute average value,effective value,square root amplitude,kurtosis factor,pulse factor,peak factor and waveform factor with bearing speed,load and defect width is obtained.(3)In this paper,the detection principle of acoustic emission technology is studied from the perspective of fault diagnosis application,and the mechanism of acoustic emission of intermediate bearing fault is deeply analyzed.Based on the relationship between information entropy and thermodynamic entropy,the concept of information exergy is further deduced and perfected.The calculation method of wavelet energy spectrum exergy is systematically established,which is applied to the fault feature extraction of acoustic emission signals.The wavelet energy spectrum exergy is used to fuse the acoustic emission signals collected at multiple speeds.Comparing and analyzing the wavelet energy spectrum exergy under different fault conditions and different measuring points,it is proved that the wavelet energy spectrum exergy has good separability and can accurately diagnose the inter-shaft bearing faults.(4)Aiming at the non-linearity and non-stationarity characteristics of fault vibration signals of inter-shaft bearings,a hierarchical permutation entropy algorithm is proposed and applied to feature extraction of fault signals of inter-shaft bearings.Through the comparative analysis of Gauss white noise and pink noise,the effect laws of embedding dimension,time scale,decomposition layer and data length on feature extraction are studied.It is also proved that the method can comprehensively extract the fault information of the low frequency part and the high frequency part of the fault signal.The effectiveness of the hierarchical entropy algorithm is verified by the combination of analog signal and experimental signal.The results show that HPE can accurately describe the complexity of the fault signal,and can effectively classify the four typical fault states of the inter-shaft bearing and show good stability.(5)In this paper,the advantages and disadvantages of the existing manifold learning algorithm are compared and analyzed.Based on the idea of information fusion,a local linear embedding algorithm is proposed to fuse and reduce the dimension of fault features of multiple sensors,and then to effectively explain the state of inter-shaft bearings.Based on this algorithm,the dimensionality of HPE extracted from vibration signal is reduced by fusion.The calculation results show that the fault eigenvector formed by HPE after LLE dimension reduction improves the separability of fault samples.The LLE algorithm effectively solves the problem that the dimension of fault eigenvector composed of multi-sensor HPE is too large,and contains a large amount of redundant information affecting the fault classification result.(6)In this paper,a multi-source heterogeneous information fusion algorithm based on improved random forest algorithm is proposed to make decision fusion of multi-measurement and multiclass sensor fault information.The algorithm is used to solve the problems of singularity and instability of vibration signals and acoustic emission signals when reflecting the fault information of inter-shaft bearings.The algorithm is not prone to over-fitting,has good generalization ability,and can describe the fault source from different angles to improve the confidence of the diagnostic system.The multi-source heterogeneous information fusion method is used to diagnose the fault of an aero-engine inter-shaft bearing,which proves that the method has good engineering application ability.