Research On Compound Fault Diagnosis Method Of Rolling Bearing Based On Weak Signal Processing

Rolling bearings as the core components of large equipment in the fields of aerospace,marine gas turbine,high-speed railway,and wind turbine generator.Rolling bearings operate continuously and at high speed for a long time,which may lead to the failure of the bearings due to fatigue wear.In the complex working environment,the signal obtained by the sensor often carries more interference information.Bearing weak signal is easy to cover up in the strong background noise,and the bearing in the actual operation process once the failure is usually a variety of faults are coupled together,seriously affecting the bearing fault diagnosis and health monitoring.Therefore,systematic research on compound fault diagnosis method of rolling bearing based on weak signal processing is of great significance to avoid major accidents and maintain the intelligent and healthy operation and maintenance of mechanical equipment.In practical engineering applications,rolling bearings are affected by a variety of interference factors,making the vibration signal more complex.The fault diagnosis problem under complex working conditions is usually difficult to be solved by applying traditional fault diagnosis methods.In view of the above problems,this paper systematically studies compound fault diagnosis of rolling bearing method based on weak signal processing.It proposes a series of ways including the fault weak shock signal detection,the fault weak shock signal extraction,and the compound fault weak signal fault diagnosis.The main research work of the paper includes the following aspects.(1)In view of the fact that the vibration signals obtained by rolling bearings under complex working conditions may be mixed with random noise,harmonic interference and other multi-interference factors,leading to the coexistence of multiple faults.The research of rolling bearing fault mechanism is proposed.On the basis of studying the existing impact response model of rolling bearing single point fault,the theoretical analysis is carried out for multiple compound faults of rolling bearings.Taking the gearbox as an example,the influence of random noise interference and harmonic interference on the bearing fault signal is analyzed,and the impact models of bearing vibration signals in different fault states under multi-interference factors are established respectively.The influence of multi-interference factors on single point fault and compound fault of rolling bearing is studied.The simulation results show that the rolling bearing is easily influenced by multi-interference factors in single point and compound fault states,which makes it difficult to identify the fault frequency and isolate the fault information of the bearing accurately.(2)For the actual operation of rolling bearings in the signal transmission process caused serious signal attenuation,resulting in rolling bearing fault signal showing strong noise,weakness,nonlinearity.A multi-frequency weak signal decomposition and reconstruction method combining variational mode decomposition(VMD)and adaptive cascaded stochastic resonance system is proposed.The original signal is Hilbert transformed to obtain the envelope signal.The high pass filtered signal is fed into the adaptive cascaded stochastic resonance system(ACSRS)for signal enhancement.The quantum particle swarm algorithm is used for adaptive optimization of the parameters in the cascaded stochastic resonance system.Decomposition of the enhanced signal after adaptive cascaded stochastic resonance system using variational modal decomposition.The energy loss coefficient and correlation coefficient are used to jointly determine the position of the intrinsic mode functions(IMF).The enhanced multi-frequency weak signals are reconstructed.The experimental results show that the proposed method can transfer the high-frequency noise energy to the low-frequency characteristic signals,so that different fault features can be distributed in different frequency bands,and the amplitude of fault features can be enhanced while the multi-frequency weak signals can be separated.(3)In view of the problem that the early fault signal of rolling bearing is weak and easily affected by multi-interference factors,which makes it difficult to extract fault features,an refined composite multiscale weighted entropy(RCMWE)based weak signal feature extraction method is proposed.The time-frequency domain features constitute a multidimensional original fault feature set,and four types of feature evaluation criteria are used to evaluate the feature sensitivity.The optimal feature subset is filtered out according to the specific feature evaluation criteria,and the sensitive feature parameters are used as the weight parameters of the refined composite multiscale weighted entropy,which is weighted and combined by the Hadamard product operation to obtain the refined composite multiscale weighted entropy.The experimental results show that the weighted entropy method has the ability to correctly estimate the signal complexity at all scales and can generate enhanced feature vectors,which can effectively distinguish between normal and abnormal states,as well as better distinguish between different fault types and the degree of fault degradation.(4)In view of the problem that early compound fault diagnosis of rolling bearings is difficult to be realized,a rolling bearing weak signal compound fault diagnosis method based on graph convolutional networks(GCNs)is proposed.The graph model based on refined composite multiscale weighted entropy is constructed with Gaussian kernel function as the edge weight between nodes.A rolling bearing weak signal fault diagnosis framework of graph convolutional networks for graph tasks is designed,and a rolling bearing compound fault diagnosis method based on graph convolutional networks model with refined composite multiscale weighted entropy is constructed.The experimental verification shows that the graph model based on refined composite multiscale weighted entropy can extract the interdependent information graphs from the entropy values,and use the adjacency matrix to describe the mapping relationship between the data,which can effectively identify the single fault and compound fault of rolling bearings.(5)Fault diagnosis experiments and accelerated degradation experiments were carried out using the bearing fault prediction test bench BPS.Firstly,the main components of the test bench were introduced.Then,the experimental data were analyzed and discussed by the above three fault diagnosis methods,respectively.The analysis results show that the effectiveness of the three diagnostic methods is further verified and supplemented.The weak signal decomposition and reconstruction method combined with VMD and cascaded stochastic resonance can accurately separate the compound fault signals of bearings under multi-interference factors.The method of refined composite multiscale weighted entropy can make the bearings with different fault types and different fault degrees distributed in different entropy intervals.The compound fault diagnosis method based on graph convolutional networks is applicable to single fault and compound fault diagnosis of rolling bearings.