| Rolling bearing is the important basic component in rotating machinery.It’s also theone of main sources of mechanical equipment malfunction.Therefore, to study the faultdiagnosis of rolling bearing can identify early faults of rolling bearing as soon aspossible, and make timely maintenance treatment,and then improve the operatingperformance and service life of mechanical equipment, has the extremely vitalsignificance.This paper is based on the numerical control machine rolling bearing as theresearch object, comprehensive analysis of the bearing fault mechanism, and discussesthe advantages and disadvantages of traditional fault diagnosis methods and intelligentdiagnosis theorys. Considering the single fault information can’t fully reflect the bearingfault characteristics, the introduction of multi-source information fusion fault diagnosistechnology. Through online access to rolling bearing vibration signal and noise signal,temperature signal and inside information of the machine, with the help ofcorresponding signal processing technology, realizing the data level and feature levelfusion of multi-source information. Finally, a complete system of rolling bearing faultdiagnosis is developed. The main job concents are as follows:The paper startes from the fundamental structure of rolling bearing, studying themain failure forms and reasons of rolling bearing, and establishing the correspondingrelation between fault signs and fault reasons. Depending on the machine tool bearingfault signal characteristics, bearing size and installation location, to determine theexternal sensor selection, usage and installation; At the same time, according to thecorrelation between the internal information of the numerical control machine tool andbearing fault, researching and extracting the bearing failure correlation larger internalinformation.Since the data collection must satisfy the principle of multi-channelsynchronous acquisition, on the basis of selecting the instance of multi-channelsynchronous acquisition board-NI PCI-6143, according to the features of theacquisition signal acquisition program debugging and parameter setting, a test platformof multi-channel data acquisition system based on NI and LabVIEW is established. Aiming at the nonlinear and non-stationary characteristics of rolling bearing signal,the paper introduces the statistical theory and empirical mode decomposition (EMD)method, of which principles are produced. For the fault signal within time domain,frequency domain and time-frequency domain, realizing the feature extraction andanalysis with statistical theory and the EMD method respectively, and with the help ofthe Microsoft Access database’s data management technology, extracting all data oftime-frequency domain features, then integrating them into the feature set.Because of multi-class and massive eigenvalues complicating the diagnosis process,this paper applys the rough set, which has advantages in removing redundantinformation and extracting the key features, to accomplish fault feature attributereduction, finally realizing the variable selection. In view of shortcomings of thetraditional BP neural networks, which is easy to fall into local minimum and slowtraining speed when it is trained, introducing the RBF neural network for fault patternrecognition, finally through experimental verification, RBF network has obviousadvantages in diagnosis accuracy and convergence speed.Combined with the four parts, namely the data collection, features extraction,features filtration and pattern recognition, the paper establishes a set of rolling bearingfault diagnosis system, which is based on the theories of the rough set, the EMD and theRBF neural network, and in which Multi-source information fusion theory acts as thecore. The test proved that the diagnosis system both in the diagnosis speed, precisionand stability are superior to the traditional method of bearing fault diagnosis methods,meanwhile this system has excellent real-time and generality in fault diagnosis. |
PDF Full Text Request