| With the advancement of science and technology, the automation degree ofmechanical equipment is higher and higher, which development trends towards complexity,Precision and high efficient. Then the function of mechanical equipment is complicated,and so does the relationship of various work units. If the failure occurred on some keyposition, which may lead to a series of chain reaction, affect the normal operation of theequipment and cause huge economic loss, even endanger human’s lives. Therefore, thefault diagnosis mechanical for equipment is great significant.When the signal is acquisited from the mechanical equipment, the sensor needs to beinstalled in a plurality of measurement points, the signal acquisited by sensors not onlycontains the signals are emitted diagnostic machines, but also includes the signals emittedby other devices nearby. Especially when multiple failures concurrent, the collected signalis often a mixture of a plurality of fault signals superimposed, it is difficult to separate themixed signals with the traditional signal processing methods, so the information of themechanical equipment can not be got completely. And Blind Source Separation(BSS) canonly use the observation signal to estimate the source signals, which can solve the problemof multiple signal aliasing separation.In this study, the independent component analysis(ICA) researched, because which isthe mainstream algorithm of the Blind Source Separation(BSS). Some indexs ofperformance are compared between FastICA algorithm and RobustICA algorithm such asthe computational complexity and robustness, the number of iterations, convergence time.The comparison show that RobustICA algorithm is better than FastICA algorithm. TheRobustICA algorithm is applied to fault diagnosis for the bearings, hydraulic pump inrotating machinery field.In this paper, the application of the RobustICA algorithm for single and multi-channelfault diagnosis is studied. In the multi-channel fault diagnosis, the inclined plate wearsingle fault, inclined plate wear and sliding wear boots composite failure of the hydraulicpump are studied, the fault characteristic frequency is extracted successfully by the RobustICA algorithm, and the effectiveness of the algorithm is verified. In single channelfault diagnosis, in order to meet conditions that the blind source separation problem needmultiple input, the ensemble empirical mode decomposition (EEMD) algorithm isintroduced. The signal can be decomposed into several intrinsic mode functions by EEMDalgorithm, then some multiple appropriate intrinsic mode components and the originalsignal compose the new input of RobustICA algorithm, the single channel problem istransformed to a multiple channel problem. This method is applied to the fault diagnosisof single fault and mixed for bearing, and the fault characteristic frequency is extractedsuccessfully, the validity of the method is verified. |
