Research On Bearing Fault Prediction Method Based On Full Vector MS-EMD

Rolling bearings are widely used in rotating machinery.At the same time,it is also one of the most vulnerable parts in the entire rotating equipment.Its operating state usually directly affects the accuracy,reliability and life of the entire machine.Rolling bearing fault prediction can pre-process faults and reduce economic and safety losses,which has important practical engineering significance.The prediction of the rolling bearing frequency spectrum structure can intuitively reflect the running state of the bearing,and can qualitatively and quantitatively determine the type and degree of failure when the bearing fails.The prediction of the structure of the spectrum requires a suitable prediction method,Extreme learning machine is a single hidden layer neural network model with simple structure,fast operation speed and high learning accuracy.It is an effective multivariate data prediction model.The prediction of the frequency spectrum structure is based on the rolling bearing vibration signal,so the collection and processing of the vibration signal will directly affect the final prediction result.The full vector spectrum technology can effectively merge the homologous dual-channel information,avoid the omission of information,and more fully and accurately reflect the running state of the bearing.In the process of bearing vibration signal processing,the mask signal method can suppress the modal aliasing phenomenon to a certain extent.In order to make the decomposition effect better,this paper further improves the mask signal method.Combining the advantages of omni-vector technology and improved mask signal method,a signal processing method of omni-vector improved mask signal method is proposed,and this method is combined with the extreme learning machine model to realize the prediction of the spectrum structure.The main research work of this article is as follows:1.The empirical mode decomposition method based on the mask signal method is studied.According to the traditional empirical modal decomposition(EMD)method,modal aliasing occurs in the decomposition results,losing the original physical meaning of the components,resulting in confusing decomposition components,including incomplete information,and affecting the subsequent component fusion results And other issues.The Masking Signal method was used to process the vibration signal.The simulation signal simulation and experimental signal analysis proved that the MS-EMD method can effectively suppress the modal aliasing in the decomposition results.In order to further optimize the decomposition results,the mask signal method needs to be improved.2.The mask signal method has been improved and combined with the full vector technology.Considering that when using the mask signal method,the decomposition result is mainly affected by the amplitude of the mask signal and the mask frequency,in order to obtain the optimal solution of the parameter,the traditional calculation method is abandoned on the basis of the energy-based method,and a Bacterial foraging algorithm optimizes the mask parameters of rolling bearing fault diagnosis method.The BFA algorithm is used to optimize the amplitude and frequency of the mask signal to obtain the optimal parameter combination.Experiments show that using the parameter-optimized mask signal to process the fault signal can more effectively suppress the modal aliasing phenomenon and obtain a cleaner spectrum.Intrinsic mode function.Aiming at the problem that the single channel information is not comprehensive,the full vector spectrum technology and the improved mask signal method are combined to establish a full vector improved mask decomposition method.Experiments show that,full-vector improved mask decomposition can get a more complete and clear spectrum.3.The extreme learning machine model is studied.After analyzing that the incompleteness of the single frequency information will affect the prediction results,the multivariate extreme learning machine prediction model suitable for spectrum structure prediction is further studied.The experimental results show that the rolling bearing vibration signal pre-processed by the improved mask signal method combined with the full vector technology can realize the accurate prediction of the rolling bearing frequency structure through the multi-variable limit learning machine prediction model.