Research On Fault Diagnosis Of Rotor Vibration Based On Neural Network

With modern machinery in the industrial production development in the direction of the high-speed and complicated structure,more and more new technology is also applied in the production of rotating machinery.Rotating machinery will inevitably have some unpredictable faults in actual operation,rotor system is a common part of rotating machinery failure,whether it can run efficiently and stably is closely related to the safety production and economic benefits of industry,therefore,it is very necessary to study the fault diagnosis method of the rotor working state of rotating machinery.Due to vibration signals often contain more fault information,this article uses the characteristics of non-stationary signals of rotor system vibration faults,takes several common typical faults that occur during rotor operation as research data,deep researches involving the diagnosis process signal analysis,feature extraction and pattern recognition methods.The main contents are as follows:Firstly,the problems existing in the traditional ensemble empirical mode decomposition are studied and improved.In the traditional ensemble empirical mode decomposition method,too much noise may cause the information in the high-frequency components to be submerged.For this,the correlation principle is first used to filter the high-frequency components of the signal obtained by decomposition,then the wavelet packet denoising is used to obtain the high frequency component,hurst index change as the choice of the layer number of wavelet packet decomposition within a certain range decomposed layers,and introduced the commonly used signal-to-noise ratio and root mean square error function to measure the effect of the improved signal decomposition.The analysis of the simulation signals shows that the improved decomposition method can make the information more obvious,and the decomposition of the artificial rotor vibration fault signals also meets the expectations,which can lay a good foundation for the subsequent fault feature extraction.Secondly,the entropy characteristics of measuring signal information are studied,and the deficiency of common signal entropy is analyzed,so that the characteristics of non-stationary and non-linear signals are presented in the form of dispersion entropy.The characteristics of the dispersion entropy are analyzed by using the non-stationary and non-linear simulation signals,which proved that it could be used as the measurement method of vibration fault signals.The component dispersion entropy values of the decomposed rotor in different states were calculated and taken as the characteristics of rotor vibration fault to facilitate the identification of the diagnosis method.Finally,the neural network method to identify fault categories is studied,and the parameters of the neural network are optimized by the optimization method.The quantum particle swarm algorithm,which has strong search traversal and can avoid local optima,is used to optimize the network parameters of the bidirectional long and short memory neural network,the appropriate best parameter combination is selected to improve the learning performance of the network,and use the optimized neural network as the rotor vibration fault diagnosis and identification model.The experimental results show that the diagnosis model can accurately identify the status of rotor vibration fault data collected on the experimental platform.Combining the combined diagnosis method in this paper with ensemble empirical mode decomposition-multiscale dispersion entropy-probabilistic neural network,improved ensemble empirical mode decomposition-singular value-probabilistic neural network,improved ensemble empirical mode decomposition-fuzzy entropy-bidirectional long and short memory neural network of three kinds of diagnostic combination method comparison.The experimental results show that the diagnostic method in this paper not only has high recognition accuracy,but also consumes less time.It can provide a certain guiding significance for the diagnosis of rotor vibration fault.