Research On Fault Signal Extraction Method Of Fan

In the production process of cement plants,high-temperature fans are one of the indispensable equipments for ventilation and dust removal.Once a wind turbine fails,it will cause the entire production process to stagnate,which will cause serious losses to the enterprise.Rolling bearings are an important component of high temperature fans and are also wearing parts.Therefore,this paper selects the rolling bearing of high temperature fan as the research object,the main contents are as follows:(1)This paper studies the principle of Local Mean Decomposition(LMD)from the perspective of time-frequency analysis.For the endpoint effect problem,the LMD method is compared with the Empirical Mode Decomposition(EMD)method.Analysis,the results show that the end effect of LMD decomposition process has much less impact on the final result than EMD.In order to further reduce the influence of the endpoint effect in LMD,the energy selection continuation method is proposed and the effectiveness of the method is verified by experiments.(2)The PF component obtained for LMD decomposition is too large for the input of the classification model,and the Singular Value Decomposition(SVD)method is introduced.SVD can compress the scale of components and has better stability and robustness.In order to verify the filtering performance of SVD,a simulation experiment was carried out.The results show that the amplitude of the curve of the noisy signal is gradually reduced after SVD processing.At the same time,its characteristics remain basically unchanged.It can be seen that the SVD method can achieve good filtering effect.(3)This paper selects the Extreme Learning Machine(ELM)algorithm with simple structure,fast training speed,high classification accuracy and good global optimization ability to classify the fault state of rolling bearings under different working conditions.The basic principle of single hidden layer feedforward neural network and ELM is studied in detail.The selection of the implicit layer activation function and the selection of ELM parameters are studied in detail.(4)For the non-stationary and nonlinear signals generated by the rolling bearing vibration,this paper proposes a fault diagnosis method based on LMD-SVD and ELM.First,the acquired vibration signal is decomposed into a series of product functions by LMD to obtain the instantaneous frequency with physical meaning.Then,the PF component is processed by SVD to compress the feature vector scale and obtain a more stable feature vector value.Finally,based on The extracted feature vectors are classified by ELM with higher computational efficiency and classification accuracy.The experimental results show that the method can effectively diagnose the rolling bearing of the wind turbine under variable working conditions.