Crane Gearbox Fault Feature Extraction Based On EMD

In recent years,the impact of lifting machinery in the national economy has been growing.But its security has been widespread concern.Once the accident occurs,there is often some difficulty to identify the nature of the accident.Crane gearbox is one of the most failure-prone components,this paper focuses on studying the crane gearbox fault feature extraction methods.Gearbox is widely used in machinery and equipment drive,as the connection and power transmission component,gear wear,cracks and broken teeth failure will make the machine cannot work properly,so the timely and accurate monitoring and diagnosing gearbox failure is very necessary.Since gearbox vibration signal with nonlinear,non-stationary characteristics,when the gear failure,there is a strong background noise in it,this will affect us to diagnose the gearbox accurately.Firstly,the use of improved wavelet threshold method based on the traditional soft and hard threshold method,to reduce the acquisition gearbox vibration signal noise as a pretreatment process,due to self-adaptability of EMD,the use of EMD method to decompose the signal can effectively extract gearbox fault characteristic frequency,by spectrum analysis and combining gear fault vibration frequency modulation and its sidebands distribution characteristics to analyze gear fault diagnosis,finally combining the EMD and BP neural network,it can accurately identify the gear fault condition.In this paper,the main contents and results include:(1)Studied gear fault exhibited a common form of damage and the resulting reason,which can accurately determine the validity of this fault detection parameters;Appearing mesh frequency modulation and sidebands distribution phenomenon based on gear failure,get the relationship between typical gear failures and the corresponding characteristic frequency of vibration signal.(2)In order to suppress the noise interference in gear fault signal,highlighting fault characteristic frequency,we use an improved wavelet threshold way to reduce noise,and compare with traditional soft and hard threshold de-noising by adding noise signal simulation experiment,demonstrate the effectiveness of this de-noising method.(3)Combining the improved wavelet threshold method and EMD to analysis vibration signal.Comprehensive comparison of the time-domain waveform,amplitude spectrum,Hilbert spectrum,marginal spectrum of the gearbox vibration signal in different fault condition to obtain fault characteristic frequency and its modulation sideband feature,completing gearbox fault diagnosis and analysis successfully.(4)Finally,this paper introduces the concept of BP neural network,the use of EMD extracted the corresponding feature vectors,as training sample and testing sample of neural network.Through the BP neural network learning and recognition,and finally can be able to classify the corresponding working state and make a determination of the corresponding failure of the gearbox.Indicating that this method is suitable for gearbox fault identification.