Research On Fault Diagnosis Method Of Scraper Conveyor Gear Based On Current Characteristic Analysis

Scraper conveyor is a typical low-speed and heavy-duty machinery,the main role of transport shearer cut coal from the coal wall.The cooperative work of scraper conveyor,shearer and hydraulic support is a necessary condition to realize high efficiency in fully mechanized mining face.As an essential transport machinery in coal mining face,once the gearbox of scraper conveyor fails,it will have a bad effect on coal mining efficiency and personal safety of miners.The traditional gear fault diagnosis technology often uses vibration signal,but because the scraper conveyor working face has explosion-proof requirements and has the characteristics of small space and bad environment,it is inconvenient to install sensors to obtain vibration signal.The Motor Current Signature Analysis(MCSA)is a fault diagnosis method that does not require sensor installation and is easier to be implemented underground.However,there is often a strong current power frequency component in the current signal collected directly,which will cause the gear fault features to be covered.In addition,the load of scraper conveyor is also affected by the amount of falling coal and working surface terrain factors and fluctuate.In view of the large load fluctuation of scraper conveyor,this paper extracts the gear fault frequency band signal by wavelet packet analysis,so as to filter out the frequency band containing load fluctuation interference.In this paper,the amplitude demodulation method is used to suppress the power frequency component and enhance the gear fault characteristics.Time-frequency analysis is a signal processing method which combines the signal distribution information in time domain and frequency domain.In this paper,the current signal can be converted into a two-dimensional time-frequency image by means of time-frequency analysis,and the gear fault identification can be realized by looking for the fault features in the time-frequency image.This means that the time-frequency analysis method can transform the problem of gear fault diagnosis into the problem of image classification,which greatly facilitates the diagnosis of gear fault.At present,convolutional neural network has been applied in the field of image feature recognition very mature,and has achieved a lot of research results.Therefore,this paper proposes a multi-fault diagnosis method for gears that combines time-frequency analysis and convolutional neural network.In addition,the variation law of gear fault features under time-varying loads is studied in this paper.This time-frequency analysis method generates time-frequency images containing gear fault features under different load conditions,which provides a theoretical basis for current selection.Based on this,the wavelet time-frequency analysis and short-time Fourier transform were used to generate the gear multi-fault classification data set,which included the time-frequency images of gear under normal conditions,wear,pitting and tooth broken conditions.Finally,the diagnostic results of the two data sets are compared,and the diagnostic results of the gear diagnosis model are analyzed visually.