Research On Fault Diagnosis Of Gear System Based On Small Experimental Platform

Gears are vital mechanical components for power transmission and motion,and are the core of power systems.Because of long-term operation under high-speed and heavy-load conditions,it is easy to be damaged and lead to gear failure,and then induce power system failure,resulting in economic losses and even casualties.The reliability of power system can be improved by monitoring the health status of gear and diagnosing the fault type of gear in time.At present,the main problems to be solved in the field of gear fault diagnosis are: 1)how to remove the noise in the fault source signal;2)how to accurately extract fault features;3)how to identify mixed faults.Around the above three problems,this paper analyzed the frequency components of gear vibration signal,and carried out the research from the aspects of signal denoising,feature extraction and fault identification.Firstly,this paper analyzes the causes of common gear failures,and establishes a simplified model of gear vibration system based on the characteristics of gear vibration.Gears are simplified into a vibrating structure composed of elastic components,mass components,and damping components.The frequency components of gear vibration signals are analyzed,laying the foundation for extracting fault features from the signals.Secondly,the time domain analysis,frequency domain analysis,and time-frequency analysis methods of signal are studied.After a lot of experiments,it is found that the wavelet decomposition method in time-frequency analysis is suitable for removing signal noise,and using appropriate wavelet basis function to denoise different fault signals can achieve better denoising effect,which is conducive to the eigenvalue extraction of fault signals.The decomposition coefficients are normalized to the fault characteristics of the vibration signals.Because the vibration signals measured from different detection points contain different characteristic information,this paper selects three detection points to measure the vibration signals,and takes the fault eigenvalues obtained from three channels as the input sample of fault identification.Finally,LabVIEW graphic programming software is selected to build the fault diagnosis software platform of gear system,and BP neural network model is trained based on data-driven.The experimental results show that the gear fault diagnosis system designed in this paper is reliable for the fault diagnosis of variable output load in a certain range,and the system can accurately identify the mixed fault of gear.