Research On Fault Feature Extraction And Fault Diagnosis Of Gear Under Variable Load Excitation

Gear is the key part of rotating machinery.The structure of the gear transmission system is complex,the working environment is bad,the gear is easy to be damaged,the fault occurs,which affects the normal operation of the rotating machinery.Therefore,there is important significance for gear fault monitoring and diagnosis.At present,the research of gear fault diagnosis is mainly aimed at the stable load conditions,but in the actual production,there is a variable load situation.Variable load excitation gear fault diagnosis technology is facing many problems.Therefore,it is necessary to study the gear fault diagnosis technology under variable load excitation.The main contents of this paper are as follows:In order to study the dynamic characteristics and dynamic mechanism of fault gear,considering the gear failure will cause the meshing stiffness and the change of the gear center of mass,the four degree of freedom dynamic model of fault gear is established.The four-order Runge-Kutta method is used to solve the gear fault model,and the vibration waveform and spectrum of the fault gear are obtained.The effects of different gear faults on the dynamic characteristics of the gears are studied by means of vibration waveform and frequency spectrum.Aiming at the problem that the gear fault feature under variable load excitation is difficult to be extracted,a method of gear fault feature extraction based on empirical mode decomposition(EMD)and fractal box dimension under variable load excitation is proposed.In this method,the EMD decomposition of the gear fault signal under the variable load excitation is carried out.The characteristics of gear fault under variable load excitation are extracted from the time domain,frequency domain,energy domain and fractal angle,which lays the foundation for gear fault classification under variable load excitation.Aiming at the problem that gear failure under variable load excitation is difficult to be classified,a gear fault classification method based on EMD and Particle Swarm Optimization Support Vector Machine(PSO-SVM)is proposed.This method firstly normalizes the gear fault characteristic parameters under the extracted variable load and then inputs them into the PSO-SVM and classifies the gear faults.The gear fault data of the large load conditions collected by the experimental platform and the gear fault data of multiple load conditions(multiple load conditions include: large load conditions and small load conditions)are verified.The experimental results show that the method can effectively classify the gear fault under the large load conditions.But,the classification of gear failure under multiple load conditions is not ideal.A fault diagnosis method based on EMD and Deep Belief Network(DBN)under multiple load conditions is proposed to solve the problem that gear fault characteristics are difficult to extract and classify under various load conditions.The method firstly removes the high frequency noise and the interference signal transmitted from other components through EMD,and then inputs the remaining gear fault signal into the deep belief network to extract and classify the gear fault feature under variable load excitation.The experimental results show that the method can effectively classify the gear faults under various load conditions.