Research On Fault Diagnosis Method Of Planetary Gearbox Based On Adaptive Hybrid Feature

Planetary gearbox is an important part of mechanical transmission system.Because of its long-term operation in a complex and harsh environment,it is easy to cause compound faults of internal parts.At the same time,the gears are meshing with each other in motion,and the weak fault features represented by cracks are easy to appear on the meshing surface.Therefore,it is of great significance to monitor the multi fault state in the planetary gearbox and diagnose in time to ensure the safe operation of the equipment.In engineering practice,a single fault will cause the concurrency of faults,and the faults affect each other,and the forms of the faults are also different.The faults are often not a single occurrence.When a weak compound fault occurs in the planetary gearbox,the components of the vibration signal are more complex,and the nonlinear coupling of the frequency is relatively strong,which makes the diagnosis and identification more difficult.In this paper,the planetary gearbox is difficult to analyze for weak compound faults.Taking planetary gearboxes as the research object,through the analysis of traditional dynamic models,the Volterra series model is established,combined model and signal analysis,and the mixed features are extracted and weak composite Research on fault diagnosis.The main research work of the thesis has the following aspects:(1)This paper analyzes the common failure modes,causes and manifestations of planetary gearbox,establishes the dynamic model of planetary gearbox and analyzes the characteristic frequency of each component.Through the simulation analysis,it is found that the single analysis method is difficult to solve the problem of weak compound fault of planetary gearbox.(2)In order to solve the problems of strong nonlinear coupling and difficult feature extraction in weak composite fault signal,the idea of "black box" is adopted,and the Volterra series model is established by using the input and output of the system.By building the planetary gearbox Composite Fault test-bed for data collection,according to the input and output of the system,the planetary gearbox model based on Volterra series is established,the kernel function is solved,and the high-order spectrum and slice spectrum are drawn respectively for comparative analysis,which preliminarily verifies the effectiveness of the method.(3)Aiming at the disadvantages of the nonlinear model,such as the large amount of calculation and the difficulty of online real-time diagnosis,a hybrid feature classification method is proposed to identify the weak composite fault.The calculation formula and physical meaning of time-domain feature,frequency-domain feature and nonlinear feature are described and extracted.In the case of limited training samples,support vector machine is used to classify and identify weak composite faults,and the mixed features are normalized.(4)In the classification and recognition of weak compound faults using support vector machines,the parameters that need to be set manually in the support vector machines are optimized,and genetic algorithms are selected for parameter optimization,and the GA-SVM parameter optimization model is constructed,which is compared with the traditional SVM model,The optimized model analysis is more reliable.In view of the redundant features that may appear in the mixed features,the compensation distance evaluation is used to eliminate and select the mixed features.(5)Through theoretical research,experiments show that adding nonlinear features has a higher recognition rate than without nonlinear features,reflecting the necessity of nonlinear features,and showing that nonlinear features are beneficial to the identification of weak composite faults.It can better reflect the nonlinear frequency coupling characteristics,and the recognition effect of the feature set after the elimination is better,which proves the practicability of the method.