Research On Data-driven Incipient Fault Diagnosis Method For Gearbox

With the rapid development of the Internet of Things and the Internet,mechanical equipment has gradually shifted to the direction of integration,intelligence and large-scale.Carrying out fault diagnosis of mechanical equipment is a key step to avoid major accidents in machinery and ensure the normal operation of mechanical equipment.In mechanical equipment,the gearbox is a key component that changes the mechanical speed and transmits power,and the failure of the gearbox accounts for 10% and 80% of the rotating mechanical failure and the transmission mechanical failure,respectively.Therefore,diagnosing and monitoring the health of the gearbox is critical to ensuring proper machine operation.In the actual operation of mechanical equipment,no matter how fierce and faulty the mechanical equipment is,these faults are evolved from minor faults.Therefore,it is very necessary to study the diagnosis of incipient faults in the gearbox.Based on the incipient fault of gearbox and the idea of data-driven,this paper studies two key problems of feature extraction of gearbox minor faults and fault classification of gearbox minor faults,and achieves the diagnosis of the incipient fault of gearbox.The main research work is as follows:Firstly,based on the research of the incipient fault vibration mechanism,the vibration signal is a non-stationary signal with multi-scale behavior,which is characterized by low amplitude of the vibration signal of the gearbox and easy to be masked by the system noise.The normal fault feature extraction methods do not extract useful gearbox minor fault features.In order to solve the problem of incipient fault feature extraction,this paper proposes a feature extraction method for gearbox minor faults with minimum entropy deconvolution combined with multifractal.At the same time,in order to make the characteristic representation of the incipient fault of the gearbox more comprehensive,the time domain feature and the frequency domain feature are introduced together as the characteristics of the minor gear fault based on the multi-fractal feature.This characterizes the incipient fault characteristics of the gearbox from multiple aspects,which can make the gearbox incipient fault feature more comprehensive and effective.Secondly,in order to realize the classification of the gearbox minor faults,then to diagnose the gearbox minor faults,a model for classifying the minor faults of thegearbox by using the support vector machine is proposed to avoid the lack of generalization ability of the model when using neural network training classification model.At the same time,the redundancy problem of the incipient fault characteristics of the gearbox will lead to the large training cost of the classifier and the low classification accuracy.It is proposed to select the characteristics of the incipient fault of the extracted gearbox with F-score first,then select the optimal feature combination.Eliminate redundant features and solve the problem of low classification accuracy and large computational overhead caused by feature redundancy.Finally,for the problem that the parameter setting of the support vector machine is fixed and the model classification is not accurate with the low universality,the particle swarm optimization algorithm is used to optimize the penalty coefficient and insensitive factor of the support vector machine.The problem of low classification accuracy of the model caused by fixed machine parameter setting is solved.At the end of the paper,the research work of this paper is summarized and the research on the diagnosis technology of gearbox incipient faults is prospected.