Font Size: a A A

Research On Electromagnetic Nondestructive Testing Technology For Grinding Burns Of Aeroengine Gears

Posted on:2020-06-04Degree:MasterType:Thesis
Country:ChinaCandidate:D HuangFull Text:PDF
GTID:2392330590477239Subject:Optical engineering
Abstract/Summary:PDF Full Text Request
The gear is the core component of the aero engine transmission system.The quality of its surface layer directly affects the function and service life of the engine.In the grinding process of gears,the cutting,scoring and sliding action of abrasive grains may cause grinding burns on the gears,results in deterioration of performance such as corrosion resistance,wear resistance and fatigue strength.,even the cracks may occur and break the teeth of the gear when severe burns occur,Serious accidents such as teeth seriously affect the flight safety of the aircraft.The factory usually uses the acid etching method to detect the burn of the gear.This method is not only affected by the operator,but also has low precision.The more unfavorable is the slight damage to the gear and the possibility of hydrogen embrittlement.Therefore,develop a kind of.Nondestructive testing methods for rapid detection and evaluation of gear surface grinding damage are particularly important.On the basis of studying the factors of gear grinding burn formation and grinding the electromagnetic properties of burned tissue,the advantages and disadvantages of eddy current,magnetic flux leakage and Barkhausen test for grinding burn detection and the detection rate of grinding burn were studied respectively.Aiming at the existing problems,the method of improving the detection rate by integrating eddy current,magnetic flux leakage and Barkhausen's detection data was proposed.Eddy current,magnetic flux leakage and Barkha usen's detection data were designed and manufactured.On the basis of Buckhausen gear comprehensive detection system and corresponding data acquisition and analysis software,three kinds of electromagnetic detection signals are normalized and processed.The analysis results of 11(946 tooth surfaces)gears show that: Due to the decrease of magnetic permeability in the burned part of grinding,the phase angle of eddy current signal decreases,the difference of magnetic permeability between the burned part and the burned part increases,which increases the leakage magnetic field intensity,the pinning effect of magnetic domain at the burned part of grinding,the noise intensity accompanied by domain deflection increases,and the root mean square value of Barkhausen noise voltage increases.Therefore,eddy current,magnetic leakage and Barkhausen noise voltage increase.Sen method can be used for grinding burn detection.When single parameter testing,the magnetic characteristic parameters of grinding burn parts are different,and the detection rate of grinding burn can not meet the requirements of high reliability in engineering practice.In this paper,a data fusion method based on the magnetic characteristic parameters of grinding burn parts is proposed.The eddy current testing,magnetic leakage testing and Barkhausen testing parameters are processed in normalization.After that,K-means clustering,Mahalanobis distance discrimination and BP neural network data fusion were carried out for the three parameters.The fusion results were as follows: K-means Clustering Fusion grinding burn judgment accuracy 97.25%,Mahalanobis distance judgment fusion grinding burn judgment accuracy 90.54%,and BP neural network data fusion judgment accuracy 94.52%,which were higher than the detection rate of single parameter detection,among which K-means clustering fusion was better than single parameter detection.The effect is the best.This paper synthesizes three kinds of electromagnetic nondestructive testing methods: Barkhausen,eddy current and magnetic flux leakage testing.Applying the data analysis method based on machine learning,the evaluation algorithm is given and verified.The traditional product detection is raised to the evaluation level,which provides a new idea for nondestructive testing of gear grinding burn.
Keywords/Search Tags:gear, grinding burn, machine learning, K-means clustering, distance discrimination, BP neural network
PDF Full Text Request
Related items