A Fault Diagnosis Method Of Gearbox Based On Hyperdisk

Gearbox is a key component of many mechanical equipment that shoulders the important task of power transmission.Reasonable and effective fault diagnosis of the gearbox plays a vital role in ensuring the healthy operation of mechanical equipment.As an important part of intelligent fault diagnosis,the research on pattern recognition tasks is of great significance.Aiming at the shortcomings of the existing pattern recognition methods in fault diagnosis,the principle and performance of the hyperdisk(HD)classifier are studied.The hyperdisk classifier uses the hyperdisk model to approximate the class region,which is different from the convex hull(CH)and Affine Hull(AH)models.The hyperdisk can more reasonably estimate the class region.Aiming at the shortcomings of the hyperdisk model,the paper improves the hyperdisk classifier and applies it to gearbox fault diagnosis.Afterwards,in view of the high time complexity of the hyperdisk classifier which is not suitable for real-time fault diagnosis,the hyperdisk classifier was improved and applied to gearbox fault diagnosis.The research content of the thesis is carried out in the following aspects:(1)The hyperdisk model and hyperdisk classifier are studied.Research on the hyperdisc model,analyze the essence of the model in the geometric sense,and compare it with convex hull and affine hull models,and analyze the advantages and disadvantages of various geometric models.After that,the classification performance of the hyperdisc classifier is explored,and the public data sets are used for numerical experiments to verify its performance superiority.(2)A gearbox fault diagnosis method based on the extensible and displaceable hyperdisk(EDHD)classifier is proposed.Research believes that the hyperdisk model cannot be easily adjusted,so the classification performance of hyperdisk classifier needs to be further improved.At the same time,in the case of outliers in the training samples,the hyperdisk classifier has a significant performance degradation,which proves that its robustness needs to be improved.The research puts forward the extensible and displaceable hyperdisk classifier,which is an improvement of the original hyperdisk classifier and can effectively improve the above-mentioned defects.Finally,the classifier is used for gearbox fault diagnosis,and its performance advantages are proved through comparative experiments.(3)A gearbox fault diagnosis method based on robust nearest neighbor hyperdisk(RNNHD)classifier is proposed.The hyperdisk classifier is a maximum margin classifier.The acquisition of the classification hyperplane requires the solution of the complicated quadratically constrained quadratic program(QCQP),which causes a decrease in time performance.Based on this,this paper combines the hyperdisk model with the nearest neighbor classification rule,and proposes the nearest neighbor hyperdisk(NNHD).At the same time,in order to avoid the degradation of classification accuracy caused by the intersection of superdiscs,the robust nearest neighbor hyperdisk classifier is proposed.The classification performance is verified by experiments on the gear data set.