Data-driven online prediction method for dynamic wear state of worm grinding wheels
TAO Yijie
LI Guolong
LEI Jun
FU Li
NIE Shaowen
LIU Huan
Abstract：[Objective]Aiming at the problems of high complexity,numerous influencing factors,and difficulty in prediction regarding the wear state of worm grinding wheels in the gear machining of new energy vehicles,an online prediction method for the wear state of worm grinding wheels was proposed.[Methods]Firstly,based on the two-dimensional fractal dimension,the replica profile morphology of worm grinding wheels was indirectly observed,and the wear degree of worm grinding wheels was quantitatively evaluated.Secondly,a jellyfish search-optimized back propagation(JS-BP)neural network was constructed,and an online prediction strategy for the wear state was formed baesd on an offline variable-parameter test training set and online real-time process parameters.Thirdly,orthogonal tests were carried out,and the influencing factors on the wear of worm grinding wheels were ranked in descending order through multi-factor analysis of variance.Finally,based on the prediction results of the back propagation(BP)neural network and the JS-BP neural network,the effectiveness of the JS-BP neural network in wear state prediction was verified.[Results]The average error of the proposed wear state prediction method is 1.133%.The proposed method can guide the improvement of gear grinding processes for new energy vehicles and promote the enhancement of gear surface integrity.
Keywords：Gear grindingWorm grinding wheelGrinding wheel wearData-drivenOnline predictionNeural network
Publication Date：2026-04-15
Online Publishing Date：2026-09-12(First online date of this platform, not the publication date of the document)
Pages：8( 172-179 )