Research On The Analysis Of Factors Affecting Hobbing Quality And Quality Prediction Methods

Gear is one of the most core key basic components in the equipment manufacturing industry,but also one of the most widely used components in manufacturing equipment,its processing accuracy and efficiency represents a country’s basic manufacturing level,but the gear processing process of small deformation and process stability control is relatively complex,which has both design factors,the impact of random factors,and manufacturing factors.As the most common gear processing method,it is important to analyze the influence mechanism of hobbing process parameters on quality inspection indicators and predict the quality of hobbing gears to improve gear processing quality,processing efficiency and quality inspection efficiency.This paper takes hobbing as the research object,combines the theories of fuzzy rough set,multi-objective differential evolutionary algorithm and Gaussian mixed regression,analyzes the correlation mechanism between hobbing process parameters and quality inspection parameters and extracts the characteristic process parameters,and further conducts in-depth research on gear quality prediction.The main research contents and works are as follows.First,aiming at the problem that the hobbing quality inspection indexes are complicated and the quality inspection efficiency is low,and the gear processing industry has not fully explored the correlation mechanism between process parameters and quality inspection parameters,this paper analyzed and studied the internal correlation relationship of hobbing quality inspection indexes,and proposed a quality inspection index analysis method based on density peak clustering(DPCA)and improved multithreshold Birch clustering(IBirch).Firstly,the internal correlation of quality inspection indicators are analyzed,and the correlation matrix of quality inspection indexes is obtained by fully mining the correlation of quality inspection indicators.On this basis,the quality inspection indicators are analyzed based on density peak clustering,and a group of relatively independent low dimensional quality inspection indexes(low dimensional quality inspection indicators of quality essence dimension)which can reflect the real processing quality of gears are obtained.On the basis of effectively reflecting the quality of hobbing gear,the dimension of quality inspection parameters is reduced,and the efficiency of gear quality detection and analysis is improved.Based on the improved multi threshold Birch clustering,the relatively independent quality inspection indicators obtained under different process parameters are clustered,and the cluster label is obtained,which provides technical support for further feature process parameter reduction.Then,in response to the current problem that the influencing factors of gear quality in hobbing processing are unknown and the selection of process parameters for gear processing quality influencing factor analysis is based on empirical selection without systematic theoretical support and no research on the selection of parameters according to their importance.Carry out the analysis of the mechanism of the influence of gear process parameters on machining quality,and propose a process parameter reduction method based on fuzzy rough set(FRS)and improved multi-objective differential evolution algorithm(IMODE).This method takes the reduction number and dependency error of the parameters as a separate objective function,which avoids the problem that the previous reduction algorithm only considers the reduction dependency error but does not consider the number of reduction parameters.The reduction problem of gear process parameters and design parameters is transformed into a multi-objective optimization problem,and the neighborhood mutation,dynamic cross adjustment mechanism and target dominated sorting mechanism are introduced.It can avoid the premature of the traditional multiobjective optimization algorithm in the late iteration,which affects the diversity and uniformity of the solution set distribution.It can also avoid the problem of ignoring the excellent infeasible solution in the traditional direct control selection strategy,so as to select a reasonable search range and direction,and make the Pareto optimal solution evolve towards the desired goal constraint direction.Finally,in the case of multi process parameters and multi quality indicators,the feature process parameters which have great influence on the quality of gear hobbing are extracted,and the importance of different process parameters on the comprehensive quality indicator is quantified.It provides technical support and theoretical basis for further gear quality prediction modeling.Finally,the hobbing process involves many influencing factors and contains random and implicit variables,thus accurately predicting gear machining errors is a challenging research point.To address this problem,we analyze the quality influencing factors involved in the gear machining process,make a bold guess that the hobbing quality inspection data obeys a certain normal distribution,and propose a hobbing quality prediction model based on improved variational inferential Gaussian mixed regression(IVIGMR).Based on the obtained characteristic process parameters and quality inspection parameters,a Gaussian mixed regression gear quality prediction model is established,and a variable inferred Gaussian mixed regression is introduced to train the gear error regression model,which avoids the problems that the traditional Gaussian mixed regression model is easy to fall into model selection,information loss limits the improvement of prediction accuracy,and the EM algorithm is difficult to guarantee that the obtained parameters are globally optimal solutions.In addition,a random perturbation error term is introduced to improve the robustness of the prediction model and finally achieve accurate prediction of hobbing gear quality.The application example shows that the method can effectively predict the gear quality with high prediction accuracy and generalization performance,which provides technical support for manufacturing shop decision analysis technology key parts machining process method and machining accuracy.In this paper,the application of fuzzy rough set attribute reduction and Gaussian mixture regression analysis in gear hobbing quality influencing factor analysis and quality prediction is deeply and systematically studied.They are respectively applied to feature parameter extraction and quality prediction of gear hobbing quality influencing factors,and combined with peak density clustering,multi threshold birch clustering,multiobjective differential evolution and other algorithms,a complete analysis and prediction model of hobbing gear quality was established.The research contents and methods of this paper have important theoretical significance and engineering application value for the quality analysis and prediction of gear processing industry.