Research On Precision Prediction Method And Processing Parameter Optimization Of Gear Hobbing Workpiece

Gear is one of the most basic parts in contemporary manufacturing industry,which is widely used in automobile production,shipbuilding and even aerospace industry.Due to low level of intelligentialize and digitalization in most Chinese gear manufacturing industries,the adaptability of processing precision and processing parameters cannot be accurately measured and controlled,which seriously restricts the improvement of accuracy,stability and reliability of gear manufacturing in China.The traditional gear processing and manufacturing technology cannot meet the needs of intelligent transformation and upgrading of gear manufacturing enterprises.Aiming at the problems of coupling correlation of gear manufacturing precision inspection parameters,difficulty in catching key precision influencing factors,difficulty in estimating gear machining precision,complexity of manufacturing process and difficulty in optimizing process parameters,a series of key technologies for gear hobbing precision prediction and processing parameter optimization has been researched in this dissertation.The main research contents are as follows:(1)The correlation and importance of gear precision inspection indexes and manufacturing parameters are analyzed.Firstly,the variables of gear hobbing parameters is analyzed,and an adaptive machining precision preprocessing method suitable for gear precision inspection scene is designed.Then,based on the improved density peak clustering algorithm,the correlation degree of precision inspection indices is analyzed and the key indices are extracted.Then,a grading function for the precision of gear is proposed,which can realize the precision grading of gear machining workpiece.Finally,the rough set method is introduced to calculate the importance of each gear manufacturing parameter,and the unimportant manufacturing parameter items are omitted.(2)A gear hobbing precision prediction model before hobbing processing is constructed.In order to accurately predict gear machining precision,an adaptive variational inference Gaussian mixture regressor(AVIGMR)and a correlation analysis random forest regressor(CARF)are proposed.Among them,the adaptive parameter generator designed in AVIGMR can adaptively generate appropriate GMR super-parameter GMC number K_A and noiseδ_Aaccording to the characteristics of input data;and CARF has the designed adaptive correlation weight parameter based on cosine similarity.(3)The research on optimization method of gear hobbing processing parameters is carried out.Firstly,the motion trajectory of hobbing cutter is analyzed in detail,the specific time used in different processing stages is calculated,thus the machining efficiency model of helical gear is established.Taking machining energy consumption and hobbing tool wear as the main influencing factors of gear hobbing process cost model,the cost model is established by analyzing the power variation trend of each power component and the processing parameters influence on tool life.According to the characteristics of helical gear production data,the appropriate regression prediction model is selected as the prediction model of gear machining precision.Then,combined with multi-objective differential evolution algorithm and non-dominated sorting genetic algorithm,an adaptive multi-objective fusion evolution algorithm AMFEA is proposed to jointly optimize the established three-objective model of production efficiency,manufacturing cost and machining accuracy.(4)The gear machining precision prediction methods in hobbing process based on real time data and deep regression network are proposed.In order to realize the on-line monitoring of gear machining precision in the machining process,an integrated deep on-line prediction method EQOPF-GH is proposed to monitor and predict the gear machining precision.Among them,a multi-layer fusion residual network MFRes Net is designed,and a series of regularization methods are proposed to improve the generalization of prediction.A deep regression layer structure is designed to realize the mapping from the feature map to machining precision value.Aiming at the small sample data scenario,a transfer learning model for gear machining precision estimation based on domain adversarial and attention mechanism is proposed.In order to reduce the wear of precision inspection equipment and reduce the cost of data acquisition,an adaptive transfer model A~2Res Net-a Coral based on domain confrontation and attention mechanism is proposed for the training task of precision prediction of small sample gear machining vibration data set.The model integrates two attention structures:sample-spatial attention and dynamic channel attention,constructs a domain adversarial layer and the corresponding loss function to induce the feature extraction network to extract the homogeneous features of the source and target domains,and forcibly shortens the distance between the source and target domains.