Research On Assembly Misalignment Recognition Method Of Spiral Bevel Gear Based On Contact Pattern Detection

Spiral bevel gear has the characteristics of large contact degree,smooth transmission,low noise and high bearing capacity.It is the key parts of automobile,aviation,aerospace,machine tool and construction machinery.The meshing quality of spiral bevel gear is mainly affected by gear manufacturing and assembly.When the geometric shape of the tooth surface after machining is in good agreement with the theoretical design,the tooth surface error can be ignored.However,in the actual assembly process,the assembly misalignment will be generated because the gear pair installation position deviates from the theoretical design position,which will affect the tooth surface contact characteristics.The gear meshing state is directly reflected as the contact pattern of the tooth surface,which is one of the important symbols to measure the quality of gear meshing.Therefore,it is of important scientific research value and engineering significance to identify the tooth surface contact pattern when the gear pair produces assembly misalignment due to assembly reasons,and then feedback it to the corresponding assembly adjustment.The traditional process of adjusting the installation position of spiral bevel gear pair requires workers to have rich adjustment experience,and the consistency and accuracy of the adjustment results cannot be guaranteed.Because the tooth surface contact marks have different physical characteristics under different assembly misalignment,the assembly misalignment recognition of spiral bevel gears can be converted into contact pattern image classification recognition.With the application of deep learning technology in engineering practice,more and more practical problems have been solved,and it is possible to use deep learning technology to recognize the assembly misalignment of spiral bevel gears detected by contact marks.Therefore,the main research work of this paper includes:(1)Construct the contact pattern data set of spiral bevel gears.Due to the lack of public standard data set for classification and identification of assembly misalignment,data set should be constructed for deep learning training neural network.Firstly,based on the principle of image acquisition,the original image of contact pattern is acquired under different assembly misalignment.Because the tooth surface of spiral bevel gear is a curved surface,an image preprocessing algorithm based on tooth surface flattening and tooth contour extraction was designed to extract more effective features.Data samples meeting the requirements were obtained through the algorithm processing,and data sets were made.At the same time,the author analyzes the problems such as data imbalance,high similarity and easy to be confused in the self-made data set,and carries out in-depth research on this basis.(2)The assembly misalignment recognition algorithm for spiral bevel gears based on the improved Efficient Net V2 is proposed.In this paper,the self-made contact pattern image data set is used as the research objects,the classification network is selected as Efficient Net V2-S,and the computation load is reduced and the computation speed is improved by simplifying the network model.By modifying the Attention mechanism,introduce Coordinate Attention(CA)attention mechanism,comprehensively consider the relationship between tooth surface impression space and position information,solve the similarity easily confused problem in the process of classification;By modifying the loss function,the Class-balanced loss(CB-Loss)function was introduced,and different weights were assigned to different categories to solve the problem of data imbalance.Furthermore,the overall recognition ability of the network model to the self-made contact pattern image data set is improved.The experimental results show that the classification accuracy of the improved algorithm on the constructed contact pattern image data set reaches 97.60%,which proves the effectiveness of the proposed algorithm.(3)Design the assembly misalignment identification system of spiral bevel gear.This paper encapsulates the Python-based image processing algorithm and image classification algorithm in the QT graphical user interface based on C++ and SQLite database,and designs the assembly misalignment recognition system of spiral bevel gear based on contact pattern detection.The system can classify the type and amount of assembly misalignment of the contact pattern image collected by the user in real time,and then feed back the result to the assembly of parts.Through the adjustment and assembly of the spiral bevel gear meshing test bench and the test of the assembly misalignment identification system,the effectiveness of the proposed algorithm and the feasibility of the assembly misalignment identification system are verified,and the classification of assembly misalignment is accurate and rapid.