Multivariate Process Quality Control And Diagnosis Method For Aviation Product Manufacturing

Aviation equipment products are characterized by numerous types,large number of parts and components,and precise manufacturing and assembly processes.In order to ensure that the aviation equipment achieves the expected design performance,it is necessary to strictly control numerous quality parameters in the manufacturing process.Therefore,the application of quality engineering in the manufacturing process is the key to improve the manufacturing capability of enterprises and ensure the expected performance of aviation equipment.Aiming at the manufacturing characteristics of aviation equipment products and by taking the multivariate process quality management technology in modern manufacturing as the research object,a general method and architecture of multivariate process quality control and diagnosis is proposed on the basis of analyzing the characteristics of manufacturing parameters,and the concerned theory of joint control of multivariate process quality components and correlation,diagnosis of multi-cause process quality component and components'correlation are detailed studied in this dissertation.The main contents are as follows.1.In order to simplify the control system of multivariate process quality in modern manufacturing,a solution based on transforming the joint control domain of quality components and their correlation is proposed.By using the principal component analysis method,the original process quality data which components are correlated is transformed into a new principal component vectors whose components are independent.The joint control domain of the transformed principal component vector is a hyper ellipsoid with each axis are parallel to the corresponding coordinate axis in high-dimensional space and is located within a hypercube formed by the intersection of the control limits of the principal components.Therefore,a T~2control chart can be used to monitor the shift of the quality components and their correlation of multivariate process quality,thus the control system can be simplified.2.As for the problem of multivariate process quality diagnosis,it is divided into two sub-problems:the diagnosis of quality components'shift abnormality and the diagnosis of correlation among components shift abnormality.For the problem of multi-cause process quality diagnosis,a method based on the complete binary tree structure of quality cause set is proposed.By successive bisection of the quality causes,the complete binary tree of the quality causes set is established,and then the sub-set of quality causes corresponding to each node in the binary tree is taken as the controlled quality causes to solve the regression equation of the process quality and the selected controlled quality causes,and then the corresponding cause-selecting control chart is built.When the process quality component shift abnormally,the range of abnormal quality causes is gradually narrowed according to the criteria for analyzing abnormal quality factors by traversing the binary tree of the cause-selecting control charts until the specific abnormal quality causes are diagnosed.The multi-cause process quality diagnosis algorithm based on the complete binary tree of the quality causes set can realize the rapid diagnosis of abnormal quality causes with the time complexity of the logarithmic function of the quality causes'number.3.In order to solve the problem of diagnosing abnormal correlation shift in multivariate processes,a diagnosis method based on quality components grouping is put forward.By analyzing the characteristics of the correlation of quality components,the overall correlation of quality components is decomposed into a group of binary correlations of all quality component pairs,and then the quality components are grouped by factor analysis method based on the correlation degree between different quality components,so that the correlation degree between quality components in the same group is as large as possible,while the correlation degree of quality components between different groups is as small as possible.Under the condition of ignoring the correlation of quality components between different groups,the quality components in the same group are combined as pairs,and the T~2control chart corresponding to them is established to form the multivariate quality correlation diagnosis system.The correlation diagnosis method based on quality components grouping can realize the diagnosis of multivariate process quality correlation abnormality with the spatial complexity of approximate linear function of quality components'number,and can avoid the defect of redundancy of diagnosis information in the existing methods.4.The multivariate process quality correlation diagnosis method based on quality components grouping ignores the correlation of quality components among different groups,so it has defects in theory.In order to solve this problem,based on the characteristic analysis of binary correlation,this dissertation proves that the binary correlation of a pair of quality components can essentially be expressed as a monotonic function of two quality components within their respective domains,and that the binary correlations of different component pairs has transitivity.Therefore,the binary correlations of all quality component pairs can be simplified as the binary correlations of component pair in a set of optimal basic correlation component pairs.Then,using the basic principle of graph theory to calculate the maximum correlation spanning tree of the binary correlation graph of multivariate process quality,the optimal set of basic correlation component pairs of multivariate process quality is solved,and T~2control charts for monitoring the shift of the binary correlation are established for each pair of quality components in the set of optimal basic correlation component pairs to form the multivariate process quality correlation diagnosis system.The correlation diagnosis algorithm based on the set of optimal basic correlation component pairs can realize the accurate diagnosis of multivariate process quality correlation anomalies with the spatial complexity of the linear function of the number of quality components.