Nonlinear Profile Monitoring Using B-spline And Modified Clustering Analysis

The application of statistical process control in the monitoring and diagnosis of product and process quality remains an important research issue in manufacturing.The development of sensing technology has created a data-rich environment for SPC.The digitalization and diversification of manufacturing process information is becoming clearer,and the application of traditional SPC in this field is limited.The challenge of quality control is gradually changing from the difficulty of data acquisition to how to analyze and extract the information and patterns of data.And because of the complexity of the continuous production process itself,the quality of the product or process may not be adequately represented by one variable or multiple variables,and it may be more appropriate to express it by the functional relationship between the response variable and one or more explanatory variables.This functional relationship is called Profile.The SPC process based on the profile for quality control of product or process is called Profile Monitoring.Profile monitoring is an important method in SPC and one of the hot spots in the field of quality control.The main purpose of Phase I analysis of profile monitoring is to identify and remove anomalous profiles to establish a stable and controlled process.In phase II,the control chart is established based on the controlled profiles obtained from phase I analysis,and the process is monitored in real time to find the abnormal operation state of the process.This thesis focuses on the Phase I process of profile monitoring.Outlier detection is the key problem of phase I monitoring.If the outliers cannot be identified and solved properly in phase I,it may lead to misleading results in Phase II.Therefore,it is very important to propose an effective outlier detection method.In the past,the research on profile monitoring mainly focused on the study of linear profiles.Due to the more complicated estimation process,the nonlinear profiles made the model more difficult,and the related research results were less.In view of this,for the problem of nonlinear profile outlier detection,this thesis proposes an outlier detection method based on Bspline fitting and modified clustering analysis algorithm.The performance of the new method under different variability conditions is analyzed in detail through simulation research.At the same time,the effectiveness and superiority of the new method are verified by comparison with existing methods.Finally,taking the classical vertical density profile(VDP)as an example,it shows that the proposed new method can be applied to the profile monitoring problem of complex shapes commonly encountered in the manufacturing process,and achieve higher accuracy in the outlier detection of complex profiles.