Manufacturing process multivariate quality diagnosis classifier based on improved fuzzy support vector machine

Abstract

Provided is a manufacturing process multivariate quality diagnosis classifier based on an improved fuzzy support vector machine. Original data of quality characteristics in the manufacturing process is collected and preprocessed. A process analysis is made of the multivariate quality characteristics of key procedures using a hybrid algorithm. The stability and whether there is abnormality in the process are determined according to data recorded in a control chart. A process abnormality source is found using an improved fuzzy support vector machine method. In order to make the classification result more accurate, a membership factor and center point and boundary point correction are added to an objective function. The classifier has the advantages of rigorous process capability coefficient condition, accurate state judgment, low algorithm complexity and fast processing. Multivariate quality, misjudgment factor and principal component factor are integrated. The classifier is more adaptive. Parameter processing is standardized, and data processing is perfect. The probability of misjudgment is reduced. The problems of data bias and unit inconsistency are solved. Abnormality can be diagnosed.

Description

technical field [0001] The invention relates to the technical field of quality diagnosis in the manufacturing process of mechanical products, in particular to an improved fuzzy support vector machine manufacturing process multivariate quality diagnosis classifier. Background technique [0002] The modern manufacturing process is multivariate and highly correlated, and the process monitoring of this kind of production process is called multivariate quality control (MQC) or multivariate statistical process control (MSPC). The process of finding the cause of the loss of control is known as MSPC diagnosis or anomaly identification. There are two main types of methods: one is statistical decomposition techniques; the other is techniques based on machine learning. Mainstream decomposition techniques include principal component analysis (PCA), feature space comparison method, MTY method, step-down method, and multi-directional kernel principal component analysis method. However, ...

AI Technical Summary

Problems solved by technology

However, these methods usually involve complex statistical procedures, which are not conducive to the application

Images