Assembly Fixture Fault Diagnosis Using Designated Component Analysis

Abstract

A new approach to fixture fault diagnosis, designated component analysis (DCA), is proposed for automotive body assembly systems using multivariate statistical analysis. Instead of estimating the fault patterns solely from the process data as in principal component analysis (PCA), DCA defines a set of mutually orthogonal vectors identified from known product/process knowledge to represent fault patterns, estimates their significance from data, and analyzes the correlation among the designated components. Hence, the sheet metal dimensional variation is mathematically decomposed into a series of mutually orthogonal rigid body motions with known patterns. Remaining deflections can be estimated by PCA after rigid body motions have been removed from the data. As a result, the designated components, along with their correlations, facilitate the diagnosis of multiple fixture faults that exist simultaneously and isolate deflections from other variation components. An application example is used to illustrate DCA’s effectiveness and potentials.