Statistical Diagnosis for Quality-Related Faults in BIW Assembly Process

Abstract

The assembly process of a car body-in-white (BIW) in a body shop is a very important link in complete vehicle manufacturing process, and BIW dimension quality control is the most important part of BIW quality control. The quality-related fixture fault diagnosis problem is proposed in smart dimension control loop [established in BMW Brilliance Automotive company (BBA)] aiming to detect locating fixture fault, which could seriously influence BIW dimension quality. This article introduces a novel statistical quality-related fault diagnosis method by combining Kalman filter and generalized likelihood ratio test, making system innovation as a basic fault diagnosis tool and establishing a hypothesis test between fixture-fault-free model and fixture-faulty model to detect potential fixture fault, meanwhile estimates fixture fault occurred station by maximum likelihood estimate. The fixture fault diagnosis flowchart is built to decrease false and missing alarm rate in actual production process. The case study based on the real BIW component assembly process data from BBA demonstrate that this fault diagnosis method can accurately post fixture fault warning and send correct order to on-site workers to maintain locating fixture in a batch of products.