Multi-Level Welding Quality Fault Discovery of an Intelligent Production Line by Using Taguchi Quality Loss Function and Signal-Noise Ratio

Abstract

The rapid and extensive pervasion of intelligent manufacturing concept has enhanced the revolution of the industry. A great interest has arisen in the past five years for the quantity and quality of an intelligent production line. Despite this fact, research areas, such as performance evaluation and fault discovery of these intelligent production lines based on different sets of criteria and techniques, are conspicuously untapped. This paper aims to contribute to the fault discovery by proposing an integrated approach combining the Taguchi quality loss function (QLF), the signal–noise ratio (SNR), and the relief method. First, in order to measure the theoretical value of quality deviation, the Taguchi QLF is introduced. By using the QLF, the information set is transformed into the quality features set. Thereby, the multiple quality features can be fused by using the SNR. Moreover, the features need to be reduced by the relief algorithm, if necessary. The Taguchi QLF-SNR allows decision makers to set tolerance thresholds for multi-levels (characteristic-level/unit-level/system-level) to discover the welding quality fault in the process of production line. Also a case study is presented to verify the feasibility and accuracy of the approach.