Quality Fluctuation Abnormal Source Identification Based on Variable Weighted Reconstruction Analysis in Spinning Process

Abstract

Due to the numerous and interconnected quality influencing parameters, the production process propagation and evolution of abnormal factors are complex, which can affect the stability of quality characteristics from multiple perspectives. This paper addresses the problem of identifying the quality fluctuations sources and proposes a variable-weighted reconstruction analysis-based method for identifying abnormal sources of quality fluctuations in the spinning process. The method monitors the degree of quality fluctuations by constructing information entropy statistics and reconstructs the weighted parameters of abnormal quality processes. On this basis, abnormal contribution ranking is performed based on the degree of change in quality characteristics before and after weight reconstruction, which achieves the identification of the abnormal source. The proposed method is validated using a spinning process dataset. It reveals that the method could accurately identify the quality fluctuation abnormal source, which indicates its practicability and feasibility and will provide a theoretical basis for the quality stability control.