| Summary: | In the production of gas polyethylene (PE) pipelines, quality characteristics such as ovality, outer diameter, and wall thickness often exhibit unknown distributions and complex inter-variable correlations. Traditional parametric control charts are prone to false alarms and missed detections under such conditions. This study proposes a Multivariate Lepage-type Projection Nonparametric (MLPN) control chart, integrating Vander Waerden and Klotz tests with a robust covariance estimation framework. The method first transforms multivariate observations into a single test statistic to jointly monitor shifts in location and scale parameters via nonparametric rank-based statistics. It then introduces a projection-based weighting scheme to mitigate the influence of outliers on covariance estimation, thereby enhancing the robustness and applicability of the control chart under complex conditions. Monte Carlo simulations and real-world PE pipe production data demonstrate that the proposed chart achieves accurate detection of process shifts, with reduced false alarm rates and improved sensitivity, providing an efficient and reliable tool for quality monitoring in continuous manufacturing processes.
|
|---|
