A robust quality-related fault detection method for nonlinear processes

Quality-related fault detection method is quite significant for the process industry in recent years. One of the most effective nonlinear quality-related fault detection methods is MKPLS. It can achieve the purpose of monitoring the process performance. However, it is sensitive to outliers. When a certain number of outliers occurs, MKPLS can not work well to build model precisely. To overcome this drawback, this paper proposes a revised version of MKPLS, which can be used in chemical plant for fault detection. The algorithm introduces a spherical strategy to ensure system robustness against outliers. In the fault detection part, the new gained regression coefficient are decomposed thoroughly into quality-related subspace and quality-unrelated subspace. Then, Hotelling's statistic is employed to monitor process. In the case study, a nonlinear numerical case and an industrial process simulator are applied to demonstrate the performance of the proposed method.