Key Performance Indicator Related Fault Detection Based on Modified KRR Algorithm

In the handling of the nonlinear systems, Kernel Ridge Regression (KRR) has recently served as an effective method to deal with multicollinearity problem, but it has not been used to solve the problem of fault diagnosis problems. KRR is unable to ensure the safety and the reliability of industrial systems. While previous fault detection method still encounters some problems for key performance indicator (KPI) related fault diagnosis of the underlying process. In order to compensate for these drawbacks, this paper proposes a new KPI-related fault detection algorithm, named Modified KRR (MKRR). MKRR can decompose accurately the measurable process variables into the KPI-related and KPI-unrelated parts and use corresponding test statistics to monitor them. The method can offer good performance in industrial systems. To prove the effectiveness of the proposed method, a nonlinear numerical example is used. The simulation results show that the proposed method performs better than traditional KPLS in KPI-related fault detection.