Ensemble one-class classification applied for anomaly detection in process control systems

In order to achieve better control performance for process control systems (PCSs) a plenty of advanced control methods have been proposed recently and most of them all need the support of process data generated during operation. While not much attention had been paid to the issue of anomaly detection for PCSs, which may be the main factor that prevents practical applications of these methods. This paper proposed a general anomaly detection method for online identifying anomalies mixed among process data of PCSs. Converting the issue of anomaly detection to one-class classification (OCC) can remit the absence of anomalies in PCSs. Exploiting the thought of ensemble can provide a more general solution for varying PCSs. Experimental results prove the effectiveness and applicableness of the proposed method and its ability to outperform state-of-the-art algorithms for detecting anomalies for PCSs.