A novel anomaly detection scheme for high dimensional systems using Kantorovich distance statistic

Abstract The partial least squares (PLS) is a commonly applied multi-variate method in anomaly detection problems. PLS strategy has been amalgamated with $$T^{2}$$ T 2 and squared prediction error (SPE) based statistical indicators to detect anomalies process. These traditional have few setbacks that made them ineffective monitoring applications. Hence, indicator on Kantorovich distance (KD) proposed for detecting sensor this study. integrates KD metric method. computes difference between the residuals of anomaly-free data uses as an anomaly. strategy’s critical feature single sufficient be integrated modeling framework. Tennessee Eastman process benchmark experimental distillation column processes are used assessing performance strategy. Further, comparisons provided KD, , SPE Generalized Likelihood Ratio indicators. results demonstrate superiority comparison framework also enhances small magnitude anomalies.