A global classifier implementation for detecting anomalies by using one-class techniques over a laboratory plant

Abstract

The energy and the product optimization of the industrial processes has played a key role during last decades. In this field, the appearance of any kind of anomaly may represent an important issue. Then, anomaly detection in an industrial plant is specially relevant. In this work, the anomaly detection over level plant control is achieved, by using three one class intelligent techniques. Different global classifiers are trained and tested with real data from a laboratory plant, whose main aim is to control the tank liquid level. The results of each classifier are assessed and validated with real anomalies, leading to good results, in general terms.