A Data-Driven Knowledge System for Anomaly Detection in the Oil & Gas Industry

Following the technological and digital developments introduced by Industry 4.0, the vast amount of information generated by an industrial plant increasingly requires more efficient and accurate management mechanisms for its real-time management. The proposed approach combines the Fuzzy Cognitive Maps (FCMs) methodology and the Gray Wolf Optimization (GWO) algorithm for the anomaly detection of an industrial plant with reference to the oil & gas sector. The power of FCMs mathematics and the flexibility and accuracy of the GWO algorithm allow real-time identification of the plant status and, if an anomaly is detected, discriminate potential causes. Moreover, although the FCM methodology is comparable to a neural network, it requires far fewer parameters to train the model, resulting in less computational time.