Performance-based fault detection approach for the dew point process through a fuzzy multi-label support vector machine

摘要

Nowadays, there are many fault detection approaches, which are taken into consideration in two categories including the data-driven and the model-based fault detection approaches. One of the well-known data-based fault detection techniques is the support vector machine that is proved to be a powerful approach in the classification. This one has weakness in dealing with a variety of complicated data. To address this concern, based on the investigation presented, an integration of the two approaches including the fuzzy and the multi-label SVM is proposed. In a word, the performance-based approach can classify the noisy and the multi-label data. This one is carried out for the dew point process with the cooling cycle in the real-world plant data in correspondence with the simulated plant via the HYSYS software environment. The proposed performance-based approach is about 10% percent more accurate than the conventional multi-label SVM in identifying the faults of the process. (C) 2019 Elsevier Ltd. All rights reserved.