A Methodology Based on Evolving Systems for Fault Detection and Identification of Dynamic Systems
Fault detection and identification, AutoCloud, online, data stream.
This work proposes a methodology for the detection and identification of failures in dynamic systems, through an online and evolutionary approach. The proposal is divided into three stages, in which data pre-processing and post-processing are carried out to increase the robustness of the methodology in the presence of outliers and noise, in the pre-processing the selection of characteristics, normalization is carried out of data, filtering and adding regressors, in post-processing time filtering is performed. In the processing stage, an adaptive and unsupervised approach is applied, through the Auto-Cloud algorithm, which performs grouping and classification of data streams. To validate this proposal, different evaluation metrics were used, such as Adjusted Rand Index (ARI), homogeneity, completeness, precision, f1_score, recall, and satisfactory results were obtained. Finally, the conclusion of this work is presented, in addition to proposals for future work.