Diagnóstico de falhas baseado em sistema inteligente evolutivo

Abstract

In many areas, the advance of technology has resulted in the emergence of machinery and complex equipments, which impose great challenges for its management and maintenance, and consequently, increase the fault occurrences. In this context, this work proposes a methodology for fault diagnosis based on evolving intelligent system aiming at its application in nonstationary dynamic systems which require performing these tasks in online mode and real time. Evolving intelligent systems are dened as systems capable to gradually determine both its structure and its parameters, by extracting knowledge from input data. According to intelligent maintenance concept, the application of evolving intelligent systems to perform fault diagnosis has shown promising. In this work we propose an evolving fuzzy classier for fault diagnosis application, based on a new approach that combines a recursive clustering algorithm and a drift detection method. This approach gives the evolving fuzzy classier the ability of continuous and incremental learning in online mode and real time, without need a prior knowledge about of the dynamic system in question, and providing greater robustness to outliers and noise present in input data. The classier proposed in this work is evaluated on two problems typically found in the industry, which are fault diagnosis in DC drive system, and fault diagnosis in interactive tanks system. The results of simulations and experiments demonstrate that the proposed classier achieves promising performance, suggesting it as a feasible alternative for application to real problems of fault diagnosis.