• Plataforma de vídeos del IIT
  • Twitter
  • LinkedIn
  • Asociación de Ingenieros del ICAI
  • Intranet
  • Spanish
Go top
Paper information

DADICC: Intelligent system for anomaly detection in a combined cycle gas turbine plant

A.L. Arranz, A. Cruz, M.A. Sanz-Bobi, P.R. Castelló, J. Coutiño

Expert Systems with Applications Vol. 34, nº. 4, pp. 2267 - 2277

Summary:
DADICC is the abbreviated name for an intelligent system able to detect on-line and diagnose anomalies as soon as possible in the dynamic evolution of the behaviour of a power plant based on a combined cycle gas turbine. In order to reach this objective, a modelling process is required for the characterization of the normal performance when any symptom of a possible fault is present. This will be the reference for early detection of possible anomalies. If a deviation in respect to the normal behaviour predicted is observed, an analysis of its causes is performed in order to diagnose the potential problem, and, if possible, its prevention. A multi-agent system supports the different roles required in DADICC. The detection of anomalies is based on agents that use models elaborated using mainly neural networks techniques. The diagnosis of the anomalies is prepared by agents based on an expert-system structure. This paper describes the main characteristics of DADICC and its operation.


Keywords: Anomaly detection; Normal behaviour; Diagnosis; Multi-agent system; Neural network; Expert system


JCR Impact Factor and WoS quartile: 2.596 (2008); 6.954 - Q1 (2020)

DOI reference: DOI icon 10.1016/j.eswa.2007.03.005

Published on paper: May 2008.



Citation:
A.L. Arranz, A. Cruz, M.A. Sanz-Bobi, P.R. Castelló, J. Coutiño. DADICC: Intelligent system for anomaly detection in a combined cycle gas turbine plant. Expert Systems with Applications. Vol. 34, nº. 4, pp. 2267 - 2277, May 2008.


    Research topics:
  • *Forecasting and data mining
  • *Modeling, simulation and optimization
  • *Artificial intelligence applied to maintenance diagnosis and reliability

pdf Preview
Request Request the document to be emailed to you.