HIBRID METHODOLOGIES OF SOFT COMPUTING AND ITS APLICATION TO FAULT DETECTION IN WASTEWATER TREATMENT PLANTS

DURATION: From 1st of December 2002 to 30th November 2005

CHIEF RESEARCHER: Ángela Nebot Castells

KEYWORDS:

Large-scale dynamical systems; time series: analysis and forecasting; fuzzy inductive reasoning; heterogeneous neural networks; delta neural networks; fault monitoring systems; water distribution network.

SUMMARY

Starting from the work done in the previous project (TAP1999-0474), the objective of this project is the deepening into the development of hybrid methodologies of soft computing for fault detection and identification in complex dynamical systems. In the previous project some methodologies based on fuzzy and connectionist techniques were refined (fuzzy inductive reasoning, feed-forward, recurrent, delta and heterogeneous neural networks), and a comparative study was performed in order to identify the strong points of every methodology. In this project the research of possible hybridizations of these techniques is proposed, together with their combination by means of cooperation methods, in order to predict the future behaviour of real systems and detect possible faults in their work. The results will be validated on standard benchmarks, laboratory models, and finally will be applied to a wastewater treatment plant.

PARTICIPANTS:

Researchers from the following institutions:

• Departamento de Lenguajes y Sistemas Informáticos (UPC)
• Departamento de Ingeniería de Sistemas, Automática e Informática Industrial (UPC)
• Departamento de Estadística e Investigación Operativa (UPC)
• Instituto de Robótica e Informática Industrial (UPC)
• Departament d'Enginyeria Química, Agrària, i Tecnologia Agroalimentària (Universitat de Girona)