A diagnostic tool able to detect faults that may occur in a gas turbine power plant at an early stage of their emergence is of a great importance for power production. In the present paper, a diagnostic tool, based on Feed Forward Neural Networks (FFNN), has been proposed for gas turbine power plants with a condition monitoring approach. The main aim of the proposed diagnostic tool is to reliably detect not only every considered single fault, but also two or more faults that may occur contemporarily. Two different FFNNs compose the proposed diagnostic tool. The first network, that is not-fully connected, operates a fault pre-processing in order to evaluate the influence of the single fault variable on the single fault condition. The second FFNN detects the fault conditions by means of an iterative process. Such a diagnosis tool has been applied to a mathematical model of a single shaft gas turbine for power generation, resulting able to detect the 100% of single faults and the 80% of combined faults
Feed Forward Neural Network-based Diagnostic Tool for Gas Turbine Power Plant / Dambrosio, Lorenzo; Bomba, Marco; Camporeale, Sergio M.; Fortunato, Bernardo. - STAMPA. - (2002), pp. GT2002-30019.1-GT2002-30019.7. (Intervento presentato al convegno Power for Land, Sea, and Air: ASME Turbo Expo 2002 tenutosi a Amsterdam, The Netherlands nel June 3-6, 2002) [10.1115/GT2002-30019].
Feed Forward Neural Network-based Diagnostic Tool for Gas Turbine Power Plant
Lorenzo Dambrosio;Sergio M. Camporeale;Bernardo Fortunato
2002-01-01
Abstract
A diagnostic tool able to detect faults that may occur in a gas turbine power plant at an early stage of their emergence is of a great importance for power production. In the present paper, a diagnostic tool, based on Feed Forward Neural Networks (FFNN), has been proposed for gas turbine power plants with a condition monitoring approach. The main aim of the proposed diagnostic tool is to reliably detect not only every considered single fault, but also two or more faults that may occur contemporarily. Two different FFNNs compose the proposed diagnostic tool. The first network, that is not-fully connected, operates a fault pre-processing in order to evaluate the influence of the single fault variable on the single fault condition. The second FFNN detects the fault conditions by means of an iterative process. Such a diagnosis tool has been applied to a mathematical model of a single shaft gas turbine for power generation, resulting able to detect the 100% of single faults and the 80% of combined faultsI documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
