Neural network based vibration analysis with novelty in data detection for a large steam turbine

Article type: Research Article

Authors: Kumar, K. P. | Rao, K.V.N.S. | Krishna, K.R. | Theja, B.

Affiliations: School of Mechanical Engineering, Vignan University, Vadlamudi, India | Machine Dynamics and Failure Analysis Laboratory, Bharath Heavy Electricals Limited (B.H.E.L), Hyderabad, India | Department of Mechanical Engineering, Sir C R Reddy College of Engineering, Eluru, India

Note: [] Corresponding author. Tel.: +91 9490223849, +91 863 2244735; Fax: +91 8812 224193; E-mail: [email protected]

Abstract: Health of rotating machines like turbines, generators, pumps and fans etc., is crucial to reliability in power generation. For real time, integrated health monitoring of steam turbine, novel fault detection data is essential to reduce operating and maintenance costs while optimizing the life of the critical engine components. This paper describes about normal and abnormal vibration data detection procedure for a large steam turbine (210 MW) using artificial neural networks (ANN). Self-organization map is trained with the normal data obtained from a thermal power station, and simulated with abnormal condition data from a test rig developed at laboratory. The optimum size of self-organization map is determined using quantization and topographic errors, which has a strong influence on the quality of the clustering. The Mat lab 7 codes are applied to generate program using neural networks toolbox.

Keywords: Integrated health monitoring, self-organization map, quantization and topographic errors, clustering

DOI: 10.3233/SAV-2012-0614

Journal: Shock and Vibration, vol. 19, no. 1, pp. 25-35, 2012