Data-driven fault detection and estimation in thermal pulse combustors

This paper presents the development of a dynamic data-driven statistical method for: (a) early detection of incipient faults and (b) parameter estimation for prognosis of forthcoming failures and operational disruptions (e.g. flame extinction) in thermal pulse combustors. From these perspectives, reduction in the tailpipe friction coefficient is estimated from time-series data of pressure oscillations. The algorithms for parameter estimation are built upon the principles of Symbolic Dynamics, Information Theory and Statistical Pattern Recognition . The proposed algorithms have been tested on an experimentally validated simulation model of a generic thermal pulse combustor.