Anomaly Detection in Thermal Pulse Combustors Using Symbolic Time Series Analysis

This paper presents symbolic time series analysis of observable process variables for anomaly detection in thermal pulse combustors. The anomaly detection method has been tested on the time series data of pressure oscillations, generated from a non-linear dynamic model of a generic thermal pulse combustor. Results are presented to exemplify early detection of combustion instability due to reduction of friction coefficient in the tailpipe, which eventually leads to flame extinction.