Output details
13 - Electrical and Electronic Engineering, Metallurgy and Materials
University of Strathclyde
An agent-based implementation of hidden Markov models for gas turbine condition monitoring
This paper presents an approach to dynamic monitoring of gas turbines using relatively low-frequency data. Traditional monitoring requires 10-minute averaged data for steady-state monitoring, and high-frequency vibration data for dynamic monitoring. The paper's industrial co-author [Scottish and Southern Energy (SSE), john.twiddle@sse.com] contributes a case study of data from in-service turbines, showing a successful side-by-side comparison of the new technique against SSEs current monitoring system. This work led directly to follow-on funding [SSE Research Partnership, £64.7k, 2010-2011] for researcher time to develop an online prototype system for dynamic gas turbine monitoring.