Session: 05-01 Advanced Controls for Propulsion Systems

Paper Number: 81310

81310 - Real Time Precursor Calculation for the Early Detection of Combustion Instabilities 

Combustion dynamics has long been recognized as one of the most impacting issues for the operation of gas turbines at low emissions. Over the past decades several control strategies have been defined by manufactures, pointing to increase the mean time between maintenance and, most importantly, avoid field failures.

Generally, control strategies can be classified as passive or active. The former requires the design of dedicated devices able to mitigate the instability. Helmholtz resonators or Quarter Wave Tubes installed along the combustion system are examples of this kind. The main drawback of such systems is that their damping effectiveness is limited around the design point and cannot ensure the mitigation of several other frequencies that can rise in the wide operational window they should cover.

On the other hand, active control strategies allow the combustor to dynamically adapt the operating conditions when the system is close to the instability. More complicated software logics and more sophisticated auxiliary systems are used to modulated fuel or air mass flow rates basing on the thermo-acoustic response of the combustor, resulting in a huge improvement of the performance.

Real-time measurements and their integration with physics-based models have been developed in the last years as the most attractive way to actively control thermo-acoustic instabilities. Neural control methods represent just the last step in this field of investigation. Similarly, low order modal decompositions are some of the most recent post-processing techniques for the analysis of both numerical and experimental results. Such kind of data management can provide not only the frequency spectra of a signal from a snapshot sequence but also be used to compute, among the other things, the growth rate of any relevant frequency the phenomenon is affected by. In the present work, the Dynamic Mode Decomposition algorithm has been adapted and optimized for the post processing of the real-time signals, acquired by the pressure pulsation probes monitoring the combustion dynamics.

The mathematical formulation was presented, followed by the application of the algorithm to the analysis of combustion dynamics recorded during the development of the NOVA LT16® annular combustor. The validation phase involved two different kind of instabilities. Firstly, thermo-acoustic tones have been considered: different unstable frequencies have been analyzed with the code showing promising results for the detection of the self-excited instability at the very first phase of the combustion dynamics.

Additionally, the algorithm was tested against a loss-of-flame event. It is well known that the lean blow out is generally anticipated by a low-frequency instability. For this kind of phenomenon, the code showed that it is possible to detect some precursor events, even some seconds before the flame out.

Presenting Author: Roberto Meloni Baker Hughes

Presenting Author Biography: Dr Roberto Meloni (m), Senior combustion Engineer with almost 10 years of experience in numerical modelling of the combustion process for internal combustion engines and gas turbines. He holds a PhD in combustion, achieved at the University of Rome “Sapienza”. He has joined in 2014 the combustion department at Baker Hughes-Nuovo Pignone (former GE Oil&Gas) where he is responsible for the definition of numerical processes able to optimize the design of combustion systems. He has been involved in different national and EU-funded projects and author of several research papers published in peer-reviewed journal and conferences.

Authors:

Roberto Meloni Baker Hughes
Nicola Giannini Baker Hughes