Session: 05-04: Thermal Measurements

Paper Number: 153560

Optimal Sensor Placement for LDA Measurement and Health Monitoring in a Recursive H2 Combustion 

The need for better combustion monitoring in gas turbines has become more acute with the latest technical requirements, standards, and policies in terms of efficiency, environment, safety, operation flexibility and operation costs. Combustion Bay One and FH JOANNEUM started the project “KI-PIRINHA” (Key enabling technologIes on Performance, EfficIency and Resilience by Artificial INtelligence for Hydrogen Applications) in 2024, which is an experimental research program about health monitoring of H2 combustion. The focus is put on protection versus detonation at ignition.

For the implementation of a health monitoring system, a reduced order model is created based on CFD simulations. To optimize efficient information gain of validation experiments with LDA measurements, methods of sparse sensing are utilized on the flow-field inside conceptual combustion chambers. We present the mathematical foundations and the application on the specific flow field within the combustion test stands of the project KI-PIRINHA.

Optimizing the most informative sensor location is based on a dimensionality reduction of the flow-field on to a tailored basis (Eigenmodes) and subsequent QR decomposition of these modes to find optimal sensor locations. Without the need to fully reconstruct overall systems states, we apply minimization of the ℓ₁-norm to find the fewest nonzero entries of the full measurement location index vector. These locations support either state reconstruction or discriminating a health status of the flow in the combustor and optimally inform the decision task for flow regime classification and identification of a precursor for detonation. We showcase the application of the method on two testcases: Firstly, a combustion chamber for H2 combustion is considered and, secondly, transient behavior of the flame front in a deflagration to detonation transition test stand is investigated.

The classification performance achieved leverages applications for health monitoring and emergency shut-down decisions as well as reignition procedures. This knowledge enables further research on sparse placement of future implementation of optic-acoustic and/or pressure probes

The paper finishes with a specific measurement plan as the design-of-experiments result. This strategy optimizes the information gain for model validation, minimizes experiment duration and cost in the development phase and leverages health monitoring and anomaly detection capabilities in operation. The exhibition of results for the specific use cases prove the suitability of our approach and the detailed presentation of used methods allows the research community for adaption to numerous applications. The sparse sensing leverages possibilities of data driven model creation and probabilistic health monitoring.

Presenting Author: Arno Fallast FH JOANNEUM GmbH, University of Applied Sciences , Institute of Aviation

Presenting Author Biography: A. Fallast is a researcher and senior lecturer at the Institute of Aviation at the University of Applied Sciences, Graz in the working group “Flight control and artificial intelligence.” He is a PhD Student at the Technical University Graz (Doctoral School of Computer Science) researching on AI methods in flight control.
He received his bachelor’s degree and his master’s degree in aviation engineering from the FH JOANNEUM. For 10 years he is working as a researcher and lecturer and his research interests include:
- System identification of unmanned vehicles and dynamical systems.
- Dynamic system anomaly detection
- AI based flight control.

Authors:

Arno Fallast FH JOANNEUM GmbH, University of Applied Sciences , Institute of Aviation
Lukas Andracher FH JOANNEUM University of Applied Sciences, Institute of Aviation
Fabrice Giuliani Combustion Bay One e.U.
Nina Paulitsch Combustion Bay One e.U.
Andrea Hofer Combustion Bay One e.U.