Online Characterization of Combustion Dynamics Using System Identification Ba- sed on Sensor Data

Institut

Professur für Thermofluiddynamik

Typ

Masterarbeit /

Inhalt

theoretisch /  

Beschreibung

Combustion systems exhibit feedback between heat release fluctuations and chamber
acoustics, which generates system-specific characteristic sound emissions – a subject
commonly categorized under the umbrella of “combustion dynamics”. At a priori unknown
conditions, this feedback can become unstable leading to high-amplitude self-sustained
pressure oscillations and highly unsteady combustion. In extreme cases, the flame can go
extinct, or the combustion chamber can be damaged – situations that must absolutely be
avoided. Since the detailed stability margins in which these high-amplitude oscillations occur
are unknown and can shift during operation (fuel flexibility, machine aging, ambient
temperature, etc.), continuous monitoring based on dynamic pressure measurements is often
carried out. Usually, the resulting measurement data is inspected in the frequency domain,
which has the advantage that (sinusoidal) periodic oscillations collapse to single amplitudes at
characteristic frequencies. Although this provides more insights than looking at the raw time
series data, it is still challenging for operators to assess trends or stability margins. There-
fore, within this research project, new methods shall be explored that fit linear dynamical
systems to the measurement data using output-only identification methods. The eigen-
values of the resulting system then potentially allow for reliable and useful conclusions
on stability margins or combustor condition. This research project shall pave the way for
the development of new tools for early warning of combustion instabilities, machine tuning
and AI-based condition monitoring. The project is a collaboration between TUM and IFTA,
a company located in Puchheim near Munich that develops and sells machine monitoring
and protection systems in the energy sector.

Voraussetzungen

Python (or similar)

System Identification

State space modeling

Signal processing

Möglicher Beginn

sofort

Kontakt

Marcel Desor
Raum: 0732
Tel.: 089 289 16232
marcel.desortum.de

Ausschreibung