Gas turbines play a major role in the field of power generation. The heart of a gas turbine engine is the combustion chamber or combustor where the combustion process takes place to produce energy. One of the key challenges in this process is the onset of a thermo-acoustic instability, which may result in reduced combustion efficiency and even structural damage. Over the years, different methods have been proposed by researchers to detect and predict combustion instability with most of them deriving a precursor which senses features that can characterize combustion instability. In this regard, a data driven method based on cross-entropy derived from symbolic time series analysis of the chemiluminescence and pressure sensor data has been recently found to be robust and fast among other existing methods for predicting instability. The excellent performance of this prediction model can be attributed to the approach utilising STSA tools to construct the probabilistic state machines called D-Markov machines to derive the cross-entropy precursor. Conventional sensors used in laboratory-scale combustors such as high-speed imaging camera for chemiluminescence and piezo-electric transducers (PZT) for pressure have been shown to provide satisfactory performance for predicting the onset of combustion instability through the above-said data-driven algorithm. However, it is difficult to incorporate a camera in a practical combustor since it needs an optical window. Also, a pressure sensor which is not susceptible to EMI is preferred in the combustion environment.

In this thesis, we have experimentally demonstrated a complete fiber optic sensor solution for the above requirement with fiber optic bundle-based chemiluminescence sensor and fiber Bragg grating (FBG) based pressure sensor. The solution utilizes the advantages of fiber optic sensors such as compactness, immunity to EMI, minimal intrusive nature etc. A key challenge while using fiber optic sensors especially the FBG pressure sensor in the combustor is the high-temperature environment. We have carried out experimental analysis to adapt and package the sensors for such a harsh environment. Proof of principle experiments are conducted to evaluate the performance of sensors and establish them as an effective replacement of the standard sensors. Through our work, we have demonstrated that the fiber optic solutions provide a viable pathway for implementing a robust and field-deployable solution to predict the onset of combustion instability in commercial gas turbines.