Diagnosis of Combustion Instability in Gas Turbine Combustor Using Auto-Encoder-Based Machine Learning

Various diagnostic models for combustion instability based on statistical techniques and machine learning using measured or monitored data are being developed to suppress instabilities in power generation gas turbines. In this paper, we proposed a combustion state diagnosis model using an auto-encoder, one of machine learning approaches. This model has been trained using two main input components; magnitude and waveform pattern of time series dynamic pressure which is measured from a real-scale single nozzle gas turbine combustor. As a result of training, the model could successfully classify the pre-processed input data into stable, transition, and unstable states onto the feature map of the latent space. Especially, in contrast with existing combustion instability monitoring tools based on statistical techniques, the current auto-encoder model provided accurate combustion state diagnosis performance for stable and transition states of dynamic pressure.