Enhancing Fault Detection in Gas Turbines Using MachineLearning Models: A Case Study on Kirkuk Gas Power Plant

This research deals with the problem of fault detection in gas turbineswhich are major sub-systems of modern power generation equipment.Therefore, the current research employs deep machine learning modelsknown as Temporal Convolutional Networks in combination with anAutoencoder aimed at improving anomaly detection accuracy andefficiency in gas turbines. To this end, this work uses vibration datagathered from accelerometers in the Kirkuk gas power plant andcompares the efficiency of the proposed TCN Autoencoder model toconventional models such as GRU Autoencoder, LSTM Autoencoder,and Variational Autoencoder (VAE). Statistically, findings highlight thatthe TCN Autoencoder model performs better in terms of identifyingsubtle anomalies than the RW model and has lower error measurementsthat thus point to enhanced PM and operational effectiveness. Theoutcomes from this research advance the ongoing endeavor of mergingmachine learning into machinery for precise and effective fault detection.

Enhancing Fault Detection in Gas Turbines Using MachineLearning Models: A Case Study on Kirkuk Gas Power Plant Keywords:

Gas turbine , Anomaly detection , Vibration analysis , Machine learning , TCN , Autoencoder