Machine Learning Health Assessment Model To Project Long-Term Degradation For A Pack Of Micro Gas Turbines

Abstract

Gas turbines are widely used for producing electricity, operating airplanes and for various industrial
applications such as in refineries and petrochemical plants. Due to the low thermal efficiency of the
fundamental gas turbine cycle, it is crucial to search for enhanced gas turbine-based cycles.
The availability and operating and maintenance expenses of tiny gas turbines can be considerably increased
and decreased using predictive health monitoring. Predictive health monitoring techniques are often created
for big gas turbines and frequently concentrate on a single setup. With this approach, the operational data from
the turbine can be analysed to predict when maintenance is necessary, allowing for more efficient and costeffective maintenance scheduling. This can also help prevent unexpected downtime and minimize the risk of
equipment failure.
Overall, using operational metrics from actual installations, this research proposes a potential data-driven
method for forecasting setup deterioration in micro gas turbines. The strategy employs linear regression
techniques to estimate and anticipate deterioration and is not dependent on data from a reference setup. The
method was evaluated on five distinct setups and showed consistent outcomes when compared to an existing
technique that makes use of data from an analogous setup.