ML based Prediction of the Decay Coefficient of a Naval Propulsion System’s Turbine

The failure of the gas turbine of the naval vessel propulsion plant can have disastrous consequences, and hence timely maintenance becomes key to prevent any such failure and associated losses. To be able to predict run time decay of turbines, well in advance and taking precautionary steps immediately, can avoid such disruptions. This study focuses on employing various ML and DL techniques that use data from different components of the propulsion system to predict their decay, enabling timely intervention and maintenance of the faulty components. As part of this research multiple techniques were assessed to find out the most optimal model that could give an insight into the decay state coefficient of the turbine with the highest precision. Feature extraction techniques like Principal Component Analysis (PCA) were also applied but were not found efficient enough. Out of all the algorithms applied, Optimized Support Vector Machine(SVM) emerged as the most powerful and efficient model capturing 99.22% of the variance in the data thereby proving the efficacy and robustness of optimized models.