Gas Turbine Design Analysis And Optimization With Novel Hybrid Model Using Classical Physics And Machine Learning

In this research Gas Turbine analytical design and optimization using physics-based models (PBM) vs machine learning-based models (MLBM) are compared for performance evaluations. The paper contributes towards the design modeling, parametric analysis, and optimization of the gas turbine operating at various altitudes. Physics based nonlinear second-order model and data-driven analytical approach using machine learning were compared with each other for designing a gas turbine. The above PBM model uses gray-box identification procedures for predicting parameters and the MLBM uses black-box procedures. A novel hybrid model utilizing Machine Learning along with physics-based modeling of gas turbines was developed. The physics-based models usually mean the first-principle models or the classical models based on empirical equations governed by the laws of physics and thermodynamics. This requires the processing of domain knowledge to identify the dependencies and their relationship between the various thermodynamic parameters and the target function whereas, the machine learning models understand the patterns between input data and results by gathering missing information and hidden data to identify the behavior of the gas turbine with changes to the thermodynamic parameters by itself through correlations. The proposed hybrid model helps in overcoming the disadvantages of PBM and MLBM. The results of the hybrid model when physics and machine learning model are combined provide a more accurate prediction of the gas turbine performance.