기계 학습을 이용한 ORC용 터빈 형상 최적화

In this study, our primary objective is to optimize a turbine employed in an Organic Rankine Cycle (ORC) system that harnesses waste heat from a ship. Traditional optimization methods are known to be time-consuming and expensive. To streamline a more practical optimization approach, we have leveraged design of experiment, machine learning, and Latin hypercube sampling. Our experimental design involves varying turbine geometries under different conditions and subsequently creating a machine learning model using the results derived from computational fluid dynamics simulations. Through the utilization of these methodologies, we have succeeded in achieving a turbine that surpasses the turbine
isentropic efficiency obtained at the initial design point by 1.3% point.