It is well known that the efficiency of a steam turbine is affected by the pressure recovery performance of its low-pressure exhaust hood,and therefore,parametric analysis of the exhaust hood is of great importance in...It is well known that the efficiency of a steam turbine is affected by the pressure recovery performance of its low-pressure exhaust hood,and therefore,parametric analysis of the exhaust hood is of great importance in the steam turbine design process.In recent years,computationally inexpensive metamodels have been widely used in the parametric analysis of exhaust hood.However,the prediction accuracy of metamodels is highly dependent on the number and distribution of sample points in the design space.The purpose of active learning is selecting informative samples sequentially to obtain an accurate metamodel within a reasonable computational budget.However,the impact of active learning on the accuracy of metamodels such as exhaust hood parameter analysis has not been fully explored.Therefore,this paper investigates and compares four representative active learning methods on the parametric modeling of turbine exhaust hoods,and the comparison results highlight the advantages of active learning and the analysis of the exhaust hood based on the metamodel with the highest accuracy.展开更多
基金National Natural Science Foundation of China(52005074)Natural Science Foundation of Liaoning Province(2022-MS-135)。
文摘It is well known that the efficiency of a steam turbine is affected by the pressure recovery performance of its low-pressure exhaust hood,and therefore,parametric analysis of the exhaust hood is of great importance in the steam turbine design process.In recent years,computationally inexpensive metamodels have been widely used in the parametric analysis of exhaust hood.However,the prediction accuracy of metamodels is highly dependent on the number and distribution of sample points in the design space.The purpose of active learning is selecting informative samples sequentially to obtain an accurate metamodel within a reasonable computational budget.However,the impact of active learning on the accuracy of metamodels such as exhaust hood parameter analysis has not been fully explored.Therefore,this paper investigates and compares four representative active learning methods on the parametric modeling of turbine exhaust hoods,and the comparison results highlight the advantages of active learning and the analysis of the exhaust hood based on the metamodel with the highest accuracy.
