다품질 CFD와 M.L. 설계 알고리즘을 이용한 홴 임펠러의 공력 및 소음 최적설계

초록

In the research a multi-fidelity strategy for optimizing fan impellers in the design space accelerated by an M.L algorithm is developed and verified by comparing their performances and noise levels with those of off-the-shelf shrouded axial fan and turbo-type centrifugal fan. Our approach is to combine selected high-fidelity CFD simulations with a large number of lower-fidelity computationally inexpensive models to minimize the circulation strength and maximize the fan static efficiency near an operation point. The high-fidelity optimization model is implemented through a reinforced machine-learning A.I. algorithm.