校准
湍流
计算机科学
数学优化
气象学
物理
数学
统计
作者
Alperen Yildizeli,Sertaç Çadırcı
标识
DOI:10.1115/imece2023-115019
摘要
Abstract Flow over a Backward Facing Step (BFS) is a classic fluid dynamics problem that has received considerable attention in the research community. In this study, the k-ω SST turbulence model is calibrated using ANSYS Fluent flow solver based on the comparison with Direct Numerical Simulation (DNS) data of a flow over a BFS which is available in the literature. On at a time (OAT) sensitivity analysis is conducted to determine six most dominant turbulence closure coefficients. Skin friction coefficient distribution is examined as quantity of interest. After determining the six most effective parameters, a deep neural network is trained with 500 CFD simulations; and multi-objective genetic algorithm is applied to reduce both RMSE and maximum absolute error of the skin friction coefficient distribution. The results of the study demonstrate that the model coefficients were successfully calibrated using multi-objective optimization. Improvements of the velocity profiles and skin friction coefficient are addressed to the optimization of closure coefficients which are related to the diffusion and production terms of turbulence model of interest. Optimum values of α∞*, β∞*, βi,1, βi,2, a1, and σω,2 were found 13.5% higher, 25% lower, 6.5 % lower, 6.5% higher, 14.5% higher and 3% higher than their original values, respectively.
科研通智能强力驱动
Strongly Powered by AbleSci AI