Scholarlink

<p>This study aims to employ a generative adversarial network (GAN) for the creation of flow fields in turbulent round jets, with a subsequent assessment of their performance. A ground-truth dataset to train a GAN model is obtained using a three-dimensional large eddy simulation. The reliability of the large eddy simulation dataset is validated against experimental data from laboratory experiments. The performance of GAN is assessed using temporally and azimuthally averaged flow quantities, e.g., temporal-azimuthal averaged velocity and turbulent kinetic energy. The normalized root mean square errors between the flow fields generated by GAN and the ground truth flow fields are 0.46% for the averaged velocity and 3.08% for the turbulent kinetic energy. Utilizing GAN, the time required to generate flow fields, encompassing both GAN training and simulation, is reduced by an impressive 99.7% compared to the duration associated with LES simulations.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

On the performance of generative adversarial network for generating turbulent round jet flow

  • 期刊论文

  • 发表日期：

    2025年2月7日

查看全文
  • Seongeun Choi,
  • Yeulwoo Kim,
  • Jin Hwan Hwang

摘要

This study aims to employ a generative adversarial network (GAN) for the creation of flow fields in turbulent round jets, with a subsequent assessment of their performance. A ground-truth dataset to train a GAN model is obtained using a three-dimensional large eddy simulation. The reliability of the large eddy simulation dataset is validated against experimental data from laboratory experiments. The performance of GAN is assessed using temporally and azimuthally averaged flow quantities, e.g., temporal-azimuthal averaged velocity and turbulent kinetic energy. The normalized root mean square errors between the flow fields generated by GAN and the ground truth flow fields are 0.46% for the averaged velocity and 3.08% for the turbulent kinetic energy. Utilizing GAN, the time required to generate flow fields, encompassing both GAN training and simulation, is reduced by an impressive 99.7% compared to the duration associated with LES simulations.