Scholarlink

<p>Creating ensembles for ensemble-based data assimilation (DA) in numerical simulations is often the greatest computational burden when applied to complex flow problems. In the present short communication, an ensemble-based DA method is presented, with the main novelty being the creation of the CFD ensemble by means of a simple scaling procedure and not additional CFD runs. This approach is in line with the common assumptions that the CFD uncertainty mainly originates from uncertain inlet conditions and that Reynolds number independence is observed for the examined flow. The developed method is applied to the case of a surface-mounted cube, immersed in the turbulent flow of an atmospheric boundary layer (ABL). Particle Image Velocimetry (PIV) measurements are assimilated into a RANS simulation. Both the mean velocities and the turbulence kinetic energy exhibit significant correction even when only 2% of the full PIV dataset is assimilated, while the computational cost remains comparable to that of a single CFD run.</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

Ensemble-Based Data Assimilation Without CFD-Generated Ensembles: A Scaling Approach for RANS Simulations Under ABL Inflow

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年5月16日

查看全文
  • Nikolaos Petros Pallas,
  • Demetri Bouris

摘要

Creating ensembles for ensemble-based data assimilation (DA) in numerical simulations is often the greatest computational burden when applied to complex flow problems. In the present short communication, an ensemble-based DA method is presented, with the main novelty being the creation of the CFD ensemble by means of a simple scaling procedure and not additional CFD runs. This approach is in line with the common assumptions that the CFD uncertainty mainly originates from uncertain inlet conditions and that Reynolds number independence is observed for the examined flow. The developed method is applied to the case of a surface-mounted cube, immersed in the turbulent flow of an atmospheric boundary layer (ABL). Particle Image Velocimetry (PIV) measurements are assimilated into a RANS simulation. Both the mean velocities and the turbulence kinetic energy exhibit significant correction even when only 2% of the full PIV dataset is assimilated, while the computational cost remains comparable to that of a single CFD run.