Computational Fluid Dynamics (CFD) discretisation errors are generally present in any simulation result, with the main issues lying in the mesh generation stage. Despite this, CFDsoftware has widespread use in the industrial design procedure. To improve modelling reliability, mesh adaptation proposes to automatically modify grids to better their accuracy, while minimising their overall size. Regardless of the efforts made by the research community to produce a reliable technique, there are limited examples of its use in an industrial setting.

The aim of this paper is to prove the validity of an anisotropic sensor based on the functional volume-mesh sensitivity quantities. In particular, these will be employed to relocate the grid nodes towards the most problematic regions. It will be shown that this results in an improvement of the functional of interest and a reduction of its volume-grid sensitivities, yet maintaining a constant node count. Moreover, it will be demonstrated that nodes cluster anisotropically towards the problematic regions. The technique’s effectiveness will be proved by means of a comparison with a consistent feature-based mesh movement approach and tested on typical 3D compressor and fan cases. The promising results will show the method’s capability in tackling more advanced industrial configurations.

计算流体动力学（CFD）离散误差通常出现在任何模拟结果中，主要问题在于网格生成阶段。尽管如此，CFD软件仍在工业设计过程中得到广泛使用。为了提高建模的可靠性，网格自适应建议自动修改网格以提高其准确性，同时最大程度地减小它们的大小。不管研究团体为生产可靠技术所做的努力，在工业环境中使用该技术的例子都很少。

本文的目的是基于体积-网格敏感度函数来证明各向异性传感器的有效性。特别是，将使用这些方法将网格节点重新定位到最有问题的区域。可以看出，这导致了所关注的功能的改进，其体积网格敏感度的降低，但仍保持了恒定的节点数。而且，将证明节点各向异性地聚集到有问题的区域。将通过与基于特征的一致网格运动方法进行比较来证明该技术的有效性，并在典型的3D压缩机和风扇箱上进行测试。令人鼓舞的结果将显示该方法在解决更高级的工业配置方面的能力。