ABSTRACT Reduced order modelling (ROM) techniques allow to reduce the cost of shape optimisation problems. In the present work, the compressible turbulent flow around a gas turbine profile is studied by a Discontinuous Galerkin (DG) scheme. The simulations are accelerated by the combination of two existing ROM approaches: Domain Decomposition and Local Proper Orthogonal Decomposition in a DG (LPOD-DG) framework. In particular, the focus is fixed on the aerofoil suction side which is deformed while the pressure side remains fixed. The proposed method allows to reduce significantly the number of degrees of freedom in the simulation. The method is evaluated by performing a set of random predictions for shapes not included in the training database and comparing the obtained results with high-fidelity simulations. The approach is also compared to a p-adaptive scheme. Finally, the use of an automatic adaptive technique is investigated in order to improve the prediction accuracy at runtime.

中文翻译：

---

涡轮机械形状设计的降阶建模

摘要 降阶建模 (ROM) 技术可以降低形状优化问题的成本。在目前的工作中，围绕燃气轮机轮廓的可压缩湍流通过不连续伽辽金 (DG) 方案进行研究。通过结合两种现有的 ROM 方法来加速模拟：DG (LPOD-DG) 框架中的域分解和局部适当正交分解。特别是，焦点固定在翼型吸入侧，该侧发生变形而压力侧保持固定。所提出的方法允许显着减少模拟中的自由度数。该方法通过对训练数据库中未包含的形状执行一组随机预测并将获得的结果与高保真模拟进行比较来评估。该方法还与 p 自适应方案进行了比较。最后，研究了自动自适应技术的使用，以提高运行时的预测精度。