In this paper, we investigate the feasibility of using DNS data and machine learning algorithms to assist RANS turbulence model development. High-fidelity DNS data are generated with the incompressible Navier–Stokes solver implemented in the spectral/hp element software framework Nektar++. Two test cases are considered: a turbulent channel flow and a stationary serpentine passage, representative of internal turbo-machinery cooling flow. The Python framework TensorFlow is chosen to train neural networks in order to address the known limitations of the Boussinesq approximation and a clustering based on flow features is run upfront to enable training on selected areas. The resulting models are implemented in the Rolls-Royce solver HYDRA and a posteriori predictions of velocity field and wall shear stress are compared to baseline RANS. The paper presents the fundamental elements of procedure applied, including a brief description of the tools and methods and improvements achieved.

Reynolds-Averaged Navier–Stokes (RANS) methods continue to be the backbone of CFD-based design; however, the recent development of high-order unstructured solvers and meshing algorithms, combined with the lowering cost of HPC infrastructures, has the potential to allow for the introduction of high-fidelity simulations in the design loop, taking the role of a virtual wind tunnel. Extensive validation and verification is required over a broad design space. This is challenging for a number of reasons, including the range of operating conditions, the complexity of industrial geometries and their relative motion. A representative industrial low pressure turbine (LPT) cascade subject to wake passing interactions is analysed, adopting the incompressible Navier–Stokes solver implemented in the spectral/hp element framework Nektar++. The bar passing effect is modelled by leveraging a spectral-element/Fourier Smoothed Profile Method. The Reynolds sensitivity is analysed, focusing in detail on the dynamics of the separation bubble on the suction surface as well as the mean flow properties, wake profiles and loss estimations. The main findings are compared with experimental data, showing agreement in the prediction of wake traverses and losses across the entire range of flow regimes, the latter within 5% of the experimental measurements.