Development and Application of Embedded Large Eddy Simulation for Internal Fluid Flows in Industrial Computational Fluid Dynamics

Abstract

Computational Fluid Dynamics can be a valuable tool in the detailed analysis of flow phenomena, for example when one needs to understand the complex flow around pipe networks. In cases where transient flow phenomena result from complex geometries it is necessary to undertake 3D Computational Fluid Dynamics (CFD) for either a range of piecewise components or larger combined scenarios. The most difficult challenge in the accurate prediction of unsteady flow arises from turbulence. For most industrial applications, it is too computationally expensive to resolve all the scales of turbulence as required in direct numerical simulation (DNS). Hence, recourse must be sought in turbulence modelling. In Large Eddy Simulation (LES) only the smallest scales are modelled, but even the application of full wall-resolved LES is still prohibitively expensive for industry. The default prediction method thus continues to be based on Reynolds-averaged Navier-Stokes (RANS) equations. In many cases RANS models are not sufficiently accurate and lack the broad frequency content required for the faithful computation of fluctuating wall temperatures, or flow induced vibrations which can only be obtained with higher fidelity methods. This thesis presents the developments of improved methodologies for Embedded Large Eddy Simulation (ELES); to enable a breakthrough in the Reynolds-number range accessible with high-accuracy flow, noise and vibration simulations. The aim of this project it to use CFD for exploring the physics of flows through pipe networks, and also to develop and validate new CFD techniques for study of such problems in industry. Firstly, improved formulations to the techniques used for RANS to LES transfer in hybrid RANS-LES are demonstrated. A number of validation cases are able to show the suitability of an improved formulation of the Synthetic Eddy Method (SEM) for use in an ELES framework. A detailed review of the existing methodologies in the framework of hybrid RANS- LES is provided including an assessment of the suitability of ELES for internal flows in industrial computations. Following this a detailed analysis of the flow structures developed downstream of three 90o bends with varying radius of curvature is demonstrated with the application of LES. Three-dimensional mode decomposition techniques, namely Proper Orthogonal Decomposition (POD) & Dynamic Mode Decomposition (DMD), are able to provide insight into the dynamics of the dominant flow structures created. The secondary motions in pipe bends give rise to low-frequency oscillations in the flow and can lead to pipe fatigue in Nuclear and gas pipe lines. The understanding of the underlying dynamics of these motions is therefore vital for a number of industries. Finally, a two-way coupled nesting procedure for Embedded Large Eddy Simulation is proposed as an alternate solution to achieving high fidelity simulation within industrial CFD. Turbulent fluctuations at the inlet to the nested LES domain are provided from a nested form of the SEM, whilst a blending region near the LES outlet provides a smooth transition back to RANS. Of particular merit with this approach is the ease of including multiple uncoupled LES regions throughout the domain, whilst ensuring global consistency of the flow. Prediction of the flow through numerous channel flow configurations and flow over a forward-backward facing step are in good agreement with reference data.

Date of Award	1 Aug 2019
Original language	English
Awarding Institution	The University of Manchester
Supervisor	Timothy Craft (Supervisor) & Alistair Revell (Supervisor)