DNS vs. LES for Nuclear Thermal Hydraulics

Comparative analysis of turbulent heat transfer in square duct with heated foil boundary conditions

This project investigates turbulent heat transfer in a square duct configuration where a finite section of a single duct boundary is heated with an infinitely thin foil. The research compares Direct Numerical Simulation (DNS) and Large Eddy Simulation (LES) approaches to understand their relative accuracy and computational efficiency for nuclear thermal hydraulics applications.

Motivation and Background

Temperature fluctuations at heated surfaces in turbulent flows matter for nuclear reactor safety analysis, for example through thermal fatigue. Understanding these phenomena requires accurate prediction of near-wall thermal boundary layer behavior, which presents significant challenges for computational methods. This work addresses the fundamental question of when LES can reliably replace DNS for thermal hydraulics calculations.

Research Objectives

  • Method Validation: Compare DNS and LES-WALE model predictions for thermal boundary layer behavior
  • Accuracy Assessment: Quantify differences in temperature statistics and fluctuation predictions
  • Computational Efficiency: Evaluate the cost-benefit trade-offs between high-fidelity DNS and practical LES
  • Design Support: Provide guidance for experimental design in thermal hydraulics research

Key Results

Temperature Prediction Differences

The standard LES-WALE model predicts about 20% higher foil temperatures and temperature fluctuations than DNS.

Near-Wall Model Behavior

A detailed analysis of the LES-WALE model’s near-wall behavior explains the discrepancy. LES-WALE is highly accurate only on finely resolved meshes, and the cost of that refinement is hard to justify when DNS is an option.

Methodology

Numerical Setup

  • Code: OpenFOAM with second-order accurate schemes
  • Approach: Passive scalar approximation for thermal transport
  • Boundary Conditions: Constant heat flux at heated foil section
  • Reynolds Number: 10,000 (fully developed turbulent flow)
Computational geometry showing square duct configuration with heated foil boundary section for DNS vs LES comparison

Comparative Framework

  • DNS: High-resolution reference solution capturing all turbulent scales
  • LES: Wall-Adapting Local Eddy-viscosity (WALE) subgrid-scale model
  • Mesh Sensitivity: Systematic grid refinement studies for both methods

Applications and Impact

This research supports:

  • Nuclear reactor thermal analysis by providing method accuracy guidelines
  • Experimental design for heated foil thermometry studies
  • CFD code validation for nuclear thermal hydraulics applications
  • Computational resource planning for industrial simulations

Experimental Connection

The simulations supported the design of the heated-foil experiment in the THELMA laboratory at Jožef Stefan Institute. LES of the planned setup set the foil heating power, and the analysis of coherent thermal structures determined the spatial resolution and frame rate the high-speed thermal camera needs.

Results

This work is published in (Kren et al., 2024; Tiselj et al., 2021; Sharma et al., 2023; Kren et al., 2022; Kren et al., 2020) and in my MSc thesis (Kren, 2020).

References

2024

  1. DNSLESCoherent.jpg
    DNS vs. LES: Turbulent flow in square duct with heated foil boundary
    Jan Kren, Blaž Mikuž, and Iztok Tiselj
    International Journal of Heat and Fluid Flow, Jul 2024

2023

  1. Thermal streak spacing in fully developed duct flow with different Reynolds and Prandtl numbers
    Mohit Pramod Sharma, Iztok Tiselj, Jan Kren, and 2 more authors
    Heat and Mass Transfer, Jul 2023

2022

  1. DNS OF SQUARE DUCT FLOW WITH HEATED FOIL BOUNDARY
    Jan Kren, Blaž Mikuž, and Iztok Tiselj
    2022

2021

  1. Investigation of turbulent flow in square duct with heated foil thermometry
    Iztok Tiselj, Mohit Pramod Sharma, Boštjan Zajec, and 4 more authors
    International Journal of Heat and Mass Transfer, Aug 2021

2020

  1. Large Eddy Simulation Of Turbulent Flow With Different Prandtl Numbers Near Uniformly Heated Wall
    Jan Kren, Blaž Mikuž, and Iztok Tiselj
    In Proceedings of the 29th International Conference Nuclear Energy for New Europe, 2020
    NENE 2020
  2. Simulacije turbulentnega prenosa toplote ob greti foliji z metodo velikih vrtincev
    Jan Kren
    Univerza v Ljubljani, Fakulteta za matematiko in fiziko, 2020