ME3621

Applied Fluid
Mechanics (+CFD)
James Tyacke
Edward Smith
Jun Xia
 This session will be recorded

Brunel University London ME3621 Applied Fluid Mechanics 2

 Feedback

• Use the QR code to go to feedback

• You can ask questions or make comments at any time, either linked to
your name (if you put it in) or anonymously

Brunel University London ME3621 Applied Fluid Mechanics 3

 Summary

Aim
• To extend understanding of turbulence associated with different problems

Objectives
• Identify key aspects of different flows
• Choose appropriate modelling
• Understand the trade-offs made

Brunel University London ME3621 Applied Fluid Mechanics 4

 This shows the mean velocity
Range of turbulence modelling gradients noted earlier are used
to approximate the Reynolds
stresses using the linear
Inviscid methods – E.g. panel
methods, thin aerofoil theory, etc proportionality constant μt

Approximated Reynolds stresses

(linear/eddy viscosity model)

Exact Reynolds stress equation

(models still contain empiricism)

In industry:
• RANS is most common due to cost

• High uncertainties in accuracy

mean LES is now being used in
targeted ways and its use will grow

Brunel University London ME3621 Applied Fluid Mechanics 5

 Key LES problem is wall streak resolution
• We have noted LES cost increases with Re

LES
y+<100 RANS •LES Cost α Re2.5*
•Resolution critical
U
•Hybrid RANS-LES Cost α Re
*Piomelli, AIAA-2008-396
•Resolution relaxed

LES RANS-LES
(streaks covered with a RANS layer)

Brunel University London ME3621 Applied Fluid Mechanics 6

 LES/LES-RANS Resolution requirements
For y+<100 only LES
Low Re
High
Re

Now
Hybrid
2014

N ≠ f(Re)

Wake

This assumes a single side of a plate of chord=span.

Given a chord-based Re, we can estimate mesh requirements
Adapted from Leschziner (2009), Piomelli and Balaras (2002)
Brunel University London ME3621 Applied Fluid Mechanics 7
 Mesh estimation example

Assume hybrid RANS-LES, Re=500,000

Blade of aspect ratio (span:chord) 5
Including endwalls

N=10million

one side = 5*10million

2 sides = 2*5*10million
= 100million
+2 endwalls (assuming AR~1 each) = + 2*10million

Total = 100 + 20 = 120 million nodes

(For LES this might be 50million for AR=1  600 million nodes)

Note this is for y+< 100 only, real mesh will be somewhat higher but not
as significant

Brunel University London ME3621 Applied Fluid Mechanics 8

 Detached Eddy Simulation
(one form of RANS-LES)
Idea is to model attached BL
using RANS (where RANS
performs well and LES is
expensive) and LES in
separated zones (where RANS
fails and LES does well)

DES covers the entire BL not just

the streaks and there are many
variants – DDES, IDDES, ZDES

Brunel University London ME3621 Applied Fluid Mechanics 9

 ZDES inspired classification

I. Geometry induced separation

II. Pressure gradient induced separation
III. Boundary layer dynamics dependent separation

S. Deck, 2012, Recent improvements in the zonal detached eddy simulation (ZDES) formulation

Brunel University London ME3621 Applied Fluid Mechanics 10

 Note these are just examples of
Flow classification flow physics in the same
machine (gas turbine)

Impact / Ease = HIGH MODERATE

A: Wake (Re independent) B: Low Re C: High Re
Bluff bodies

High Pressure
Turbine
Wall jet Low Pressure Labyrinth seals
Turbine

Jets
Rim seals

Mixed
LES LES RANS-LES

Brunel University London ME3621 Applied Fluid Mechanics 11

 Inflow classification

A: Wake (Re independent) B: Low Re C: High Re

Ribbed passages

Wall jet High Pressure

Turbine
High importance

Low Pressure Lab seals

Jets Turbine

Rim seals

Low importance Crucial Low importance

Brunel University London ME3621 Applied Fluid Mechanics 12

 Case overview – High Pressure Turbine (HPT) blade

Flow type C: High Re

Method RANS-(I)LES
Inflow type Critical: Turbulent BL,
combustor eddies

Benefits Moderate: Modelling

cooling holes challenging

Uses Understand flow physics,

improvement of lower
Re ≈ 0.6 million, order modelling
N = 5.5 million

Brunel University London ME3621 Applied Fluid Mechanics 13

 Validation – HPT blade

Validation - No Cavity

10% DS
Cp

Interface

x/cax
25% span 2.3% span

50% span 12.5% span

Brunel University London ME3621 Applied Fluid Mechanics 14

 LES uses – RANS validation and improvement
Validation with LES with Cavity (no Experiments available)

z/h RANS

Real inflow:

RANS-HJ

LES

Cpt,PA

Jefferson-Loveday et al.
ASME J. Turbomach.

Brunel University London ME3621 Applied Fluid Mechanics 15

 LES/DNS uses – real wall roughness

Smooth
Sand
grain

Smooth Rough u+
y+

•80% higher skin friction

•Improve RANS modelling

Brunel University London ME3621 Applied Fluid Mechanics 16

 Case overview – labyrinth seal

Flow type C: Wake+high Re BL

Method RANS-(I)LES

Inflow type Simple: Large geometric

scales form rapidly

Benefits High: Accurate and

consistent for all
operating conditions.
Lower cost.
Uses Test new designs, add to
and improve existing
databases, improve low-
order modelling
RANS
(Correlations poor in
some cases)

Brunel University London ME3621 Applied Fluid Mechanics 17

 Validation – Labyrinth seal LES/LES-RANS

LES

Vortex shedding from

seal teeth after severe
acceleration
Brunel University London ME3621 Applied Fluid Mechanics 18
 Labyrinth seal optimisation with LES

Brunel University London ME3621 Applied Fluid Mechanics 19

 Baseline

Optimised

Brunel University London ME3621 Applied Fluid Mechanics 20

 Case overview Low Pressure Turbine blade

Flow type B: Low Re, transitional

Method (I)LES

Inflow type Critical: Wakes, FS

turbulence, acoustic
waves, endwalls

Benefits High: LES crucial to

capture and understand
complex flow physics for
future technologies, low
cost
Uses Understand transition
processes, real surface
y/L roughness effects,
improve performance
estimates.
x/L

Brunel University London ME3621 Applied Fluid Mechanics 21

 Case overview – internal cooling passages

Flow type A: Wake

Method (I)LES
Inflow type Simple: Large geometric
scales form rapidly

Benefits High: Accurate and

consistent HT prediction
for complex geometries.
Low cost.

Uses Study of developing

flows and local hot spots.
Improve designs directly
Nu

- fast turnaround.
Optimisation of existing
designs. Refine lower
order models
x/h

Brunel University London ME3621 Applied Fluid Mechanics 22

 Validation – internal cooling passages

LES

u’/U0
U/U0

y/D
x/D x/P

Nu
x/D

w/U0 x/h

Brunel University London ME3621 Applied Fluid Mechanics 23

 Case overview – wall jet heat transfer

Mixing flow
Flow type A: Wake
Method (I)LES

Inflow type Simple: Large

geometric scales form
z/s rapidly from
turbulators+pedastals
x/s
Benefits High: Accurate and
consistent HT
RANS prediction for complex
geometries and a
variety of BR. Low
cost
η

Uses Improve designs

directly - fast
turnaround. Refine
LES lower order models.

x/s
Brunel University London ME3621 Applied Fluid Mechanics 24
 LES computational cost (CPU time + £)

•500 CPU cores

(Modest cluster)
•Mean quantities ~ half
runtime
•GPU 1/10th cost CPU
•Lab seal rotor cost >15-20k

Flow type LES cost using CPUs

for turbulence and HT data

Ribbed, ducts, CBTEs A: Wake <£0.25k (1 day)

LPT B: Low Re, Highly complex, <£1k (0.5 week)

incoming wakes
Lab seals, cavities C: Wake+high Re BL, inlet swirl <£1k (0.5 week)

Brunel University London ME3621 Applied Fluid Mechanics 25

 Summary

• Use of zonalized RANS-(I)LES and (I)LES will increase in

industrial CFD
• Flow classification based on key aspects:
• Turbulence modelling
• Inflow
• See greater use in RANS + lower order model
calibration/development
• LES still needs physical insight by analysts but much less than
RANS  best practices: easier within a confined application
• Economic costs cheaper than many rig tests
• Note: large meshes are now of order 10BIllion cells on >100k
CPU cores

Brunel University London ME3621 Applied Fluid Mechanics 26