Multiphase flow from a civil engineering perspective Benjamin - - PowerPoint PPT Presentation

www.hece.ulg.ac.be

Multiphase flow from a civil engineering perspective

Benjamin Dewals, Sébastien Erpicum, Pierre Archambeau & Michel Pirotton

HECE – Hydraulics in Environmental and Civil Engineering University of Liege

University of Liege

Founded in 1817 All faculties, 20,000 students

Department ArGEnCo

Architecture, Geology, Environmental and Civil Engineering 250 staff members, 30 academics Research group HECE

APPLIED HYDRODYNAMICS AND HYDRAULIC CONSTRUCTIONS Michel Pirotton Full Professor HYDROLOGY, FREE SURFACE AND PRESSURIZED FLOW Pierre Archambeau Research Associate ENVIRONMENTAL HYDRAULICS Benjamin Dewals Associate Professor LABORATORY OF ENGINEERING HYDRAULICS Sébastien Erpicum Laboratory manager

www.hece.ulg.ac.be

The modelling system WOLF, developed by the team, enables to study a wide range of flows in civil engineering

HYDRO WOLF

Rainfall-runoff modelling

1D WOLF

Channel networks

2D / 3D WOLF

Detailed flow analysis

• Turbulence modelling
• Non-hydrostatic features
• Bottom curvature

Clear water Transport of air, sediments, pollutant, …

Self-developed pre- and post-processing user interface

www.hece.ulg.ac.be

Composite modelling enhances both understanding of basic processes and performance of numerical models

Pumps room Main test slab

Test slabs 1100 m² Storage tank 400 m³ Regulated pumps 400 l/s 3 (tilting) flumes

up to 20 m long

Workshops

(synthetic materials, wood, steel, concrete)

www.hece.ulg.ac.be

Three main types of multiphase flow are commonly addressed in civil engineering

Aerated flow

• n hydraulic structures

e.g., on spillways and in pipes

Debris flow, geophysical granular flow

e.g., waste dump failure

Sediment transport

e.g., reservoir sedimentation, flushing

www.hece.ulg.ac.be

The type of weir crest significantly influences the location of the inception point on a stepped spillway

Ogee-crest Piano Key Weir 1 Piano Key Weir 2

E.g., Riou dam, France

IAHR Media library

Experimental facility at our laboratory

www.hece.ulg.ac.be

Air-water interactions alter the flow dynamics and hence the discharge capacity in bottom outlets of dams

• b. View of the

physical model

3D view of a typical bottom outlet gallery 1:30 scale model Example: design of the bottom outlet of a dam

www.hece.ulg.ac.be

Between smooth stratified flow and pressurized flow, intermediate flow show distinct multiphase characteristics

0,0 0,8 5 50

Upstream pressure head [m] Flow discharge [l/s]

Smooth stratified flow Bubbly flow Pure water fully pressurized flow Intermittent flow: Plug/Slug flow Wavy stratified flow Upstream reservoir bottom level

Instability

0.3 0.53 0.4 0.7

Interval of variation Without air vent With air vent

0,0 0,8 5 50

Upstream pressure head [m] Flow discharge [l/s]

Smooth stratified flow Bubbly flow Pure water fully pressurized flow Intermittent flow: Plug/Slug flow Wavy stratified flow Upstream reservoir bottom level

0.3 0.53 0.4 0.7

Without air vent With air vent Classic Preissmann

0,0 0,8 5 50

Upstream pressure head [m] Flow discharge [l/s]

Smooth stratified flow Bubbly flow Pure water fully pressurized flow Intermittent flow: Plug/Slug flow Wavy stratified flow Upstream reservoir bottom level

0.3 0.53 0.4 0.7

Without air vent With air vent Negative Preissmann Classic Preissmann

W a t e r F l

• w

Water Flow W a t e r F l

• w

Water Flow W a t e r F l

• w

Air-water interactions alter the discharge capacity

• f the bottom outlet

Numerical result obtained from a 1D Homogeneous Equilibrium model

Kerger et al. Adv. Eng. Soft. (2011)

www.hece.ulg.ac.be

Modelling the failure of waste dumps lies at the frontier between geotechnical and hydraulic engineering

Jupille, near Liege (1961) Run out of ~ 200,000 t

Mining tip collapse in England (1967)

Fly ash dump

www.hece.ulg.ac.be

Modelling the failure of waste dumps lies at the frontier between geotechnical and hydraulic engineering

Topography: Steep-sided valley ~ 20 – 30 Mild-sloped thalweg ~ 3 Physical properties:  = 1000 – 1400 kg/m³ d50 = 40 µm Quasi-static mechanical characteristics:  = 20 Jupille, near Liege (1961) Run out of ~ 200,000 t

• ver ~ 700 m

Material: fly ashes 700m

Fly ash dump

Lateral dikes in the deposits 10m deep deposits

www.hece.ulg.ac.be

Three main types of multiphase flow are commonly addressed in civil engineering

Aerated flow

• n hydraulic structures

e.g., on spillways and in pipes

Debris flow, geophysical granular flow

e.g., waste dump failure

Sediment transport

e.g., reservoir sedimentation, flushing ► Basic research,

with practical applications in mind …

Reservoir sedimentation: a worldwide challenge

2000 4000 6000 1900 1950 2000

Volume worldwide ( billion m³)

Total storage worldwide Net storage capacity Sedimentation worldwide  1-2% of worldwide storage capacity is lost every year!

ICOLD (2009)

World bank: “Last century was used to build reservoirs. This one will be used to solve sediment problems...”

www.hece.ulg.ac.be

Complexity in morphodynamic modelling stems from the multiscale processes in space and time

Multiscale in space Mutliscale in time Soil erosion

• n the watershed

1 to 105 km² Reservoir sedimentation 1 to 103 m Transport in rivers 10 to 103 km

Flushing operations planned during hours, days or weeks Bank failures: within seconds! Reservoir sedimentation: within years or decades

differences of up to 9 orders of magnitude! Multiple processes

• Bed load
• Suspended load
• Sediment exchange
• Bank failures
• ...

www.hece.ulg.ac.be

Morphodynamic modelling system enabling four levels of coupling between sub-models

Increasing relative time scale of morphological changes

Sediment transport +morphodynamics (steady approach) ► steady state Flow sub-model

Next iteration

Synchronous resolution Sequential resolution Iterative process Stochastic particle tracking

Sediment transport +morphodynamics (unsteady, 1 step) ► steady state Flow sub-model

Next time step

Sediment transport +morphodynamics (unsteady, 1 step) Flow sub-model (unsteady, 1 step)

Next time step

Sediment transport

• nly

(unsteady, 1 step) Flow sub-model (unsteady, 1 step)

Next time step Maintenance Bank failure ~ 100s Flushing ~ 101-103s Sedimentation ~ 104-108s → ∞

www.hece.ulg.ac.be

Shallow rectangular reservoirs

Flow pattern

• Experimental observations
• Numerical prediction
• Theoretical stability

analysis

Sedimentation pattern

• Experimental observations
• Example of practical

implication

• Morphodynamic modelling

Feedback from sedimentation

• n flow pattern

www.hach.ulg.ac.be

Physical modelling

Dufresne, Dewals et al. (2010)

Water depth = 20 cm Velocity at inlet = 0.28 m/s Inlet channel width = 0.3 m Basin width = 1 m

Lmax = 7 m

www.hach.ulg.ac.be

Flow pattern classification

2 . 6

40 . 60 .

b B L 8 . 6

40 . 60 .

b B L

Dufresne, Dewals et al. (2010)

Q L B b B

www.hece.ulg.ac.be

Shallow rectangular reservoirs

Flow pattern

• Experimental observations
• Numerical prediction
• Theoretical stability

analysis

Sedimentation pattern

• Experimental observations
• Example of practical

implication

• Morphodynamic modelling

Feedback from sedimentation

• n flow pattern

www.hach.ulg.ac.be

No disturbance at inflow Slight disturbance at inflow (1%)

Can those flow patterns be predicted by a 2DH flow model?

Shallow-water equations (WOLF 2D) Finite volume scheme:

• 2nd order accurate
• self-developed FVS

Eddy viscosity:

• algebraic model: T = h u*
• two-length-scale k- model

Bottom and wall friction

Observed flow field

Dewals, Kantoush et al. (2008) Dufresne, Dewals et al. (2011)

www.hach.ulg.ac.be

All flow patterns well reproduced, based on a seed for asymmetry

”Short” reservoirs S0 pattern Intermediate length A1 pattern ”Long” reservoirs A2 pattern

Dufresne, Dewals et al. (2011)

www.hach.ulg.ac.be

Co-existence of two stable flow patterns depending on flow history

Symmetric IC Deviated IC

Dufresne, Dewals et al. (2011)

Symmetric IC Deviated IC

www.hece.ulg.ac.be

Shallow rectangular reservoirs

Flow pattern

• Experimental observations
• Numerical prediction
• Theoretical stability

analysis

Sedimentation pattern

• Experimental observations
• Example of practical

implication

• Morphodynamic modelling

Feedback from sedimentation

• n flow pattern

0 m 1 m 0 m 1 m 2 m

m 1 m 2 m 3 m 4 m 5 m 6 m 7 m Experiment ST1-a

Flow pattern strongly influences trapping efficiency as can be deduced from experimental data and numerical simulations

Observed location of deposits Dufresne, Dewals et al. (2011)

Material = walnut shell

s = 1.5

d50 = 50 m ws = 1 mm/s Cin = 3.0 g/l Mostly suspended load ▲ Material = Granular plastic (Styrolux 656 C)

s = 1.020 d50 = 2.5 mm ws = 25 mm/s

Cin = 0.5 g/l Mainly bedload

Observed deposits patterns Kantoush (2008), PhD thesis, EPFL

www.hece.ulg.ac.be

Sudden rise in trapping efficiency as flow bifurcates from S0 to A1

www.hece.ulg.ac.be

Sudden rise in trapping efficiency as flow bifurcates from S0 to A1

TE ~ 50% Global TE ~ 65%

i.e. increase by almost 1/3 for a similar spatial extent!

BUT if S0 flow pattern, Global TE < 15-20% !! 1 – TE ~ 70% ▲ 1 – TE ~ 70% 70% ▲ TE = Trapping efficiency 1 – TE ~ 70% 70% 70% ▲

www.hece.ulg.ac.be

Results of morphodynamic modelling (on-going)

Simulated maps of thickness of sediment deposits

L-R configuration C-R configuration

Satisfactory agreement in the main jet Underestimation of sediment deposits thickness

in the quiescent zones, where TKE is underpredicted

www.hece.ulg.ac.be

Shallow rectangular reservoirs

Flow pattern

• Experimental observations
• Numerical prediction
• Theoretical stability

analysis

Sedimentation pattern

• Experimental observations
• Example of practical

implication

• Morphodynamic modelling

Feedback from sedimentation

• n flow pattern

www.hece.ulg.ac.be

Effect of inlet and outlet location on flow patterns

C-C L-L L-R C-R

Experiments (Camnasio): Rein = 112,000; Fr in =0.1; h/b= 0.8 - WITH SEDIMENTS Experiments (Camnasio): Rein = 112,000; Fr in = 0.1; h/b= 0.8 - WITHOUT SEDIMENTS

www.hece.ulg.ac.be

Flow history has a significant role

Experiment (Camnasio): Rein = 112,000; Fr in = 0.1; h/b= 0.8 - WITHOUT SEDIMENTS [L-R configuration]

Starting from an empty basin tends to facilitate flow reattachement

www.hece.ulg.ac.be

Feedback of morphodynamics on the flow pattern

Mean flow Sediment transport Morpho- dynamics Reservoir geometry

Experiments (Camnasio): WITHOUT SEDIMENTS Experiments (Camnasio): WITH SEDIMENTS (< 30 min)

www.hece.ulg.ac.be

Feedback of morphodynamics on the flow pattern

Mean flow Sediment transport Morpho- dynamics Reservoir geometry

No change in the computed flow patterns EXCEPT FOR C-DX after morphological evolution during 4h, implemented as a time- varying topography

Mean flow Sediment transport Morpho- dynamics Reservoir geometry Roughness

www.hece.ulg.ac.be

Feedback of morphodynamics on the flow pattern

Mean flow Sediment transport Morpho- dynamics Reservoir geometry Roughness

ks (m) cf (-) S (-) Smooth bottom 4.4 × 10-3 0.02 Roughness corresponding to d50, d90 89 - 215 × 10-6 4.6 - 4.8 × 10-3 0.02 Roughness including bed forms effect 0.005 (maximum assumed bed form height) 8.3 × 10-3 0.04

No change in the computed flow patterns when significant changes are introduced in bed roughness In agreement with Babarutsi et al. (1989 and 1991) and Chu (2004) for unilateral expansions: Bed friction number S = cf (B-b) / 4 h < 0.05 flow classified as “non-frictional”

www.hece.ulg.ac.be

Feedback of sediment transport on the flow pattern

Mean flow Sediment transport Morpho- dynamics Reservoir geometry Roughness Turbulence

• Computed flow pattern tends to match experimental tests

with suspended load for reduced turbulence parameter

• Possible explanation: damping of turbulence caused by

suspended sediments

• “Effects of suspended load on

turbulence can become appreciable even for volumetric concentrations as low as 10-6” Cao & Carling (2001)

• In the present case: volumetric

concentration ~ 1.3 × 10-3

www.hece.ulg.ac.be

► Simple geometries, such as rectangular basins, lead to complex flow patterns involving flow instabilities ► Prerequisite to succeed in predicting sedimentation: predict flow pattern and assess the flow stability ► Depth-averaged flow model with simple turbulence closure achieves prediction of global flow pattern ► Morphodynamic model provides depth of deposits and feed-back on the flow field ► Modelling of diffusion needs to be enhanced accounting for (i) 3D effects and (ii) interactions with suspended load Some conclusions … + transitions