A model for unsteady mixed flows in non uniform closed water pipes - - PowerPoint PPT Presentation

a model for unsteady mixed flows in non uniform closed
SMART_READER_LITE
LIVE PREVIEW

A model for unsteady mixed flows in non uniform closed water pipes - - PowerPoint PPT Presentation

A model for unsteady mixed flows in non uniform closed water pipes and a well-balanced finite volume scheme M. Ersoy 1, Christian Bourdarias 2 and St ephane Gerbi 3 LMB, Besan con, the 10 February 2011 1. BCAM, Spain, mersoy@bcamath.org 2.


slide-1
SLIDE 1

A model for unsteady mixed flows in non uniform closed water pipes and a well-balanced finite volume scheme

  • M. Ersoy 1, Christian Bourdarias 2 and St´

ephane Gerbi 3 LMB, Besan¸ con, the 10 February 2011

  • 1. BCAM, Spain, mersoy@bcamath.org
  • 2. LAMA–Savoie, France, christian.bourdarias@univ-savoie.fr
  • 3. LAMA–Savoie, France, stephane.gerbi@univ-savoie.fr
slide-2
SLIDE 2

Outline of the talk

Outline of the talk

1 Unsteady mixed flows : PFS equations (Pressurized

and Free Surface)

Previous works Formal derivation of the free surface and pressurized model A coupling : the PFS-model

2 Finite Volume discretization

Discretization of the space domain Explicit first order VFRoe scheme

  • 1. The Case of a non transition point
  • 2. The Case of a transition point
  • 3. Update of the cell state
  • 4. Approximation of the convection matrix

3 Numerical experiments 4 Conclusion and perspectives

  • M. Ersoy (BCAM)

PFS-model and VFRoe solver LMB, Besan¸ con, the 10 February 2011 2 / 40

slide-3
SLIDE 3

Outline

Outline

1 Unsteady mixed flows : PFS equations (Pressurized

and Free Surface)

Previous works Formal derivation of the free surface and pressurized model A coupling : the PFS-model

2 Finite Volume discretization

Discretization of the space domain Explicit first order VFRoe scheme

  • 1. The Case of a non transition point
  • 2. The Case of a transition point
  • 3. Update of the cell state
  • 4. Approximation of the convection matrix

3 Numerical experiments 4 Conclusion and perspectives

  • M. Ersoy (BCAM)

PFS-model and VFRoe solver LMB, Besan¸ con, the 10 February 2011 3 / 40

slide-4
SLIDE 4

Unsteady mixed flows in closed water pipes ?

Free surface area (SL) sections are not completely filled and the flow is incompressible. . .

  • M. Ersoy (BCAM)

PFS-model and VFRoe solver LMB, Besan¸ con, the 10 February 2011 4 / 40

slide-5
SLIDE 5

Unsteady mixed flows in closed water pipes ?

Free surface area (SL) sections are not completely filled and the flow is incompressible. . . Pressurized area (CH) sections are non completely filled and the flow is compressible. . .

  • M. Ersoy (BCAM)

PFS-model and VFRoe solver LMB, Besan¸ con, the 10 February 2011 4 / 40

slide-6
SLIDE 6

Unsteady mixed flows in closed water pipes ?

Free surface area (SL) sections are not completely filled and the flow is incompressible. . . Pressurized area (CH) sections are non completely filled and the flow is compressible. . . Transition point

  • M. Ersoy (BCAM)

PFS-model and VFRoe solver LMB, Besan¸ con, the 10 February 2011 4 / 40

slide-7
SLIDE 7

Examples of pipes

Orange-Fish tunnel Sewers . . . in Paris Forced pipe problems . . . at Minnesota http://www.sewerhistory.org/grfx/ misc/disaster.htm

  • M. Ersoy (BCAM)

PFS-model and VFRoe solver LMB, Besan¸ con, the 10 February 2011 5 / 40

slide-8
SLIDE 8

Outline

Outline

1 Unsteady mixed flows : PFS equations (Pressurized

and Free Surface)

Previous works Formal derivation of the free surface and pressurized model A coupling : the PFS-model

2 Finite Volume discretization

Discretization of the space domain Explicit first order VFRoe scheme

  • 1. The Case of a non transition point
  • 2. The Case of a transition point
  • 3. Update of the cell state
  • 4. Approximation of the convection matrix

3 Numerical experiments 4 Conclusion and perspectives

  • M. Ersoy (BCAM)

PFS-model and VFRoe solver LMB, Besan¸ con, the 10 February 2011 6 / 40

slide-9
SLIDE 9

Previous works

For free surface flows :

Generally

Saint-Venant equations :    ∂tA + ∂xQ = 0, ∂tQ + ∂x Q2 A + gI1(A)

  • = 0

with A(t, x) : wet area Q(t, x) : discharge I1(A) : hydrostatic pressure g : gravity

Advantage

Conservative formulation − → Easy numerical implementation

Hamam and McCorquodale (82), Trieu Dong (91), Musandji Fuamba (02), Vasconcelos et al (06)

  • M. Ersoy (BCAM)

PFS-model and VFRoe solver LMB, Besan¸ con, the 10 February 2011 7 / 40

slide-10
SLIDE 10

Previous works

For pressurized flows :

Generally

Allievi equations :    ∂tp + c2 gS ∂xQ = 0, ∂tQ + gS∂xp = 0 with p(t, x) : pressure Q(t, x) : discharge c(t, x) : sound speed S(x) : section

Advantage

Compressibility of water is taking into account = ⇒ Sub-atmospheric flows and over-pressurized flows are well computed

Drawback

Non conservative formulation = ⇒ Cannot be, at least easily, coupled to Saint-Venant equations

Winckler (93), Blommaert (00)

  • M. Ersoy (BCAM)

PFS-model and VFRoe solver LMB, Besan¸ con, the 10 February 2011 8 / 40

slide-11
SLIDE 11

Previous works

For mixed flows :

Generally

Saint-Venant with Preissmann slot artifact :    ∂tA + ∂xQ = 0, ∂tQ + ∂x Q2 A + gI1(A)

  • = 0

Advantage

Only one model for two types of flows.

Drawbacks

Incompressible Fluid = ⇒ Water hammer not well computed Pressurized sound speed ≃

  • S/Tfente =

⇒ adjustment of Tfente Depression = ⇒ seen as a free surface state

Preissmann (61), Cunge et al. (65), Baines et al. (92), Garcia-Navarro et al. (94), Capart et al. (97), Tseng (99)

  • M. Ersoy (BCAM)

PFS-model and VFRoe solver LMB, Besan¸ con, the 10 February 2011 9 / 40

slide-12
SLIDE 12

Our goal :

Use Saint-Venant equations for free surface flows

  • M. Ersoy (BCAM)

PFS-model and VFRoe solver LMB, Besan¸ con, the 10 February 2011 10 / 40

slide-13
SLIDE 13

Our goal :

Use Saint-Venant equations for free surface flows Write a pressurized model

◮ which takes into account the compressibility of water ◮ which takes into account the depression ◮ similar to Saint-Venant equations

  • M. Ersoy (BCAM)

PFS-model and VFRoe solver LMB, Besan¸ con, the 10 February 2011 10 / 40

slide-14
SLIDE 14

Our goal :

Use Saint-Venant equations for free surface flows Write a pressurized model

◮ which takes into account the compressibility of water ◮ which takes into account the depression ◮ similar to Saint-Venant equations

Get one model for mixed flows To be able to simulate, for instance :

  • C. Bourdarias and S. Gerbi

A finite volume scheme for a model coupling free surface and pressurized flows in pipes.

  • J. Comp. Appl. Math., 209(1) :109–131, 2007.
  • M. Ersoy (BCAM)

PFS-model and VFRoe solver LMB, Besan¸ con, the 10 February 2011 10 / 40

slide-15
SLIDE 15

Outline

Outline

1 Unsteady mixed flows : PFS equations (Pressurized

and Free Surface)

Previous works Formal derivation of the free surface and pressurized model A coupling : the PFS-model

2 Finite Volume discretization

Discretization of the space domain Explicit first order VFRoe scheme

  • 1. The Case of a non transition point
  • 2. The Case of a transition point
  • 3. Update of the cell state
  • 4. Approximation of the convection matrix

3 Numerical experiments 4 Conclusion and perspectives

  • M. Ersoy (BCAM)

PFS-model and VFRoe solver LMB, Besan¸ con, the 10 February 2011 11 / 40

slide-16
SLIDE 16

Derivation of the free surface model

3D Incompressible Euler equations

ρ0div(U) = ρ0(∂tU + U · ∇U) + ∇p = ρ0F Method :

1 Write Euler equations in curvilinear coordinates. 2 Write equations in non-dimensional form using the small parameter ǫ = H/L

and takes ǫ = 0.

3 Section averaging U 2 ≈ U U and U V ≈ U V . 4 Introduce Asl(t, x) : wet area, Qsl(t, x) discharge given by :

Asl(t, x) =

  • Ω(t,x)

dydz, Qsl(t, x) = Asl(t, x)u(t, x) u(t, x) = 1 Asl(t, x)

  • Ω(t,x)

U(t, x) dydz

  • M. Ersoy (BCAM)

PFS-model and VFRoe solver LMB, Besan¸ con, the 10 February 2011 12 / 40

slide-17
SLIDE 17

Derivation of the free surface model

3D Incompressible Euler equations

ρ0div(U) = ρ0(∂tU + U · ∇U) + ∇p = ρ0F Method :

1 Write Euler equations in curvilinear coordinates. 2 Write equations in non-dimensional form using the small parameter ǫ = H/L

and takes ǫ = 0.

3 Section averaging U 2 ≈ U U and U V ≈ U V . 4 Introduce Asl(t, x) : wet area, Qsl(t, x) discharge given by :

Asl(t, x) =

  • Ω(t,x)

dydz, Qsl(t, x) = Asl(t, x)u(t, x) u(t, x) = 1 Asl(t, x)

  • Ω(t,x)

U(t, x) dydz

J.-F. Gerbeau, B. Perthame Derivation of viscous Saint-Venant System for Laminar Shallow Water ; Numerical Validation. Discrete and Continuous Dynamical Systems, Ser. B, Vol. 1, Num. 1, 89–102, 2001.

  • F. Marche

Derivation of a new two-dimensional viscous shallow water model with varying topography, bottom friction and capillary effects. European Journal of Mechanic B/Fluid, 26 (2007), 49–63.

  • M. Ersoy (BCAM)

PFS-model and VFRoe solver LMB, Besan¸ con, the 10 February 2011 12 / 40

slide-18
SLIDE 18

Derivation of the free surface model

3D Incompressible Euler equations

ρ0div(U) = ρ0(∂tU + U · ∇U) + ∇p = ρ0F Method :

1 Write Euler equations in curvilinear coordinates. 2 Write equations in non-dimensional form using the small parameter ǫ = H/L

and takes ǫ = 0.

3 Section averaging U 2 ≈ U U and U V ≈ U V . 4 Introduce Asl(t, x) : wet area, Qsl(t, x) discharge given by :

Asl(t, x) =

  • Ω(t,x)

dydz, Qsl(t, x) = Asl(t, x)u(t, x) u(t, x) = 1 Asl(t, x)

  • Ω(t,x)

U(t, x) dydz

  • M. Ersoy (BCAM)

PFS-model and VFRoe solver LMB, Besan¸ con, the 10 February 2011 12 / 40

slide-19
SLIDE 19

Derivation of the free surface model

3D Incompressible Euler equations

ρ0div(U) = ρ0(∂tU + U · ∇U) + ∇p = ρ0F Method :

1 Write Euler equations in curvilinear coordinates. 2 Write equations in non-dimensional form using the small parameter ǫ = H/L

and takes ǫ = 0.

3 Section averaging U 2 ≈ U U and U V ≈ U V . 4 Introduce Asl(t, x) : wet area, Qsl(t, x) discharge given by :

Asl(t, x) =

  • Ω(t,x)

dydz, Qsl(t, x) = Asl(t, x)u(t, x) u(t, x) = 1 Asl(t, x)

  • Ω(t,x)

U(t, x) dydz

  • M. Ersoy (BCAM)

PFS-model and VFRoe solver LMB, Besan¸ con, the 10 February 2011 12 / 40

slide-20
SLIDE 20

The free surface model

               ∂tAsl + ∂xQsl = 0, ∂tQsl + ∂x Q2

sl

Asl + psl(x, Asl)

  • =

−gAsl d Z dx + Prsl(x, Asl) − G(x, Asl) with psl = gI1(x, Asl) cos θ : hydrostatic pressure law Prsl = gI2(x, Asl) cos θ : pressure source term G = gAslz d dx cos θ : curvature source term

  • M. Ersoy (BCAM)

PFS-model and VFRoe solver LMB, Besan¸ con, the 10 February 2011 13 / 40

slide-21
SLIDE 21

The free surface model

               ∂tAsl + ∂xQsl = 0, ∂tQsl + ∂x Q2

sl

Asl + psl(x, Asl)

  • =

−gAsl d Z dx + Prsl(x, Asl) − G(x, Asl) − gK(x, Asl)Qsl|Qsl| Asl

  • friction added after the derivation

with psl = gI1(x, Asl) cos θ : hydrostatic pressure law Prsl = gI2(x, Asl) cos θ : pressure source term G = gAslz d dx cos θ : curvature source term K = 1 K2

sRh(Asl)4/3

: Manning-Strickler law

  • M. Ersoy (BCAM)

PFS-model and VFRoe solver LMB, Besan¸ con, the 10 February 2011 13 / 40

slide-22
SLIDE 22

Derivation of the pressurized model

3D isentropic compressible equations

∂tρ + div(ρU) = 0 ∂t(ρU) + div(ρU ⊗ U) + ∇p = ρF with p = pa + ρ − ρ0 c2 with c sound speed Method :

1 Write Euler equations in curvilinear coordinates. 2 Write equations in non-dimensional form using the small parameter ǫ = H/L

and takes ǫ = 0.

3 Section averaging ρU ≈ ρU and ρU 2 ≈ ρU U. 4 Introduce Ach(t, x) : equivalent wet area, Qch(t, x) discharge given by :

Ach(t, x) = ρ ρ0 S(x), Qch(t, x) = Ach(t, x)u(t, x) u(t, x) = 1 S(x)

  • Ω(x)

U(t, x) dydz

  • M. Ersoy (BCAM)

PFS-model and VFRoe solver LMB, Besan¸ con, the 10 February 2011 14 / 40

slide-23
SLIDE 23

Derivation of the pressurized model

3D isentropic compressible equations

∂tρ + div(ρU) = 0 ∂t(ρU) + div(ρU ⊗ U) + ∇p = ρF with p = pa + ρ − ρ0 c2 with c sound speed Method :

1 Write Euler equations in curvilinear coordinates. 2 Write equations in non-dimensional form using the small parameter ǫ = H/L

and takes ǫ = 0.

3 Section averaging ρU ≈ ρU and ρU 2 ≈ ρU U. 4 Introduce Ach(t, x) : equivalent wet area, Qch(t, x) discharge given by :

Ach(t, x) = ρ ρ0 S(x), Qch(t, x) = Ach(t, x)u(t, x) u(t, x) = 1 S(x)

  • Ω(x)

U(t, x) dydz

  • M. Ersoy (BCAM)

PFS-model and VFRoe solver LMB, Besan¸ con, the 10 February 2011 14 / 40

slide-24
SLIDE 24

Derivation of the pressurized model

3D isentropic compressible equations

∂tρ + div(ρU) = 0 ∂t(ρU) + div(ρU ⊗ U) + ∇p = ρF with p = pa + ρ − ρ0 c2 with c sound speed Method :

1 Write Euler equations in curvilinear coordinates. 2 Write equations in non-dimensional form using the small parameter ǫ = H/L

and takes ǫ = 0.

3 Section averaging ρU ≈ ρU and ρU 2 ≈ ρU U. 4 Introduce Ach(t, x) : equivalent wet area, Qch(t, x) discharge given by :

Ach(t, x) = ρ ρ0 S(x), Qch(t, x) = Ach(t, x)u(t, x) u(t, x) = 1 S(x)

  • Ω(x)

U(t, x) dydz

  • M. Ersoy (BCAM)

PFS-model and VFRoe solver LMB, Besan¸ con, the 10 February 2011 14 / 40

slide-25
SLIDE 25

Derivation of the pressurized model

3D isentropic compressible equations

∂tρ + div(ρU) = 0 ∂t(ρU) + div(ρU ⊗ U) + ∇p = ρF with p = pa + ρ − ρ0 c2 with c sound speed Method :

1 Write Euler equations in curvilinear coordinates. 2 Write equations in non-dimensional form using the small parameter ǫ = H/L

and takes ǫ = 0.

3 Section averaging ρU ≈ ρU and ρU 2 ≈ ρU U. 4 Introduce Ach(t, x) : equivalent wet area, Qch(t, x) discharge given by :

Ach(t, x) = ρ ρ0 S(x), Qch(t, x) = Ach(t, x)u(t, x) u(t, x) = 1 S(x)

  • Ω(x)

U(t, x) dydz

  • M. Ersoy (BCAM)

PFS-model and VFRoe solver LMB, Besan¸ con, the 10 February 2011 14 / 40

slide-26
SLIDE 26

The pressurized model

               ∂tAch + ∂xQch = 0, ∂tQch + ∂x Q2

ch

Ach + pch(x, Ach)

  • =

−gAch d Z dx + Prch(x, Ach) − G(x, Ach) with pch = c2(Ach − S) : acoustic type pressure law Prch = c2 Ach S − 1 d S dx : pressure source term G = gAchz d dx cos θ : curvature source term

  • M. Ersoy (BCAM)

PFS-model and VFRoe solver LMB, Besan¸ con, the 10 February 2011 15 / 40

slide-27
SLIDE 27

The pressurized model

               ∂tAch + ∂xQch = 0, ∂tQch + ∂x Q2

ch

Ach + pch(x, Ach)

  • =

−gAch d Z dx + Prch(x, Ach) − G(x, Ach) − gK(x, S)Qch|Qch| Ach

  • friction added after the derivation

with pch = c2(Ach − S) : acoustic type pressure law Prch = c2 Ach S − 1 d S dx : pressure source term G = gAchz d dx cos θ : curvature source term K = 1 K2

sRh(S)4/3

: Manning-Strickler law

  • M. Ersoy (BCAM)

PFS-model and VFRoe solver LMB, Besan¸ con, the 10 February 2011 15 / 40

slide-28
SLIDE 28

Outline

Outline

1 Unsteady mixed flows : PFS equations (Pressurized

and Free Surface)

Previous works Formal derivation of the free surface and pressurized model A coupling : the PFS-model

2 Finite Volume discretization

Discretization of the space domain Explicit first order VFRoe scheme

  • 1. The Case of a non transition point
  • 2. The Case of a transition point
  • 3. Update of the cell state
  • 4. Approximation of the convection matrix

3 Numerical experiments 4 Conclusion and perspectives

  • M. Ersoy (BCAM)

PFS-model and VFRoe solver LMB, Besan¸ con, the 10 February 2011 16 / 40

slide-29
SLIDE 29

the PFS model

Models are formally close . . .

               ∂tAsl + ∂xQsl = 0, ∂tQsl + ∂x Q2

sl

Asl + psl (x, Asl)

  • =

−g Asl d Z dx + Prsl (x, Asl) −G(x, Asl ) −gK(x, Asl )Qsl|Qsl| Asl                ∂tAch + ∂xQch = 0, ∂tQch + ∂x Q2

ch

Ach + pch (x, Ach)

  • =

−g Ach d Z dx + Prch (x, Ach) −G(x, Ach ) −gK(x, S )Qch|Qch| Ach

  • M. Ersoy (BCAM)

PFS-model and VFRoe solver LMB, Besan¸ con, the 10 February 2011 17 / 40

slide-30
SLIDE 30

the PFS model

Models are formally close . . .

               ∂tAsl + ∂xQsl = 0, ∂tQsl + ∂x Q2

sl

Asl + psl (x, Asl)

  • =

−g Asl d Z dx + Prsl (x, Asl) −G(x, Asl ) −gK(x, Asl )Qsl|Qsl| Asl                ∂tAch + ∂xQch = 0, ∂tQch + ∂x Q2

ch

Ach + pch (x, Ach)

  • =

−g Ach d Z dx + Prch (x, Ach) −G(x, Ach ) −gK(x, S )Qch|Qch| Ach

Continuity criterion

  • M. Ersoy (BCAM)

PFS-model and VFRoe solver LMB, Besan¸ con, the 10 February 2011 17 / 40

slide-31
SLIDE 31

the PFS model

Models are formally close . . .

               ∂tAsl + ∂xQsl = 0, ∂tQsl + ∂x Q2

sl

Asl + psl (x, Asl)

  • =

−g Asl d Z dx + Prsl (x, Asl) −G(x, Asl ) −gK(x, Asl )Qsl|Qsl| Asl                ∂tAch + ∂xQch = 0, ∂tQch + ∂x Q2

ch

Ach + pch (x, Ach)

  • =

−g Ach d Z dx + Prch (x, Ach) −G(x, Ach ) −gK(x, S )Qch|Qch| Ach

« mixed »condition

  • M. Ersoy (BCAM)

PFS-model and VFRoe solver LMB, Besan¸ con, the 10 February 2011 17 / 40

slide-32
SLIDE 32

the PFS model

Models are formally close . . .

               ∂tAsl + ∂xQsl = 0, ∂tQsl + ∂x Q2

sl

Asl + psl (x, Asl)

  • =

−g Asl d Z dx + Prsl (x, Asl) −G(x, Asl ) −gK(x, Asl )Qsl|Qsl| Asl                ∂tAch + ∂xQch = 0, ∂tQch + ∂x Q2

ch

Ach + pch (x, Ach)

  • =

−g Ach d Z dx + Prch (x, Ach) −G(x, Ach ) −gK(x, S )Qch|Qch| Ach

To be coupled

  • M. Ersoy (BCAM)

PFS-model and VFRoe solver LMB, Besan¸ con, the 10 February 2011 17 / 40

slide-33
SLIDE 33

The PFS model

The « mixed »variable

We introduce a state indicator E = 1 if the flow is pressurized (CH), if the flow is free surface (SL)

  • M. Ersoy (BCAM)

PFS-model and VFRoe solver LMB, Besan¸ con, the 10 February 2011 18 / 40

slide-34
SLIDE 34

The PFS model

The « mixed »variable

We introduce a state indicator E = 1 if the flow is pressurized (CH), if the flow is free surface (SL) and the physical section of water S by : S = S(Asl, E) = S if E = 1, Asl if E = 0.

  • M. Ersoy (BCAM)

PFS-model and VFRoe solver LMB, Besan¸ con, the 10 February 2011 18 / 40

slide-35
SLIDE 35

The PFS model

The « mixed »variable

We introduce a state indicator E = 1 if the flow is pressurized (CH), if the flow is free surface (SL) and the physical section of water S by : S = S(Asl, E) = S if E = 1, Asl if E = 0. We set A = ¯ ρ ρ0 S = S(Asl, 0) = Asl if SL ¯ ρ ρ0 S(Asl, 1) = Ach if CH : the « mixed »variable Q = Au : the discharge

  • M. Ersoy (BCAM)

PFS-model and VFRoe solver LMB, Besan¸ con, the 10 February 2011 18 / 40

slide-36
SLIDE 36

The PFS model

The « mixed »variable

We introduce a state indicator E = 1 if the flow is pressurized (CH), if the flow is free surface (SL) and the physical section of water S by : S = S(Asl, E) = S if E = 1, Asl if E = 0. We set A = ¯ ρ ρ0 S = S(Asl, 0) = Asl if SL ¯ ρ ρ0 S(Asl, 1) = Ach if CH : the « mixed »variable Q = Au : the discharge

Continuity of S at transition point

  • M. Ersoy (BCAM)

PFS-model and VFRoe solver LMB, Besan¸ con, the 10 February 2011 18 / 40

slide-37
SLIDE 37

The PFS model

Construction of the « mixed »pressure

Continuity of S = ⇒ continuity of p at transition point − → p(x, A, E) = c2(A − S) + gI1(x, S) cos θ

  • M. Ersoy (BCAM)

PFS-model and VFRoe solver LMB, Besan¸ con, the 10 February 2011 19 / 40

slide-38
SLIDE 38

The PFS model

Construction of the « mixed »pressure

Continuity of S = ⇒ continuity of p at transition point − → p(x, A, E) = c2(A − S) + gI1(x, S) cos θ Similar construction for the pressure source term : Pr(x, A, E) = c2 A S − 1 d S dx + gI2(x, S) cos θ

  • M. Ersoy (BCAM)

PFS-model and VFRoe solver LMB, Besan¸ con, the 10 February 2011 19 / 40

slide-39
SLIDE 39

The PFS model

                             ∂t(A) + ∂x(Q) = 0 ∂t(Q) + ∂x Q2 A + p(x, A, E)

  • = −g A d

dxZ(x) +Pr(x, A, E) −G(x, A, E) −g K(x, S) Q|Q| A

  • C. Bourdarias, M. Ersoy and S. Gerbi

A model for unsteady mixed flows in non uniform closed water pipes and a well-balanced finite volume scheme.

  • Int. J. On Finite Volumes, 6(2) :1–47, 2009.
  • M. Ersoy (BCAM)

PFS-model and VFRoe solver LMB, Besan¸ con, the 10 February 2011 20 / 40

slide-40
SLIDE 40

The PFS model

Mathematical properties

The PFS system is strictly hyperbolic for A(t, x) > 0. For regular solutions, the mean speed u = Q/A verifies ∂tu + ∂x u2 2 + c2 ln(A/S) + g H(S) cos θ + g Z

  • = −g K(x, S) u |u|

and for u = 0, we have : c2 ln(A/S) + g H(S) cos θ + g Z = cte where H(S) is the physical water height. There exists a mathematical entropy E(A, Q, S) = Q2 2A + c2A ln(A/S) + c2S + gz(x, S) cos θ + gAZ which satisfies ∂tE + ∂x (E u + p(x, A, E) u) = −g A K(x, S) u2 |u| 0

  • M. Ersoy (BCAM)

PFS-model and VFRoe solver LMB, Besan¸ con, the 10 February 2011 21 / 40

slide-41
SLIDE 41

Outline

Outline

1 Unsteady mixed flows : PFS equations (Pressurized

and Free Surface)

Previous works Formal derivation of the free surface and pressurized model A coupling : the PFS-model

2 Finite Volume discretization

Discretization of the space domain Explicit first order VFRoe scheme

  • 1. The Case of a non transition point
  • 2. The Case of a transition point
  • 3. Update of the cell state
  • 4. Approximation of the convection matrix

3 Numerical experiments 4 Conclusion and perspectives

  • M. Ersoy (BCAM)

PFS-model and VFRoe solver LMB, Besan¸ con, the 10 February 2011 22 / 40

slide-42
SLIDE 42

Outline

Outline

1 Unsteady mixed flows : PFS equations (Pressurized

and Free Surface)

Previous works Formal derivation of the free surface and pressurized model A coupling : the PFS-model

2 Finite Volume discretization

Discretization of the space domain Explicit first order VFRoe scheme

  • 1. The Case of a non transition point
  • 2. The Case of a transition point
  • 3. Update of the cell state
  • 4. Approximation of the convection matrix

3 Numerical experiments 4 Conclusion and perspectives

  • M. Ersoy (BCAM)

PFS-model and VFRoe solver LMB, Besan¸ con, the 10 February 2011 23 / 40

slide-43
SLIDE 43

The mesh and the unknowns

Geometric terms and unknowns are piecewise constant approximations on the cell mi at time tn : Geometric terms

◮ Zi, Si, cos θi

unknowns

◮ (An

i , Qn i ), un i = Qn i

An

i

Notation :“unknown”vector

◮ Wn

i = (Zi, cos θi, Si, An i , Qn i )t

  • M. Ersoy (BCAM)

PFS-model and VFRoe solver LMB, Besan¸ con, the 10 February 2011 24 / 40

slide-44
SLIDE 44

Outline

Outline

1 Unsteady mixed flows : PFS equations (Pressurized

and Free Surface)

Previous works Formal derivation of the free surface and pressurized model A coupling : the PFS-model

2 Finite Volume discretization

Discretization of the space domain Explicit first order VFRoe scheme

  • 1. The Case of a non transition point
  • 2. The Case of a transition point
  • 3. Update of the cell state
  • 4. Approximation of the convection matrix

3 Numerical experiments 4 Conclusion and perspectives

  • M. Ersoy (BCAM)

PFS-model and VFRoe solver LMB, Besan¸ con, the 10 February 2011 25 / 40

slide-45
SLIDE 45

Non-conservative formulation

Adding the equations ∂tZ = 0, ∂t cos θ = 0 and ∂tS = 0, the PFS-model under a non conservative form reads : ∂tW + D(W)∂xW = TS(W) (1) where W = (Z, cos θ, S, A, Q)t TS(W) =

  • 0, 0, 0, 0, −g K(x, S) Q|Q|

A

  • D(W) =

      1 gA gAH(S) Ψ(W) c2(W) − u2 2u       where Ψ(W) = gS∂SH(S) cos θ − c2(W)A S and c(W) =      c for pressurised flow

  • g

A T(A) cos θ for free surface flow

  • M. Ersoy (BCAM)

PFS-model and VFRoe solver LMB, Besan¸ con, the 10 February 2011 26 / 40

slide-46
SLIDE 46

The Finite Volume scheme

Integrating conservative PFS-System over ]xi−1/2, xi+ 1

2 [×[tn, tn+1[, we can write

a Finite Volume scheme as follows : Wn+1

i

= Wn

i − ∆tn

hi

  • F(W∗

i+1/2(0−, Wn i , Wn i+1)) − F(W∗ i−1/2(0+, Wn i−1, Wn i ))

  • +TS(Wn

i )

(1) W∗

i+1/2(ξ = x/t, Wi, Wi+1) is the exact or an approximate solution to the

Riemann problem at interface xi+1/2.

  • M. Ersoy (BCAM)

PFS-model and VFRoe solver LMB, Besan¸ con, the 10 February 2011 26 / 40

slide-47
SLIDE 47

Interface quantities AM, QM, AP, QP depend on two types of interfaces

W ∗(0+, Wi, Wi+1) = (Zi+1, cos θi+1, Si+1, AP, QP)t and W ∗(0−, Wi, Wi+1) = (Zi+1, cos θi+1, Si+1, AM, QM)t depend on two types of interfaces : a non transition point : the flow on both sides of the interface is of the same type a transition point : the flow changes of type through the interface

  • M. Ersoy (BCAM)

PFS-model and VFRoe solver LMB, Besan¸ con, the 10 February 2011 26 / 40

slide-48
SLIDE 48

Outline

Outline

1 Unsteady mixed flows : PFS equations (Pressurized

and Free Surface)

Previous works Formal derivation of the free surface and pressurized model A coupling : the PFS-model

2 Finite Volume discretization

Discretization of the space domain Explicit first order VFRoe scheme

  • 1. The Case of a non transition point
  • 2. The Case of a transition point
  • 3. Update of the cell state
  • 4. Approximation of the convection matrix

3 Numerical experiments 4 Conclusion and perspectives

  • M. Ersoy (BCAM)

PFS-model and VFRoe solver LMB, Besan¸ con, the 10 February 2011 27 / 40

slide-49
SLIDE 49

The linearized Riemann problem

approximating the convection matrix D(W) by D, to compute (AM, QM), (AP, QP), we solve the linearized Riemann problem :    ∂tW + D ∂xW = W = Wl = (Zl, cos θl, Sl, Al, Ql)t if x < 0 Wr = (Zr, cos θr, Sr, Ar, Qr)t if x > 0 (1) with (Wl, Wr) = (Wi, Wi+1) and D = D(Wl, Wr) = D( W) where W is some approximate state of the left Wl and the right Wr state.

  • M. Ersoy (BCAM)

PFS-model and VFRoe solver LMB, Besan¸ con, the 10 February 2011 28 / 40

slide-50
SLIDE 50

The convection matrix

The eigenvalues of D are λi = 0, i = 1, 2, 3, λ4 = u − c( W), λ5 = u + c( W) where c(W) =      c for pressurised flow

  • g

A T(A) cos θ for free surface flow

AM QM AM QM AM QM AP QP AP QP AP QP

W W W W W

l r l r l r

(1),(2),(3) (1),(2),(3) (1),(2),(3) (4) (4) (4) (5) (5) (5) W

u < − c ~ − c < u < c u > c ~ ~ ~ ~ ~ ~

  • M. Ersoy (BCAM)

PFS-model and VFRoe solver LMB, Besan¸ con, the 10 February 2011 28 / 40

slide-51
SLIDE 51

AM, QM, AP, QP are given by

We obtain, for instance in the sub-critical case (when −c( W) < u < c( W)), we have : AM = Al + g A 2 c( W) (c( W) − u) ψr

l +

u + c( W) 2 c( W) (Ar − Al) − 1 2 c( W) (Qr − Ql) QM = QP = Ql − g A 2 c( W) ψr

l +

u2 − c( W)2 2 c( W) (Ar − Al) − u − c( W) 2 c( W) (Qr − Ql) AP = AM + g A

  • u2 − c(

W)2 ψr

l

where ψr

l is the upwinded source term

Zr − Zl + H( S)(cos θr − cos θl) + Ψ( W)(Sr − Sl).

  • M. Ersoy (BCAM)

PFS-model and VFRoe solver LMB, Besan¸ con, the 10 February 2011 28 / 40

slide-52
SLIDE 52

Outline

Outline

1 Unsteady mixed flows : PFS equations (Pressurized

and Free Surface)

Previous works Formal derivation of the free surface and pressurized model A coupling : the PFS-model

2 Finite Volume discretization

Discretization of the space domain Explicit first order VFRoe scheme

  • 1. The Case of a non transition point
  • 2. The Case of a transition point
  • 3. Update of the cell state
  • 4. Approximation of the convection matrix

3 Numerical experiments 4 Conclusion and perspectives

  • M. Ersoy (BCAM)

PFS-model and VFRoe solver LMB, Besan¸ con, the 10 February 2011 29 / 40

slide-53
SLIDE 53

Two Riemann problems

Assumption the propagation of the interface (pressurized-free surface or free surface-pressurized) has a constant speed w during a time step. Consequently the half line x = w t is the discontinuity line of D(Wl, Wr). Setting w = Q+ − Q− A+ − A− with U− = (A−, Q−) and U+ = (A+, Q+) the (unknown) states resp. on the left and on the right hand side of the line x = w t

click .

Remark Both states Ul and U− (resp. Ur and U+) correspond to the same type of flow Thus it makes sense to define the averaged matrices in each zone as follows :

◮ for x < w t, we set

Dl = D(Wl, Wr) = D( Wl) for some approximation Wl which connects the state Wl and W−.

◮ for x > w t, we set

Dr = D(Wl, Wr) = D( Wr) for some approximation Wl which connects the state W+ and Wr.

  • M. Ersoy (BCAM)

PFS-model and VFRoe solver LMB, Besan¸ con, the 10 February 2011 30 / 40

slide-54
SLIDE 54

four cases

Then we formally solve two Riemann problems and use the Rankine-Hugoniot jump conditions through the line x = w t which writes : Q+ − Q− = w (A+ − A−) (1) F5(A+, Q+) − F5(A−, Q−) = w (Q+ − Q−) (2) with F5(A, Q) = Q2 A + p(x, A). According to (U−, UM) and (U+, UP ) (unknowns) at the interface xi+1/2 and the sign of the speed w, we have to deal with four cases : pressure state propagating downstream

click ,

  • M. Ersoy (BCAM)

PFS-model and VFRoe solver LMB, Besan¸ con, the 10 February 2011 31 / 40

slide-55
SLIDE 55

four cases

Then we formally solve two Riemann problems and use the Rankine-Hugoniot jump conditions through the line x = w t which writes : Q+ − Q− = w (A+ − A−) (1) F5(A+, Q+) − F5(A−, Q−) = w (Q+ − Q−) (2) with F5(A, Q) = Q2 A + p(x, A). According to (U−, UM) and (U+, UP ) (unknowns) at the interface xi+1/2 and the sign of the speed w, we have to deal with four cases : pressure state propagating downstream

click ,

pressure state propagating upstream, free surface state propagating downstream, free surface state propagating upstream.

  • M. Ersoy (BCAM)

PFS-model and VFRoe solver LMB, Besan¸ con, the 10 February 2011 31 / 40

slide-56
SLIDE 56

Outline

Outline

1 Unsteady mixed flows : PFS equations (Pressurized

and Free Surface)

Previous works Formal derivation of the free surface and pressurized model A coupling : the PFS-model

2 Finite Volume discretization

Discretization of the space domain Explicit first order VFRoe scheme

  • 1. The Case of a non transition point
  • 2. The Case of a transition point
  • 3. Update of the cell state
  • 4. Approximation of the convection matrix

3 Numerical experiments 4 Conclusion and perspectives

  • M. Ersoy (BCAM)

PFS-model and VFRoe solver LMB, Besan¸ con, the 10 February 2011 32 / 40

slide-57
SLIDE 57

State update

Given n, ∀i, An

i and En i are known. Then

if En

i = 0 then

if An+1

i

< Si then En+1

i

= 0 else En+1

i

= 1

if En

i = 1 then

if An+1

i

Si then En+1

i

= 1 else En+1

i

= En

i−1En i+1

  • M. Ersoy (BCAM)

PFS-model and VFRoe solver LMB, Besan¸ con, the 10 February 2011 33 / 40

slide-58
SLIDE 58

Outline

Outline

1 Unsteady mixed flows : PFS equations (Pressurized

and Free Surface)

Previous works Formal derivation of the free surface and pressurized model A coupling : the PFS-model

2 Finite Volume discretization

Discretization of the space domain Explicit first order VFRoe scheme

  • 1. The Case of a non transition point
  • 2. The Case of a transition point
  • 3. Update of the cell state
  • 4. Approximation of the convection matrix

3 Numerical experiments 4 Conclusion and perspectives

  • M. Ersoy (BCAM)

PFS-model and VFRoe solver LMB, Besan¸ con, the 10 February 2011 34 / 40

slide-59
SLIDE 59

The classical choice

The classical approximation D( W) of the Roe matrix DRoe(Wl, Wr) = 1 D(Wr + (1 − s)(Wl − Wr)) ds defined by

  • D = D(

W) = D Wl + Wr 2

  • preserve the still water steady state only for

constant section pipe and Z = 0.

  • M. Ersoy (BCAM)

PFS-model and VFRoe solver LMB, Besan¸ con, the 10 February 2011 35 / 40

slide-60
SLIDE 60

Construction of an exactly well-balanced scheme

Let us start with the consideration : the still water steady state is perfectly maintained : there exists n such that for every i, if Qn

i = 0 and ∀i,

A1 : c2 ln An

i+1

Si+1

  • + gH(Sn

i+1) cos θ + gZi+1 =

c2 ln An

i

Si

  • + gH(Sn

i ) cos θ + gZi,

A2 : AM n

i+1/2 = AP n i−1/2,

A3 : Qn

i+1/2 = Qn i−1/2,

then, for all l > n the conditions A1, A2 and A3 holds.

  • M. Ersoy (BCAM)

PFS-model and VFRoe solver LMB, Besan¸ con, the 10 February 2011 35 / 40

slide-61
SLIDE 61

Defining

( An

i−1/2,

An

i+1/2) as the solution of the non-linear system :

               = ∆An

i+1/2 + g

2 An

i+1/2ψi+1 i

  • c2

i+1/2

+

  • An

i−1/2ψi i−1

  • c2

i−1/2

  • =

g 2 An

i−1/2 ψi i−1

  • ci−1/2

  • An

i+1/2 ψi+1 i

  • ci+1/2
  • +

∆An

i+1/2

2

  • ci−1/2 −

ci+1/2

  • (3)

the numerical scheme is exactly well-balanced.

  • M. Ersoy (BCAM)

PFS-model and VFRoe solver LMB, Besan¸ con, the 10 February 2011 35 / 40

slide-62
SLIDE 62

For small ∆x, we show that

  • An

i+1/2 ≈ An i + An i+1

2

  • M. Ersoy (BCAM)

PFS-model and VFRoe solver LMB, Besan¸ con, the 10 February 2011 35 / 40

slide-63
SLIDE 63

Outline

Outline

1 Unsteady mixed flows : PFS equations (Pressurized

and Free Surface)

Previous works Formal derivation of the free surface and pressurized model A coupling : the PFS-model

2 Finite Volume discretization

Discretization of the space domain Explicit first order VFRoe scheme

  • 1. The Case of a non transition point
  • 2. The Case of a transition point
  • 3. Update of the cell state
  • 4. Approximation of the convection matrix

3 Numerical experiments 4 Conclusion and perspectives

  • M. Ersoy (BCAM)

PFS-model and VFRoe solver LMB, Besan¸ con, the 10 February 2011 36 / 40

slide-64
SLIDE 64

Well-balanced scheme and the averaged approximation for P

  • 2.5e-09
  • 2e-09
  • 1.5e-09
  • 1e-09
  • 5e-10
5e-10 1e-09 1.5e-09 2e-09 2.5e-09 10 20 30 40 50 Discharge (m^3/s) Time (s) Discharge at x = 40 m Classical approximation EWBS

Well-balanced scheme and the averaged approximation for FS

  • 0.00000500
  • 0.00000400
  • 0.00000300
  • 0.00000200
  • 0.00000100
0.00000000 0.00000100 0.00000200 0.00000300 10 20 30 40 50 Discharge (m^3/s) Time (s) Discharge at x = 40 m Classical approximation EWBS

Depression for a contracting pipe

0.9 0.95 1 1.05 1.1 1.15 1.2 1.25 1.3 20 40 60 80 100 A/S x (m) A/S = 1 t = 16.870 t = 17.189 t = 17.508 t = 17.828 t = 18.149 t = 18.471 t = 18.794 t = 19.117 t = 19.442 t = 19.767 t = 20.093 t = 20.419 t = 20.747 t = 21.074 t = 21.403 t = 21.732

Depression for an uniform pipe

0.9 0.95 1 1.05 1.1 1.15 1.2 20 40 60 80 100 A/S x (m) A/S=1 t = 23.495 t = 23.569 t = 23.642 t = 23.715 t = 23.788 t = 23.860 t = 23.932 t = 24.003 t = 24.075 t = 24.146 t = 24.216 t = 24.287 t = 24.357 t = 24.427 t = 24.497 t = 24.567

Depression for an expanding pipe

0.96 0.98 1 1.02 1.04 1.06 1.08 1.1 20 40 60 80 100 A/S x (m) A/S=1 t = 28.172 t = 28.227 t = 28.283 t = 28.339 t = 28.395 t = 28.452 t = 28.510 t = 28.568 t = 28.626 t = 28.685 t = 28.745
  • M. Ersoy (BCAM)

PFS-model and VFRoe solver LMB, Besan¸ con, the 10 February 2011 37 / 40

slide-65
SLIDE 65

Outline

Outline

1 Unsteady mixed flows : PFS equations (Pressurized

and Free Surface)

Previous works Formal derivation of the free surface and pressurized model A coupling : the PFS-model

2 Finite Volume discretization

Discretization of the space domain Explicit first order VFRoe scheme

  • 1. The Case of a non transition point
  • 2. The Case of a transition point
  • 3. Update of the cell state
  • 4. Approximation of the convection matrix

3 Numerical experiments 4 Conclusion and perspectives

  • M. Ersoy (BCAM)

PFS-model and VFRoe solver LMB, Besan¸ con, the 10 February 2011 38 / 40

slide-66
SLIDE 66

Conclusion

Conservative and simple formulation (easy implementation even if source terms are complex) Well-balanced numerical scheme Very good agreement for uniform case Compressibility of water for pressurized flows

Water hammer Depression

  • M. Ersoy (BCAM)

PFS-model and VFRoe solver LMB, Besan¸ con, the 10 February 2011 39 / 40

slide-67
SLIDE 67

Conclusion and perspectives

Conservative and simple formulation (easy implementation even if source terms are complex) Well-balanced numerical scheme Very good agreement for uniform case Compressibility of water for pressurized flows

Water hammer Depression

Perspectives : cavitation condensation evaporation

  • M. Ersoy (BCAM)

PFS-model and VFRoe solver LMB, Besan¸ con, the 10 February 2011 39 / 40

slide-68
SLIDE 68

Thank you

Thank you

for your

for your

attention

attention

  • M. Ersoy (BCAM)

PFS-model and VFRoe solver LMB, Besan¸ con, the 10 February 2011 40 / 40