Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work A New Class of Well-Balanced Finite Volume schemes for Conservation laws with source terms Siddhartha Mishra Centre of Mathematics for Applications (CMA), University of Oslo, Norway Siddhartha Mishra A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work Joint Work with: I Kenneth Hvistendahl Karlsen (CMA, Oslo). I Nils Henrik Risebro (CMA, Oslo). Siddhartha Mishra A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work Siddhartha Mishra A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work Basic Equations Ut + (f (U))x + (g (U))y + (h(U))z = S(x, U) I System of Conservation laws in multi-D. I Together with source terms. I Source can be spatially dependent (maybe singular). I Also termed Balance laws. Siddhartha Mishra A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work Flow on a non-trivial topography h h b Non−Trivial Smooth Bottom Topography Siddhartha Mishra b Discontinuous Bottom Topography A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work An Example • Shallow water equations with Non-trivial Bottom Topography. ht + (hu)x + (hv )y (hu)t + (hu 2 + 12 gh2 )x + (huv )y (hv )t + (huv )x + (hv 2 + 12 gh2 )y = 0 = −ghbx = −ghby I h is height of the free surface. I (u, v ) is the velocity vector. I g - gravity constant. I b Topography function (can be discontinuous). Siddhartha Mishra A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work The Model Equation • Single conservation law in 1-d. ut + f (u)x = A(x, u) I Unknown u, flux f and source A. I Source can even be singular. (A can be a measure) Siddhartha Mishra A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work Special Cases • Autonomous source ut + f (u)x = g (u) • Scalar “Shallow Water” equations ut + (f (u))x = z 0 (x)b(u) I z is the topography function (possibly discontinuous) • Singular Sources ut + (f (u))x = z 0 (x) • z Heaviside funtion ⇒ RHS is a measure. Siddhartha Mishra A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work Weak Solutions • Well Defined when A(x, u) ∈ L∞ . • u ∈ L∞ (R × R+ ) ∩ L1loc is a weak solution if for all test functions ϕ, Z Z Z uϕt + f (u)ϕx + A(x, u)ϕ dxdt + u(x, 0)ϕ(x, 0) = 0 (1) R+ R R • Special attention when A ∈ / L∞ . • Make sense of the non-conservative product z 0 (x)b(u) Siddhartha Mishra A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work Entropy Solutions • Well Defined when A(x, u) ∈ L∞ . • u ∈ L∞ (R × R+ ) ∩ L1loc is a entropy solution if for all test functions ϕ ≥ 0, Z Z Z 0 S(u)ϕt +Q(u)ϕx +S (u)A(x, u)ϕ dxdt+ u(x, 0)ϕ(x, 0) ≥ 0 R+ R R • For any entropy-entropy flux pair (S, Q). • Entropy solutions exist and are unique when A ∈ L∞ . • No general theory in the singular case except when A(x, u) = z 0 (x) Siddhartha Mishra A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work A Naive Numerical Scheme ut + f (u)x = z 0 (x)b(u) Siddhartha Mishra A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work A Naive Numerical Scheme ut + f (u)x = z 0 (x)b(u) I Explicit Euler in Time. Siddhartha Mishra A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work A Naive Numerical Scheme ut + f (u)x = z 0 (x)b(u) I Explicit Euler in Time. I Godunov type numerical fluxes for the flux. Siddhartha Mishra A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work A Naive Numerical Scheme ut + f (u)x = z 0 (x)b(u) I Explicit Euler in Time. I Godunov type numerical fluxes for the flux. I Central differences for the source. Siddhartha Mishra A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work A Numerical Experiment ut + f (u)x = z 0 (x)b(u) • With 1 2 = 2u cos(πx) if 4.5 < x < 5.5 −z(x) = 0 Otherwise u(t, 0) = 2 f (u) b(u) = u u(0, x) = 0 • Explicit steady state is given by u(x) = 2 + z(x) Siddhartha Mishra A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work At the steady state 3.5 3 2.5 2 1.5 1 Exact:−−−−−−−−−−−−− CS :− − − − − − − BT: + + + + + + 0.5 0 −0.5 −1 0 2 4 Siddhartha Mishra 6 8 10 A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work Key Numerical Issues I Resolution of steady states. Siddhartha Mishra A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work Key Numerical Issues I Resolution of steady states. I At a steady state ⇔ Flux-Source balance. Siddhartha Mishra A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work Key Numerical Issues I Resolution of steady states. I At a steady state ⇔ Flux-Source balance. I f (u)x ≈ A(x, u) Siddhartha Mishra A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work Key Numerical Issues I Resolution of steady states. I At a steady state ⇔ Flux-Source balance. I f (u)x ≈ A(x, u) I Numerical schemes have to preserve Flux-Source balance. Siddhartha Mishra A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work Key Numerical Issues I Resolution of steady states. I At a steady state ⇔ Flux-Source balance. I f (u)x ≈ A(x, u) I Numerical schemes have to preserve Flux-Source balance. I Centered Source/Operator splitting doesn’t respect it. Siddhartha Mishra A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work Key Numerical Issues I Resolution of steady states. I At a steady state ⇔ Flux-Source balance. I f (u)x ≈ A(x, u) I Numerical schemes have to preserve Flux-Source balance. I Centered Source/Operator splitting doesn’t respect it. I Search for better schemes Siddhartha Mishra A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work I Greenberg, Leroux. I Greenberg, Leroux, Baraille and Noussair. I Gosse, Leroux. I Botchorischvili, Perthame and Vasseur. (BPV) I Bermudez, Vasquez I Perthame, Bouchut, Bristeau, Klien, Audusse. I Russo, Noelle, Kurganov, Levy and many others. Siddhartha Mishra A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work WBS (condensed) • Consider Scalar Shallow water equations, ut + f (u)x = z 0 (x)b(u) • The steady state is formally, ⇒ ⇒ ⇒ f (u)x f 0 (u)ux f 0 (u) b(u) ux D(u)x D(u) = = = = = z 0 (x)b(u) z 0 (x)b(u) z 0 (x) z 0 (x) R u f 0 (s) b(s) ds • Steady State evaluated from D − z = Constant Siddhartha Mishra A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work WBS (Condensed) I At each time step, the cell values are projected unto “local” steady states i.e Siddhartha Mishra A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work WBS (Condensed) I I At each time step, the cell values are projected unto “local” steady states i.e At nth time step let vjn be the cell-averages and zj be averages of the topography, then define “local” steady states solving D(vjn −) − zj D(vjn +) − zj Siddhartha Mishra n )−z = D(vj−1 j−1 n )−z = D(vj+1 j+1 A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work WBS (Condensed) I I At each time step, the cell values are projected unto “local” steady states i.e At nth time step let vjn be the cell-averages and zj be averages of the topography, then define “local” steady states solving D(vjn −) − zj D(vjn +) − zj I n )−z = D(vj−1 j−1 n )−z = D(vj+1 j+1 Use the local steady states to define a Godonov type scheme with update Siddhartha Mishra A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work WBS (Condensed) I I At each time step, the cell values are projected unto “local” steady states i.e At nth time step let vjn be the cell-averages and zj be averages of the topography, then define “local” steady states solving D(vjn −) − zj D(vjn +) − zj I I n )−z = D(vj−1 j−1 n )−z = D(vj+1 j+1 Use the local steady states to define a Godonov type scheme with update ∆t (F (vjn , vjn +) − F (vjn −, vjn )) ∆x • with F being Standard (Godunov, Enquist-Osher) flux corresponding to f vjn+1 = vjn − Siddhartha Mishra A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work WBS (Advantages) Siddhartha Mishra A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work WBS (Advantages) I Discrete steady states are preserved exactly. Siddhartha Mishra A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work WBS (Advantages) I Discrete steady states are preserved exactly. I Shown to Converge to entropy solutions (via Kinetic formulation). Siddhartha Mishra A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work WBS (Advantages) I Discrete steady states are preserved exactly. I Shown to Converge to entropy solutions (via Kinetic formulation). I Basis for WBS for systems. Siddhartha Mishra A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work WBS (Problems) Siddhartha Mishra A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work WBS (Problems) I Expensive: 2 Algebraic equations to be solved for each mesh point. Siddhartha Mishra A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work WBS (Problems) I Expensive: 2 Algebraic equations to be solved for each mesh point. I Complicated: Steady state equations may have no solutions/multiple solutions. Siddhartha Mishra A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work WBS (Problems) I Expensive: 2 Algebraic equations to be solved for each mesh point. I Complicated: Steady state equations may have no solutions/multiple solutions. I Specialized: Difficult to extend when source is not in product form. Siddhartha Mishra A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work WBS (Problems) I Expensive: 2 Algebraic equations to be solved for each mesh point. I Complicated: Steady state equations may have no solutions/multiple solutions. I Specialized: Difficult to extend when source is not in product form. I Non-entropic: In some cases with discontinuous z. Siddhartha Mishra A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work WBS (Problems) I Expensive: 2 Algebraic equations to be solved for each mesh point. I Complicated: Steady state equations may have no solutions/multiple solutions. I Specialized: Difficult to extend when source is not in product form. I Non-entropic: In some cases with discontinuous z. I Possible loss of accuracy away from steady states. Siddhartha Mishra A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work WBS (Problems) I Expensive: 2 Algebraic equations to be solved for each mesh point. I Complicated: Steady state equations may have no solutions/multiple solutions. I Specialized: Difficult to extend when source is not in product form. I Non-entropic: In some cases with discontinuous z. I Possible loss of accuracy away from steady states. I Subtle deficiences (see sequel) Siddhartha Mishra A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work Along the lines of I Greenberg, Leroux, Baraille and Noussair. (Singular Sources) I Noussair. I LeVeque. I Bale, LeVeque, Mitran, Rossmanith (Flux - Differencing) I Adimurthi, Gowda, Mishra (Singular Sources) Siddhartha Mishra A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work 1−D Finite Volume Grid tn + 2 n+ 1 U j n Uj n U j −1 xj − 3/2 xj − 1/2 Siddhartha Mishra t n Uj +1 F (u j , u j + 1) x j + 1/2 n+1 tn x j + 3/2 A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work The Scheme: Design I At time level n, let ujn be the cell averages, • Step 1: Freeze the source at t n and define the piecewise constant X u n (x) = ujn 1{Ij } (x) j with Ij being the jth cell. Formally (“local” in time) we have the equation ut + (f (u))x = A(x, u n (x)) Siddhartha Mishra A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work The Scheme: Design I At time level n, let ujn be the cell averages, • Step 1: Freeze the source at t n and define the piecewise constant X u n (x) = ujn 1{Ij } (x) j with Ij being the jth cell. Formally (“local” in time) we have the equation ut + (f (u))x = A(x, u n (x)) I Primitive Reconstruction: Define the function Z x n B̃ (x) = A(y , u n (y ))dy • We obtain the following discontinuous flux problem, Siddhartha Mishra A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work Scheme: Design ut + (f (u))x = (B̃ n (x))x Siddhartha Mishra A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work Scheme: Design ut + (f (u))x = (B̃ n (x))x I Local Discontinuous flux problems: By sampling define X B n (x) = B̃ n (xj )1{Ij } (x) j • We obtain the following discontinuous flux problem, ut + (f (u) − B n (x))x = 0, u(x, t n ) = u n (x) Siddhartha Mishra A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work Scheme: Design ut + (f (u))x = (B̃ n (x))x I Local Discontinuous flux problems: By sampling define X B n (x) = B̃ n (xj )1{Ij } (x) j • We obtain the following discontinuous flux problem, ut + (f (u) − B n (x))x = 0, u(x, t n ) = u n (x) I Local Riemann problems at each interface ut + (f (u) − Bjn )x n ) ut + (f (u) − Bj+1 x = 0 = 0 Siddhartha Mishra u(x, 0) = ujn n u(x, 0) = uj+1 x < xj+1/2 x > xj+1/2 A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work Shape of Adjacent fluxes f+ f − Siddhartha Mishra A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work Scheme: Design I Use a Exact Riemann Solver to solve the discontinuous flux problem Siddhartha Mishra A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work Scheme: Design I I Use a Exact Riemann Solver to solve the discontinuous flux problem RPs are simple to solve as the flux is additive. Siddhartha Mishra A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work Scheme: Design I I I Use a Exact Riemann Solver to solve the discontinuous flux problem RPs are simple to solve as the flux is additive. The update formula is ∆t n (F − F n j − 1/2) ∆x j+1/2 being the corresponding Godunov flux. ujn+1 = ujn − • with Fj+1/2 Siddhartha Mishra A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work Scheme: Design I I I Use a Exact Riemann Solver to solve the discontinuous flux problem RPs are simple to solve as the flux is additive. The update formula is ∆t n (F − F n j − 1/2) ∆x j+1/2 • with Fj+1/2 being the corresponding Godunov flux. Explicit formulas are available in most cases e.g (f convex) then ujn+1 = ujn − I n Fj+1/2 = max(f (max(uj , θ) − Bjn , f (min(uj+1 , θ)) − Bj+1 )) Siddhartha Mishra A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work Scheme: Properties I Discrete steady states of the scheme n n f (uj+1 ) − f (ujn ) = Bj+1 − Bjn • Reflects Flux-Source balance. Siddhartha Mishra A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work Scheme: Properties I Discrete steady states of the scheme n n f (uj+1 ) − f (ujn ) = Bj+1 − Bjn • Reflects Flux-Source balance. I Rankine-Hugoniot Conditions + Jump entropy conditions ⇒ Entropic Discrete steady states are preserved . Siddhartha Mishra A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work Scheme: Properties I Discrete steady states of the scheme n n f (uj+1 ) − f (ujn ) = Bj+1 − Bjn • Reflects Flux-Source balance. I Rankine-Hugoniot Conditions + Jump entropy conditions ⇒ Entropic Discrete steady states are preserved . I Flexibility in the averaging steps to obtain equivalent discrete steady states. Siddhartha Mishra A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work Scheme: Properties I Growth assumptions of flux and source + Boundedness of A ⇒ L∞ bounds. Siddhartha Mishra A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work Scheme: Properties I Growth assumptions of flux and source + Boundedness of A ⇒ L∞ bounds. I Entropy inequalities + Properties of the Riemann solution ⇒ Rate of Blow-up of BV -norm. Siddhartha Mishra A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work Scheme: Properties I Growth assumptions of flux and source + Boundedness of A ⇒ L∞ bounds. I Entropy inequalities + Properties of the Riemann solution ⇒ Rate of Blow-up of BV -norm. I Blow-up estimates on BV -norm ⇒ Compactness of Approximations (Compensated Compactness). Siddhartha Mishra A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work Scheme: Properties I Growth assumptions of flux and source + Boundedness of A ⇒ L∞ bounds. I Entropy inequalities + Properties of the Riemann solution ⇒ Rate of Blow-up of BV -norm. I Blow-up estimates on BV -norm ⇒ Compactness of Approximations (Compensated Compactness). I Jump entropy conditions + Structure of the scheme ⇒ Convergence to entropy solutions if A ∈ L∞ Siddhartha Mishra A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work Experiment 1 (Continuous Bottom) ut + f (u)x = z 0 (x)b(u) • With 1 2 = 2u cos(πx) if 4.5 < x < 5.5 −z(x) = 0 if Otherwise u(t, 0) = 2 f (u) b(u) = u u(0, x) = 0 • Explicit steady state is given by u(x) = 2 + z(x) • Comparision of Central Sources (CS), Existing Well-Balanced Scheme (BPV) and New Well-Balanced Scheme (AWBS) Siddhartha Mishra A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work At Steady State: ∆x = 0.1 3.5 3 2.5 2 1.5 1 AWBS:−−−−−−−−−−−−−−− BPV:................................ CS :− − − − − − − BT: o o o o o o o o o o o 0.5 0 −0.5 −1 0 2 4 Siddhartha Mishra 6 8 10 A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work Errors at Steady State CS AWBS BPV L∞ 0.1652 4.37 × 10−14 8.45 × 10−14 Siddhartha Mishra L1 0.4824 2.22 × 10−13 2.26 × 10−13 A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work Transients: ∆x = 0.1 Figure: Left:AWBS, Siddhartha Mishra Right:BPV A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work Transient snapshots 3 3.5 3 t = 2 Delta x = 0.1 2.5 t=5 Delta x = 0.1 2.5 AWBS:−−−−−−−−−−−−−−−−− CS :o o o o o o o o o BPV:− − − − − − − − 2 2 1.5 1.5 1 1 0.5 0.5 0 0 −0.5 −1 0 −0.5 2 4 6 8 10 Siddhartha Mishra −1 0 AWBS:−−−−−−−−−−− BPV: − − − − − − − CS :o o o o o o 2 4 6 8 10 A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work BPV at high resolution 3 2.5 t = 3 BPV(Delta x =0.1):− − − − − BPV(Delta x=0.01):−−−−−−−−− 2 1.5 1 0.5 0 −0.5 0 2 4 Siddhartha Mishra 6 8 10 A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work Experiment 2 (Discontinuous Bottom) ut + f (u)x = z 0 (x)b(u) • With 1 2 = 2u cos(πx) if 5 < x < 6 −z(x) = 0 if Otherwise u(t, 0) = 2 f (u) b(u) = u u(0, x) = 0 • Explicit steady state is given by u(x) = 2 + z(x) Siddhartha Mishra A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work At Steady State: ∆x = 0.1 3.5 3 2.5 2 1.5 AWBS:−−−−−−−−−−−−−−− BPV: o o o o o o o o 1 0.5 CS :− − − − − − − BT:−.. −. −. − . −. −. −.. −.. 0 −0.5 −1 0 2 4 Siddhartha Mishra 6 8 10 A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work Errors at Steady State CS1 AWBS BPV L∞ 0.8027 1.87 × 10−12 2.53 × 10−9 L1 1.6449 8.12 × 10−13 6.34 × 10−10 Table: Errors at the steady state for the three schemes with ∆x = 0.1 in Experiment 2 Siddhartha Mishra A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work A Transient snapshot 2.5 3.5 3 2 2.5 AWBS:−−−−−−−−−− BPV:− − − − − CS:o o o o o o 1.5 t =6 2 1 1.5 1 0.5 0.5 0 0 −0.5 0 AWBS:−−−−−−−−−−−−− BPV:− − − − − − − CS: o o o o o o o t −0.5 = 2 2 4 6 8 10 Siddhartha Mishra 0 2 4 6 8 10 A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work Experiment 3 (Another Discontinuous Bottom) ut + f (u)x = z 0 (x)b(u) • With 1 2 = 2u − cos(πx) if 5 < x < 6 −z(x) = 0 if Otherwise u(t, 0) = 2 f (u) b(u) = u u(0, x) = 0 • Explicit steady state is given by u(x) = 2 + z(x) Siddhartha Mishra A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work At Steady State: ∆x = 0.1 3.5 3 2.5 2 1.5 1 0.5 AWBS:−−−−−−−−−−−−−−− BPV:− − − − − − − t = 10 CS: o o o o o o o BT:−.. −... −... −.. −..−.. −. 0 −0.5 −1 0 2 4 Siddhartha Mishra 6 8 10 A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work Transients Figure: Left:AWBS, Siddhartha Mishra Right:BPV A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work Experiment 4 (Non-Monotone D) ut + f (u)x = z 0 (x)b(u) • With 1 3 = 3u cos(πx) if 4.5 < x < 5.5 −z(x) = 0 if Otherwise u(t, 0) = 2 f (u) b(u) = u u(0, x) = 0 • Difficult to define BPV as D = u 2 is not monotone. No problems with AWBS Siddhartha Mishra A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work Numerical Results at : ∆x = 0.1 1.5 2 1.5 1 0.5 1 0 −0.5 AWBS;−−−−−−−−−− t=3 0.5 0 AwBS:−−−−−−−−− BT:− − − − −1 2 4 6 8 10 Siddhartha Mishra −1.5 0 − t = 10 2 4 6 8 10 A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work Experiment 5 (Source not in Product form) ut + f (u)x = A(x, u) • With f (u) = 21 u 2 u(t, 0) = 1 A(x, u) = sin(2πxu 2 ) u(0, x) = 0 • Unclear how to define BPV in this case (other than using ODE solvers at each mesh point) whereas AWBS is well-defined Siddhartha Mishra A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work Numerical results with : ∆x = 0.1 1.3 1.2 1.1 AWBS:−−−−−−−−−−−− RK4 :o o o o o o o 1 0.9 0.8 0.7 0.6 0.5 0.4 0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 Siddhartha Mishra A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work Subtle problems with existing well-balanced schemes I Incorrect Shock speeds and strengths due to non-linear transformations. Siddhartha Mishra A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work Subtle problems with existing well-balanced schemes I Incorrect Shock speeds and strengths due to non-linear transformations. I Problems at resonance u = 0 Siddhartha Mishra A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work Summary • New class of Well-Balanced Schemes are Siddhartha Mishra A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work Summary • New class of Well-Balanced Schemes are I Very Simple to implement (Explicit formulas, No extra equations) Siddhartha Mishra A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work Summary • New class of Well-Balanced Schemes are I Very Simple to implement (Explicit formulas, No extra equations) I Robust and proved to be convergent. Siddhartha Mishra A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work Summary • New class of Well-Balanced Schemes are I Very Simple to implement (Explicit formulas, No extra equations) I Robust and proved to be convergent. I Very General: Work with different type of fluxes and sources. Siddhartha Mishra A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work Summary • New class of Well-Balanced Schemes are I Very Simple to implement (Explicit formulas, No extra equations) I Robust and proved to be convergent. I Very General: Work with different type of fluxes and sources. I Tailormade for discontinuous and singular sources. Siddhartha Mishra A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work Summary • New class of Well-Balanced Schemes are I Very Simple to implement (Explicit formulas, No extra equations) I Robust and proved to be convergent. I Very General: Work with different type of fluxes and sources. I Tailormade for discontinuous and singular sources. I Numerically efficient at both transients and steady states. Siddhartha Mishra A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work Ongoing and Future Work Siddhartha Mishra A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work Ongoing and Future Work I Discontinuous and Singular A. Siddhartha Mishra A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work Ongoing and Future Work I Discontinuous and Singular A. I Higher order schemes. Siddhartha Mishra A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work Ongoing and Future Work I Discontinuous and Singular A. I Higher order schemes. I Multi Dimensional problems. Siddhartha Mishra A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work Ongoing and Future Work I Discontinuous and Singular A. I Higher order schemes. I Multi Dimensional problems. I Systems: Shallow Water, Euler. Siddhartha Mishra A New Class of Well-Balanced Finite Volume schemes for Cons Outline The Problem Numerical Difficulties Existing Well-Balanced Schemes New Well-Balanced Schemes Numerical Experiments Summary and Future Work Ongoing and Future Work I Discontinuous and Singular A. I Higher order schemes. I Multi Dimensional problems. I Systems: Shallow Water, Euler. I Stiff Source terms. Siddhartha Mishra A New Class of Well-Balanced Finite Volume schemes for Cons