PART I: Modeling of Production of Aluminium

PART I: Modeling of Production of Aluminium PART II: The one-fluid MHD equations - Analysis and Numerics PART III: The two-fluid MHD equations - Analysis and Numerics PART IV: Simulation of Aluminium Electrolysis PART V: Summary Production of Aluminium: Modeling, Analysis and Numerics Andreas Prohl (U Tübingen) L. Banas (HW Edinburgh) Andreas Prohl (U Tübingen) PART I: Modeling of Production of Aluminium PART II: The one-fluid MHD equations - Analysis and Numerics PART III: The two-fluid MHD equations - Analysis and Numerics PART IV: Simulation of Aluminium Electrolysis PART V: Summary Contents PART I: Modeling of Production of Aluminium PART II: The one-fluid MHD equations - Analysis and Numerics PART III: The two-fluid MHD equations - Analysis and Numerics PART IV: Simulation of Aluminium Electrolysis PART V: Summary Andreas Prohl (U Tübingen) PART I: Modeling of Production of Aluminium PART II: The one-fluid MHD equations - Analysis and Numerics PART III: The two-fluid MHD equations - Analysis and Numerics PART IV: Simulation of Aluminium Electrolysis PART V: Summary ... the industrial problem Modeling of the problem Continuum Model Part I: Modeling of Production of Aluminium Andreas Prohl (U Tübingen) PART I: Modeling of Production of Aluminium PART II: The one-fluid MHD equations - Analysis and Numerics PART III: The two-fluid MHD equations - Analysis and Numerics PART IV: Simulation of Aluminium Electrolysis PART V: Summary ... the industrial problem Modeling of the problem Continuum Model Motivation: Production of Aluminium I electrolysis: reduce aluminium oxid to aluminium I two non-miscible, conducting, incompressible fluids I high temperatures & high currents: no experimental data I industrial challenge: reduce electric power waste Andreas Prohl (U Tübingen) PART I: Modeling of Production of Aluminium PART II: The one-fluid MHD equations - Analysis and Numerics PART III: The two-fluid MHD equations - Analysis and Numerics PART IV: Simulation of Aluminium Electrolysis PART V: Summary I ... the industrial problem Modeling of the problem Continuum Model anod-metal distance: controlled movement of interface Andreas Prohl (U Tübingen) PART I: Modeling of Production of Aluminium PART II: The one-fluid MHD equations - Analysis and Numerics PART III: The two-fluid MHD equations - Analysis and Numerics PART IV: Simulation of Aluminium Electrolysis PART V: Summary ... the industrial problem Modeling of the problem Continuum Model ... the industrial problem I chem. reaction A: at surface of carbon anod 3 3 3 O 2− + C → CO2 (gas) + 6 e− 2 2 I chem. reaction B: at interface between the two fluids Al2 O3 + 6 e− → 2 Al 3+ + 3 O 2− + 6 e− → 2 Al + 3 O 2− I Global balance: 2 Al2 O3 + 3 C → 4 Al + 3 CO2 I I needed: ≈ 1000 C 0 , high intensity of current ”the higher intensity, the higher production” Andreas Prohl (U Tübingen) PART I: Modeling of Production of Aluminium PART II: The one-fluid MHD equations - Analysis and Numerics PART III: The two-fluid MHD equations - Analysis and Numerics PART IV: Simulation of Aluminium Electrolysis PART V: Summary I ... the industrial problem Modeling of the problem Continuum Model Goals/Problem: reduce power waste - reduce anod-metal distance: small distance between anod and surface of aluminium layer (‘a few centimeters’) - strong Lorentz forces: motion of the interface; instabilities - avoid short -circuits: no touching of fluid-fluid interface and anod I Questions: how stabilize the position of the interface? I Control parameter: height of anod, intensity of current, geometry I experimental observations difficult to obtain Andreas Prohl (U Tübingen) PART I: Modeling of Production of Aluminium PART II: The one-fluid MHD equations - Analysis and Numerics PART III: The two-fluid MHD equations - Analysis and Numerics PART IV: Simulation of Aluminium Electrolysis PART V: Summary ... the industrial problem Modeling of the problem Continuum Model Modeling of the problem I Physical phenomena and simplifying assumptions: - magnetohydrodynamics (MHD) - moving interface - electrochemistry (concentration of chemical species not homogeneous throughout liquids - three-phase flows (bubbles of carbon oxydes at surface of anods) - solidification process at boundaries - temperature effects (influence of physical parameters) Andreas Prohl (U Tübingen) PART I: Modeling of Production of Aluminium PART II: The one-fluid MHD equations - Analysis and Numerics PART III: The two-fluid MHD equations - Analysis and Numerics PART IV: Simulation of Aluminium Electrolysis PART V: Summary ... the industrial problem Modeling of the problem Continuum Model Continuum Model I Maxwell equations coupled to multi-fluid Navier-Stokes equations ∂t ρ + div(ρu) = 0, 1 ∂t (ρu) + div(ρu ⊗ u) − div 2ηε(u) + ∇p = ρf + curl B × B µ div u = 0 div B = 0 ∂t B + curl E = 0 B −∂t (εE) + curl = j µ Andreas Prohl (U Tübingen) for j = σ (E + u × B), PART I: Modeling of Production of Aluminium PART II: The one-fluid MHD equations - Analysis and Numerics PART III: The two-fluid MHD equations - Analysis and Numerics PART IV: Simulation of Aluminium Electrolysis PART V: Summary I I ... the industrial problem Modeling of the problem Continuum Model ρ density of fluids ε (u) = 21 (∇u + ∇uT ) B magnetic field σ electric conductivity j electric current η viscosity Assumption: ’low frequency hypothesis’: neglect ∂t (εE). ∂t B + curl( 1 curl B) = curl(u × B), µ0 σ div B = 0 I ⇒ ∂t B + curl( µ01σ curl B) = curl(u × B), div B = 0 magnetic boundary conditions on ∂ Ω: B·n , Andreas Prohl (U Tübingen) curl B × n PART I: Modeling of Production of Aluminium PART II: The one-fluid MHD equations - Analysis and Numerics PART III: The two-fluid MHD equations - Analysis and Numerics PART IV: Simulation of Aluminium Electrolysis PART V: Summary Analytical results/Numerical Strategies Numerical Strategies to solve one-fluid MHD equation PART II: The one-fluid MHD equations – Analysis and Numerics Andreas Prohl (U Tübingen) PART I: Modeling of Production of Aluminium PART II: The one-fluid MHD equations - Analysis and Numerics PART III: The two-fluid MHD equations - Analysis and Numerics PART IV: Simulation of Aluminium Electrolysis PART V: Summary Analytical results/Numerical Strategies Numerical Strategies to solve one-fluid MHD equation Analytical results/Numerical Strategies I M. Sermange & R. Temam [1983]: I I global weak solutions for Ω ⊂ R3 local strong solutions for Ω ⊂ R3 I Goal: Develop schemes that approximate weak solutions I for simplicity here: only temporal discretization Andreas Prohl (U Tübingen) PART I: Modeling of Production of Aluminium PART II: The one-fluid MHD equations - Analysis and Numerics PART III: The two-fluid MHD equations - Analysis and Numerics PART IV: Simulation of Aluminium Electrolysis PART V: Summary I Analytical results/Numerical Strategies Numerical Strategies to solve one-fluid MHD equation Scheme A: Let n ≥ 1. Find (un , pn , bn , r n ) such that 1 dt un − ∆un + (un−1 · ∇)un + (div un−1 )un 2 +bn−1 × curl bn + ∇pn = gn div un = 0 div bn = 0 dt bn + curl(curl bn ) − curl(un × bn−1 ) − ∇r n = 0 Andreas Prohl (U Tübingen) PART I: Modeling of Production of Aluminium PART II: The one-fluid MHD equations - Analysis and Numerics PART III: The two-fluid MHD equations - Analysis and Numerics PART IV: Simulation of Aluminium Electrolysis PART V: Summary Analytical results/Numerical Strategies Numerical Strategies to solve one-fluid MHD equation Numerical Strategies to solve one-fluid MHD equation I discrete energy law: “multiply 1st eqn. by un , and 3rd eqn. by bn ” i i kh 1 h n 2 ||dt un ||2 + ||dt bn ||2 dt ||u || + ||bn ||2 + 2 2 n 2 +||∇u || + ||curl bn ||2 = (gn , un ) Andreas Prohl (U Tübingen) PART I: Modeling of Production of Aluminium PART II: The one-fluid MHD equations - Analysis and Numerics PART III: The two-fluid MHD equations - Analysis and Numerics PART IV: Simulation of Aluminium Electrolysis PART V: Summary I Analytical results/Numerical Strategies Numerical Strategies to solve one-fluid MHD equation Result [P.’08]: Subconvergence (k , h → 0) to weak solution of one-fluid MHD system – drawback of scheme: system coupled +/- iterative decoupling strategy: restrictive mesh-constraint k ≤ Ch4 needed for convergence Andreas Prohl (U Tübingen) PART I: Modeling of Production of Aluminium PART II: The one-fluid MHD equations - Analysis and Numerics PART III: The two-fluid MHD equations - Analysis and Numerics PART IV: Simulation of Aluminium Electrolysis PART V: Summary I Analytical results/Numerical Strategies Numerical Strategies to solve one-fluid MHD equation Scheme B: decoupled scheme 1 dt un − ∆un + (un−1 · ∇)un + (div un−1 )un 2 +bn−1 × curl bn−1 − ∇pn = gn dt bn + curl(curl bn ) − curl(un−1 × bn−1 ) − ∇r n = 0 I Property: A perturbed energy law holds in a fully discrete setting, for k ≤ Ch3 . I Result [P. ’08]: Subsequence convergence (k , h → 0) to weak solution of one-fluid MHD-system. Andreas Prohl (U Tübingen) PART I: Modeling of Production of Aluminium PART II: The one-fluid MHD equations - Analysis and Numerics PART III: The two-fluid MHD equations - Analysis and Numerics PART IV: Simulation of Aluminium Electrolysis PART V: Summary The Model PART III: The two-fluid MHD equations – Analysis and Numerics Andreas Prohl (U Tübingen) PART I: Modeling of Production of Aluminium PART II: The one-fluid MHD equations - Analysis and Numerics PART III: The two-fluid MHD equations - Analysis and Numerics PART IV: Simulation of Aluminium Electrolysis PART V: Summary The Model The Model I Navier-Stokes with variable density and viscosity with Maxwell’s equation 1 curl b × b, µ̄ (ρu)t + div(ρu ⊗ u) − div η(ρ)D(u) = −∇p + g + div u = 0, ρt + div(ρu) 1 1 bt + curl curl b ξ (ρ) µ̄ div b = 0, = curl(u × b), = 0, together with IC’s & BC’s Andreas Prohl (U Tübingen) (1) PART I: Modeling of Production of Aluminium PART II: The one-fluid MHD equations - Analysis and Numerics PART III: The two-fluid MHD equations - Analysis and Numerics PART IV: Simulation of Aluminium Electrolysis PART V: Summary I Assumptions: 0 < ξ¯− ≤ ξ ≤ ξ¯+ . 1. 0 < η̄− ≤ η ≤ η̄+ , 2. ρ0 = I The Model ρ 1 > 0, ρ 2 > 0, constant on Ω1 , constant on Ω2 , with Ω1 ∪Ω2 = Ω, meas (Ωi ) > 0 Properties: 1. non-negativity, boundedness of ρ ρ1 ≤ ρ ≤ ρ2 in ΩT 2. energy law: h ρ|u|2 1 d R 2 dt Ω 2 + |b|2 µ̄ i dx + R h = Andreas Prohl (U Tübingen) i η(ρ)|εε (u)|2 + µ̄ 2 ξ1(ρ) |curl b|2 dx Ω R Ω ρg · udx PART I: Modeling of Production of Aluminium PART II: The one-fluid MHD equations - Analysis and Numerics PART III: The two-fluid MHD equations - Analysis and Numerics PART IV: Simulation of Aluminium Electrolysis PART V: Summary The Model Analytical results I weak formulation: For I I I I Ω ⊂ R3 polyhedral domain n o u0 , b0 ∈ H := ξ ∈ L2 (Ω) : divξξ = weakly in Ω,ξξ · n = 0 on ∂ Ω ρ0 as above f ∈ L2 (0, T ; L2 ) exists a weak solution u ∈ L∞ (0, T ; H) ∩ L2 (0, T ; J) b ∈ L∞ (0, T ; H) ∩ L2 (0, T ; X) ρ ∈ L∞ (ΩT ) ∩ C([0, T ]; Lp ) for all p ≥ 1 which satisfies an energy inequality. Andreas Prohl (U Tübingen) PART I: Modeling of Production of Aluminium PART II: The one-fluid MHD equations - Analysis and Numerics PART III: The two-fluid MHD equations - Analysis and Numerics PART IV: Simulation of Aluminium Electrolysis PART V: Summary The Model ... Tools employed I I I general Galerkin method Aubin-Lions compactness result Compactness result by J. Di Perna & P.L. Lions [’89]: Solvability of ρt + div(uρ) = f in ΩT , ρ(0, ·) = ρ0 ∈ L∞ (Ω) Let {ρk }k ≥0 ⊂ L∞ (0, T ; L∞ (Ω)) solve (ρk )t + div(uk ρk ) + [divuk ]ρk = fk (ρk )(0, ·) = (ρk ) Andreas Prohl (U Tübingen) in Ω in ΩT (2) PART I: Modeling of Production of Aluminium PART II: The one-fluid MHD equations - Analysis and Numerics PART III: The two-fluid MHD equations - Analysis and Numerics PART IV: Simulation of Aluminium Electrolysis PART V: Summary The Model Assume: (i) {uk }k ⊂ L1 0, T ; W01,2 , and u ∈ L1 0, T ; W1,2 , such that 0 uk → u in L1 0, T ; L2 divuk → divu in L1 0, T ; L2 (ii) fk → f in L1 0, T ; L2 (ρk )0 → ρ0 in L2 (Ω) Then ρk → ρ in L2 0, T ; L2 where ρ : Ω → R is unique solution of (2). Andreas Prohl (U Tübingen) PART I: Modeling of Production of Aluminium PART II: The one-fluid MHD equations - Analysis and Numerics PART III: The two-fluid MHD equations - Analysis and Numerics PART IV: Simulation of Aluminium Electrolysis PART V: Summary The Model Problems to construct a convergent Finite Element Discretization I Discrete Energy Law: In continuous setting: multiply (11 ) with u (13 ) with 12 |u|2 Observation: 12 |u|2 no admissible test function in FE-discretization Idea (N. Walkington [’07]): reformulation (ρu)t +div (ρu ⊗ u) = o 1n ρut +[ρu·∇]u+(ρu)t +div (ρu ⊗ u) 2 Andreas Prohl (U Tübingen) PART I: Modeling of Production of Aluminium PART II: The one-fluid MHD equations - Analysis and Numerics PART III: The two-fluid MHD equations - Analysis and Numerics PART IV: Simulation of Aluminium Electrolysis PART V: Summary I The Model (uniform) positivity and L∞ - boundedness of discrete densities: Idea: M-matrix property of stiffness matrix related to 1 (dt ρ n , χ)h +(Un · ∇ρ n , χ)+ ([divUn ]ρ n , χ)+αhα (∇ρ n , ∇χ) = 0 2 Tools: I I I I α > 0: M–matrix properly gives 0 < ρ1 ≤ ρ n ≤ ρ2 < ∞ numerical integration regularization term to NSE β2 hβ2 (∇dt Un , ∇Wn ) β2 ≥ 0 discrete version of compactness result by R. DiPerna & P.L. Lions: Idea (N. Walkington [’07]) Andreas Prohl (U Tübingen) PART I: Modeling of Production of Aluminium PART II: The one-fluid MHD equations - Analysis and Numerics PART III: The two-fluid MHD equations - Analysis and Numerics PART IV: Simulation of Aluminium Electrolysis PART V: Summary The Model U → divu in L1 (0, T ; L2 ) Tool: to validate divU we add the regularization term to NSE β1 k β1 (divUn , divWn ) I , β1 > 0 a compactness result of J.L. Lions to control temporal changes of iterates: There exists κ > 0 Z Th i U + −U U − ||2 + ||B B + −B B − ||2 ds ≤ Ck κ ρ 1 ||U 0 Tool: I I I inverse estimates mesh-constraints: F (k , h) ≥ 0 Problems already experienced for one-fluid MHD equation: H(curl), H(div), Tool: discrete compactness result by F. Kikuchi [’89] Andreas Prohl (U Tübingen) PART I: Modeling of Production of Aluminium PART II: The one-fluid MHD equations - Analysis and Numerics PART III: The two-fluid MHD equations - Analysis and Numerics PART IV: Simulation of Aluminium Electrolysis PART V: Summary The Model Result: A stable, convergent FE–based discretization I I Let {(ρ n , Un , Bn )} be solution of FE –based fully discrete scheme L. Banas, A.P. [’08]: Let I I I T be strongly acute F (k , h; α, β1 , β2 , d) ≥ 0 be valid ρ 0 → ρ0 in L2 , (U0 , B0 ) * (u0 , b0 ) in [L2 ]2 For (k , h) → 0 exist a convergent subsequence, and (ρ, u, b), s.t. ∗ B * b in L∞ (0, T ; L2 ), σ * ρ in L∞ (0, T ; L∞ ) U * u in L∞ (0, T ; L2 ),B Where (u, b, ρ) is weak solution to (1). Andreas Prohl (U Tübingen) PART I: Modeling of Production of Aluminium PART II: The one-fluid MHD equations - Analysis and Numerics PART III: The two-fluid MHD equations - Analysis and Numerics PART IV: Simulation of Aluminium Electrolysis PART V: Summary The Model From Discretization to a fully practical Scheme I Status: + a convergent discretization of (1) – a nonlinear algebraic problem I Tool: A fixed point strategy for every n ≥ 1 + + + + + linear, decoupled problems for ρ n,l , un,l , bn,l n,l a thresholding criterion contraction property for restrictive F (k , h) ≥ 0 perturbed discrete energy law, and uniform upper and lower l ∞ –bounds for ρ n,l n,l Andreas Prohl (U Tübingen) PART I: Modeling of Production of Aluminium PART II: The one-fluid MHD equations - Analysis and Numerics PART III: The two-fluid MHD equations - Analysis and Numerics PART IV: Simulation of Aluminium Electrolysis PART V: Summary PART IV: Simulation of Aluminium Electrolysis Andreas Prohl (U Tübingen) PART I: Modeling of Production of Aluminium PART II: The one-fluid MHD equations - Analysis and Numerics PART III: The two-fluid MHD equations - Analysis and Numerics PART IV: Simulation of Aluminium Electrolysis PART V: Summary Aluminium Electrolysis – Two-Fluid MHD I evolution of interface between two conducting fluids I I top: lighter cryolite, bottom: heavier liquid aluminium magnitude of magnetic field Andreas Prohl (U Tübingen) PART I: Modeling of Production of Aluminium PART II: The one-fluid MHD equations - Analysis and Numerics PART III: The two-fluid MHD equations - Analysis and Numerics PART IV: Simulation of Aluminium Electrolysis PART V: Summary I 2D Velocity profile I 2D Magnetic field profile Andreas Prohl (U Tübingen) PART I: Modeling of Production of Aluminium PART II: The one-fluid MHD equations - Analysis and Numerics PART III: The two-fluid MHD equations - Analysis and Numerics PART IV: Simulation of Aluminium Electrolysis PART V: Summary PART V: Summary I Aluminium production: Two Fluid–MHD problem I Goal/Problem: Develop convergent FE–based discretization such that computed iterates satisfy relevant properties I Problems: Discrete energy law, upper and lower bounds for density, compactness result by R. DiPena & P.L. Lions I Tools: reformulation, M–matrix property, numerical integration, regularization terms, mesh constraints Andreas Prohl (U Tübingen) PART I: Modeling of Production of Aluminium PART II: The one-fluid MHD equations - Analysis and Numerics PART III: The two-fluid MHD equations - Analysis and Numerics PART IV: Simulation of Aluminium Electrolysis PART V: Summary Thank you for your attention! Andreas Prohl (U Tübingen)

PART I: Modeling of Production of Aluminium

Products

Support

PART I: Modeling of Production of Aluminium

Add this document to collection(s)

Add this document to saved

Suggest us how to improve StudyLib