Computations of Fluid Dynamics using the Interface Tracking Method Zhiliang Xu Email: zxu2@nd.edu Department of Mathematics University of Notre Dame Outline Computational Fluid Dynamics Front Tracking Method Compressible & incompressible flows Governing equations Numerical methodology Formulation Improving the accuracy Conclusions and Future Plans Compressible & Incompressible Flows 1. 2. Approximations & Governing equations • Continuum assumption • The fundamental laws (basis): Conservation • Thermo-dynamical equation of state (EOS) e.g. PV=RT Compressibility Mach number: M = v/c M > 0.3: compressible flow Compressible, inviscid flow: Euler equations Incompressible viscous flow: Incompressible Navier-Stokes equations • No turbulence modeling Nonlinear Hyperbolic Conservation Laws Euler equations: (Gas dynamics) U F 0 t x U ( , u, E )T F ( u, p u 2 , ( E p)u )T Equation of state: Scalar examples: (Traffic flow) (Burgers’ equation) p p ( , e) 1 2 E e u 2 t ( u ) x 0 u ( ) umax (1 / max ),0 max ut (u 2 / 2) x 0 Scalar Conservation Laws Conservation equation: ut f (u ) x 0, f 0 f: Flux function d x2 u ( x, t )dx f (u ( x1 , t )) f (u ( x2 , t )), dt x1 Linear Advection Equation: ut c0u x 0, C0=const. > 0 u ( x,0) u0 ( x) Solution: u( x, t ) u0 ( x c0t ) u u(x,t) t x u(x,0) Nonlinear Scalar Equation ut c(u)u x 0, Along a characteristic curve which has slope: c(u ) f (u ) with u ( x,0) u0 ( x) dx c(u ) dt d u ( x(t ), t ) x x u t u 0 dt The total derivative: t where u (x,t) is const. along this curve. Along this line, u = u0(x0) x0 x Solve x x0 c(u0 ( x0 ))t for u ( x, t ) u0 ( x0 ) x0 Breakup of Continuous Solution Assume: c (u ) 0. t dx du c(u ), 0 dt dt Characteristics for nonlinear equations t3 Characteristics cross, the wave “breaks”. tB t1 x u Breaking solution: successive profiles corresponding to the times 0, t1, tB, t3 u(x,0) t1 tB t3 x Weak Solutions u t tB x Weak solutions: {u t x f (u ) x }dxdt 0, R R Jump Condition (Rankine-Hugoniot Condition): (ul ur ) (t ) f (ul ) f (ur ) (t ) : the propagation speed of the discontinui ty (Lax) Entropy Condition & Shock t f (ul ) (t ) To pick physically relevant solutions. Ml f (ul ) (t ) f (ur ) Mr f (ur ) x Shock: A discontinuity that satisfies the jump condition and the entropy condition. Riemann Problem (Scalar Case) Init. value problem with piecewise const. data: Admit: Similarity solution: ut f (u ) x 0, f 0 u , x 0 u0 l ur , x 0 u ( x, t ) u (x, t ), const 0. u u ( ), x / t Riemann Solution (Scalar Case) Case 1: Const. State: ul u r u ( ) const . Shock speed s t Case 2: Shock wave: ul u r ul , x st u( x, t ) ur , x st s f (ul ) f (ur ) ul u r x 0 Case 3: Rarefaction wave: ul ur u f (u ) 0 x f (ul )t ul , u ( x, t ) v( x / t ), f (ul )t x f (ur )t , f (v( )) u, x f (ur )t r Rarefaction wave t f (u ( )) x 0 Numerical Computation Milestones: • Computing discontinuous solutions by Peter Lax (1950s) (Lax-Friedrichs scheme, LaxWendroff scheme) (SIAM Reviews Vol. 11, No. 1. 1969) • Godunov’s scheme, upwind schemes • High order schemes: TVD, MUSCL, PPM, ENO, WENO, etc • Interior or Free Boundary Tracking 1. 1D, 2D interface tracking by Richtmyer and Morton (1960s) 2. Front tracking by Glimm, McBryan etc. (1980s) 3. Others (level set, VOF, etc.) Numerical Solution: Finite Volume Method uin 1 tn+1 f xi1 / 2 f xi1 / 2 Space-time Boundary of the Volume tn Space-time Volume Xi : Cell Center Xi+1/2 : Cell edge u in Xi Xi-1/2 Xi+1/2 Xi+1 1D Finite Volume Scheme {ut f (u ) x }dxdt 0, uin 1 uin 1 xi1 / 2 u : u ( x, t n )dx x i 1 / 2 x n i t 1 tn1 1 tn1 [ f (u ( xi 1/ 2 , t )) dt f (u ( xi 1/ 2 , t )) dt ] x t tn t tn Average of exact flux u n 1 i (Cell average value). f xi1/ 2 t ˆ u ( f i 1/ 2 fˆi 1/ 2 ) x n i Numerical Flux (1) Computing Discontinuous Solutions Conservation: u t ˆ u ( f i 1/ 2 fˆi 1/ 2 ) x n 1 i n i Single valued flux on each cell edge (…,Xi+1/2,…). N u n 1 i i Consistency: N u in i f f 0 fˆi 1/ 2 (u, u) f (u) The Entropy Condition: The CFL condition: with t f 1 x ( 2) Computing Discontinuous Solutions Examples Godunov’s Method (1959): Semi-Discrete Method: fˆi 1/ 2 h(u * (u i1 / 2i , u i1 / 2 )) du L( u ) dt fˆi 1/ 2 (u i , ui 1 ) f (u* (u i , ui 1 )) du i 1 [ f i 1/ 2 f i 1/ 2 ] dt x Spatial ENO/WENO reconstruction Temporal direction: TVD Runge-Kutta Dynamic Interface Tracking Rayleigh-Taylor Mixing The Level Set Method Level Set: v 0 t ( x , t 0) d Interface : (t ) : 0 Discrete Representation of Tracking y Volume filling rectangular x mesh (Eulerian Coord.) (N-1) dimensional Lagrangian mesh (interface) ij : xi 1 / 2 y j 1 / 2 1 ( x, y )dxdy x y i 1 / 2 j 1 / 2 xy Y (i,j) X A 2D Representation Front Tracking: Hybrid method, 2 meshes. A 3D Interface Time Marching & Coupling I n 1 tn+1 tn Xi-1 Xi I n Xi+1 Advancing solution in 1D To advance the numerical solution in Front Tracking: (1) Explicit procedure for interface propagation + (2) Updating states (grid cell center) Two way coupling: 1. Interface dynamics to ambient region (interior). 2. Non-interface solution variation to interface dynamics. Separation of Interface Propagation y x Normal Tangent • Operator Splitting to separate normal and tangential propagation • Normal propagation to move interface position & coupling • Tangent propagation to include information flowing tangentially along the curve. Normal Propagation of Interface Point Move the point position and couple the interior wave solution to interface dynamics. Step 1: Updated left and right states of the point Step 2: (Material interface) t t t Contact Sl0 Sr0 New position t t Sl dx uc dt n S2 S1 S 0 S 0 S r 0 S1 S2 Left and right states of the point t dx u dt S 1 S f S0 S0 S2 Sr dx u c dt n Sb S1 S r Riemann solution Method of characteristics (Coupling) S2 Advancing Eulerian Grid Solution Ghost cell method: Coupling interface dynamics to interior tn+1 : Fluid 1 Extrapolate : Fluid 2 tn Xi-1 Xi : Interface Xi+1 Cell edge Ui n 1 t L Ui ( Fi 1/ 2 Fi 1/ 2 ) x n L Fi 1/ 2 Fˆ (U in ,U ileft 1 ) U n 1 i 1 U n i 1 R t ( Fi 3 / 2 Fi 1/ 2 ) x R Fi 1/ 2 Fˆ (U iright ,U in1 ) Conservative Front Tracking - Formulation Discontinuity t Space-time volume V Xi UL UR V ( U F (U ))dV 0 t Space-time interface Xi+1/2 Xi+1 x A moving discontinuity surface bounds a time-dependent volume V. Fi L1/ 2 F (U L ) (t ) U L Fi R1/ 2 F (U R ) (t ) U R Redefine the flux through the discontinuity by R-H condition. 2D Space-Time Volumes Top face t y x Space-time hexahedron Improved Accuracy Theorem: The conservative tracking method improves accuracy by at least one order. 1D Numerical Validation Init. Condition: Shock-Rarefaction 2D Axisymmetric Richtmyer-Meshkov Instability Light gas Heavy gas Material interface Shock wave Init. Condition (Density Plot) Conservative tracking simulation Non-conservative tracking simulation 2D Axisymmetric Richtmyer-Meshkov Instability Conservative Tracking, 100*200 grid Non-Conservative Tracking, 100*200 grid Non-Conservative Tracking, 200*400 grid Amplitude (a): the height of the interface perturbation. a ( hsp hbb ) h_sp and h_bb are distances from origin to the tips of the spike and the bubble respectively. Computations of Incompressible Flows ut u u p u f (u ) 0 What is the role of the pressure? Hodge Decomposition Projection Methods * u u where (u ) 0 Projection Method 1. Advancing the momentum equation in time to determine an intermediate velocity which is not required to be divergence-free. 2. Project the intermediate velocity field onto the space of divergencefree field. The gradient part is used to update the pressure. The Numerical Method u d x L ( u , ) pdS t ij ij L(u , ) u u dS D dS k s ndx fdx ij Advancing the front: Advancing materiel properties: ij The Numerical Method Compute the intermediate velocity: Projection: p Surface tension: n 1/ 2 p n 1/ 2 t B k s ndx A knds (t A B B tA ) The Blood Flow Modelling (ut u u ) p u f (u ) 0 c u c D c t Conclusions & Future Plans The front tracking method to describe the interface. On the tracking method: To achieve uniform high order accuracy. On the application: To develop a blood flow model in the multiscale context