Title: Approximate Riemann Solvers for Multi-component flows
1Approximate Riemann Solvers for Multi-component
flows
- Ben Thornber
- Academic Supervisor D.Drikakis
- Industrial Mentor D. Youngs (AWE)
- Aerospace Sciences
- Fluid Mechanics Computational Science
2Aims
- Describe the derivation of a new approximate
Riemann solver for multi-component flows - Present a series of test cases illustrating the
performance of the scheme for two different model
equations - Compare and contrast the Mass Fraction and Total
Enthalpy Conservation of the Mixture models.
3Outline
- Introduction
- Governing equations
- Godunov method
- Higher Order Extensions
- Characteristics-Based Solver
- Test Cases and Validation
- Conclusions
4Governing equations
- Begin with the Euler equations in primitive
variables
5Governing Equations
- Augment them with two multicomponent models
- Mass Fraction
- See, for example, Abgrall (1988) or Larrouturou
(1989)
6Governing Equations
- 2) Total Enthalpy Conservation of the Mixture
(ThCM) - See Wang, S.P. et al (2004)
7Method of Solution
- Godunov finite volume method
- Dual time stepping method
Godunov (1959)
Jameson (1991)
8Higher Order Accuracy
- Utilise the MUSCL method (Van Leer, 1977)
- With 2nd order Superbee, Minmod, Van Leer, Van
Albada and 3rd order Van Albada limiters (See
Toro, 1997)
9Characteristics Based Approximate Riemann Solver
- An extension of Eberles scheme (Eberle, 1987)
- As the governing equations are identical then the
derivation holds for both models - Considering the Euler equations split
directionally, thus solving -
- The time derivative is replaced by the
Characteristic Derivative
10Non-Conservative Invariants
- After some manipulation this gives six
characteristic equations for six unknown flow
variables
11Converting to conservative form
- Now we convert the equations to conservative form
using the chain rule of differentiation
12Converting to Conservative Form
- For pressure this is a little more complex
- Giving
- Where
13Compact form
- After considerable manipulation the
characteristics based variables with which the
Godunov fluxes are calculated are
14Compact form
15Numerical Tests
- Used 5 test cases to examine the performance of
the new scheme and the multi-component models
employed - A ) Weak Post-shock Contact Discontinuity
- See Wang et al (2004)
- B ) Shock-Contact surface interaction
- See Karni (1994), Abgrall (1996), Shyue (2001),
Wang et al (2004) - C ) Modified Sod shock tube
- See Abgrall and Karni (2000), Chargy et al
(1990), Karni (1996) and Larroururou (1989) - D ) Shock interaction with a Helium Slab
- See Abgrall (1996), Wang et al (2004)
- E ) Convection of an SF6 Slab
16Test A Weak Post-shock Contact Discontinuity
17Test A
- 2nd order accuracy with Minbee characteristic
bump in the MF density profile
18Test A Limiters
- Density profile at the contact surface a) 1st
order, b) Superbee, c) Van Albada, d) Van Leer,
e) 3rd order Van Albada
19Test B Shock-Contact surface interaction
20Test B
- Oscillation free results for all limiters
- Mass fraction model captures the contact surface
over fewer points
21Test C Modified Sod shock tube
22Test C
- All profiles are captured reasonably well
23Test C Density and velocity profiles
- Mass fraction model has a typical density
undershoot and a velocity jump at the contact
surface - Slight oscillations in the ThCM model
24Test D Shock interaction with a Helium Slab
25Test D
- Very complex problem oscillatory results for
the Mass Fraction model - Dissipative solution for the ThCM model
26Test D - Convergence
- Dissipative solution for the ThCM model, with
2000 points it is more dissipative than the mass
fraction model with 400 points
27Test E Convection of an SF6 slab
28Test E Results after 1 time step
- Pressure equilibrium is not maintained for the
ThCM model or the Mass Fraction model
29Test E Results after 1 time step
- Considering a convected contact surface computed
using finite volume upwind method - Where this fraction 0.6 in the case of SF6 to
air
30Conclusions
- A new multi-component approximate Riemann solver
has been developed and validated - The Total Enthalpy Conservation of the Mixture
model is better for flows where g is not close to
1, and the difference in gas densities is low. - The Mass Fraction model captures discontinuities
in fewer points - Neither model preserves pressure equilibrium
exactly in the case of a convected contact
surface, however the extent of the error depends
on the gases simulated.
31References
32References
33References
34References