next up previous contents
Next: The solution algorithm Up: Time-discretization Previous: Introduction

The -method

  In this section we restrict ourselves to the time discretization of the momentum equations. The treatment of the transport equations will be done in exactly the same manner. Furthermore, for brevity the arguments will be dropped from the operator S. Application of the -method to (8.1), (8.2) gives

  

where lies between zero and unity, n denotes the preceding time level, n+1 the new time level and is the time step. For and we obtain the first order explicit and implicit Euler schemes, respectively, and for we have the second order Crank-Nicolson scheme. The -method is unconditionally stable for . In the range a time-step restriction is necessary. At this moment has not been tested, except .
To solve (8.4), (8.5) it is necessary to linearize the term . In ISNaS, the convective terms are linearized by a Newton linearization as given in formula (4.7). Coefficients that depend on the solution, like for example the viscosity, are evaluated at the preceding time-level.
Practical implementation:
Instead of solving (8.4), (8.5) immediately, we introduce an intermediate level by:

 

If we assume that is linearized, i.e. can be written as , then (8.4) reduces to:

 

Substitution of (8.6) into (8.7) and (8.5) gives

 

 

From (8.6) it then follows that:

 

Once the momentum equations have been solved for , each of the transport equations for is solved for one time step. Exactly the same methodology as described above is employed. Quantities already computed, like the velocity are substituted in these equations, thus improving the stability. However, it should be noted that the source term contains deferred corrections which are evaluated explicitly, so that no contribution to coefficients of the system to be solved for the new time level is involved.


Tatiana Tijanova
Wed Mar 26 10:36:42 MET 1997