What is an optimal choice for the model parameters of a non-reflective boundary ?
|
A perfectly nonreflecting boundary condition sets the amplitude of incoming characteristic waves to zero. Unfortunately, this kind of condition is not robust, so instead the wave amplitude is modified to 'feel' the far-field pressure slightly: L1 = K (P-P_infinity) When you create a nonreflecting pressure outlet, the pressure you set is actually P_infinity. K is a dimensional 'reflection coefficient' which controls how strongly it is felt: K = sigma (1-M^2) c/L where - sigma ('Reflection Factor') is a dimensionless coefficient. The default is 0.25, as recommended by Poinsot and Lele, and is usually appropriate. 0 gives the perfectly nonreflecting condition and infinite recovers the usual pressure outlet. - M ('Reflection Mach Scale') is a factor designed to make the outlet 'perfectly nonreflecting' as sonic speed is approached. In this limit, however, a supersonic outlet should be used instead, so we default this to zero. - c is the local sound speed as calculated by the solver - L ('Reflection Length Scale') is a final factor to make things dimensionally consistent. By default it is the maximum extent of the geometry,but can be replaced by somethings else if this is not relevant to the outlet length scale. The defaults are usually appropriate for all of these; we haven't used non-default values except for sanity checks. PRESSURE EXCITATION is somewhat different - it is used for the calculation of transfer matrices. You can add some pressure excitation at the outlet and then measure the corresponding pressure and velocity wave amplitudes at the inlet. This feature was implemented at the request for modelling acoustic behaviour in combustion chambers. If one is not interested in transfer matrices, then leave this unset. |
||
|