Page 143 - Introduction to Computational Fluid Dynamics
P. 143

P1: IWV
                                                                                                12:28
                                                                                   May 20, 2005
                                        0 521 85326 5
                           CB908/Date
            0521853265c05
                     122
                                                                 2D CONVECTION – CARTESIAN GRIDS
                               Special care is, however, needed for the mass residual. On staggered grids, the
                            mass residual R m is checked via Equation 5.30 [51]. That is,
                                                                        0.5

                                                                      2
                                                     R m =        ( ˙ m P )  .                 (5.72)
                                                            all nodes
                            However,oncollocatedgrids,onecannotusethisequationdirectlybecause ˙ m i, j  = 0
                            even at convergence. Therefore, Equation 5.72 is written as
                                                ⎡                                 ⎤ 0.5
                                                                                2


                                          R m =  ⎣       AP p     −     A k p     ⎦   ,        (5.73)
                                                              m,i, j        m,k
                                                  all nodes           k
                            where AP and A k are coefficients of the pressure-correction equation. It will
                            be recognized that this equation simply represents the discretised version of the
                            left-hand side of Equation 5.32 (or see Equation 5.28 with ˙ m R = 0). Thus, R m is
                            evaluated after p    is recovered in step 4 of the calculation procedure. This is
                                           m,i, j
                            an important departure from the staggered-grid practice that a casual reader may
                            overlook.

                            5.3.3 Underrelaxation

                            Global Relaxation
                            As mentioned in Chapter 2, in steady-state problems ( t →∞), underrelaxation
                            is effected by augmenting Su and Sp as
                                                                             (1 − α)
                                                l
                              Su i, j = Su i, j + B   ,  Sp i, j = Sp i, j + B,  B =  (AP i, j + Sp i, j ),
                                                i, j
                                                                                α
                                                                                               (5.74)
                            where α is the underrelaxation factor and l is the iteration level. The value of α is
                            the same for all nodes but it may be different for different  s. This is called global,
                            or constant, underrelaxation.


                            False Transient
                            In multidimensional problems, underrelaxation is often effected in another way.
                            Thus, consider a steady-state problem in which  t =∞ and, therefore, the transient
                            term is zero. However, one can imagine that the steady state is achieved following
                            a transient and each time step is likened to a change in iteration level by one. In
                            this case,   o  may be viewed as   l  and the time step  t as the false-transient
                                       i, j                 i, j
                            step. Then, combining Equation 5.65 with Equation 5.74, we can deduce that the
                            resulting equation may be viewed as one in which

                                                              AP i, j + Sp i, j
                                              α eff,i, j =                        ,            (5.75)
                                                                       o
                                                      AP i, j + Sp i, j + (ρ  V/ t) i, j
   138   139   140   141   142   143   144   145   146   147   148