Page 261 - Fundamentals of The Finite Element Method for Heat and Fluid Flow
P. 261

CONVECTION IN POROUS MEDIA
                        The Step 2 pressure calculation becomes
                                       1 ∂ 2   n+θ      1      1      C   |V|     ∂ ˜u i      253
                                            (p )   =       +       + √                      (8.59)
                                         ∂x 2           t    ReDa      Da   3/2  ∂x i
                                          i
                        Step 3 is also different and is given as

                                          1      1      C   |V|   n+1
                                             +       + √         u i  =
                                          t    ReDa      Da   3/2
                                              1      1      C   |V|       1 (∂p )
                                                                                n+θ
                                                 +       + √         ˜ u i −                (8.60)
                                              t    ReDa      Da   3/2       ∂x i
                        Although extra complications were introduced in the semi-implicit form at Step 1 for steady
                        state solutions, we can avoid simultaneous solution of the algebraic equations by taking
                        the coefficient

                                                     1      1       C   |V|
                                            CO =        +       + √     3/2                 (8.61)
                                                     t    ReDa      Da
                        on to the RHS. Thus, the system can be enabled for the mass lumping procedure (Nithiarasu
                        and Ravindran 1998) when discretized in space. The final matrix form of the three steps are

                        Step 1: Intermediate velocity calculation
                        x 1 momentum component

                                                 {u 1 }   −1                           n
                                  [M p ]{˜ u 1 }= [M p ]  + CO  −[C p ]{u 1 }− [K p ]{u 1 }+ {f 1 }  (8.62)
                                                   t
                        x 2 momentum component

                                                 {u 2 }   −1                           n
                                  [M p ]{˜ u 2 }= [M p ]  + CO  −[C p ]{u 2 }− [K p ]{u 2 }+ {f 2 }  (8.63)
                                                   t
                        Step 2: Pressure field

                                             n+1    CO                             n
                                      [K p1 ]{p}  =−     [G p1 ]{˜ u 1 }+ [G p2 ]{˜ u 2 }− {f 3 }  (8.64)
                                                      t
                        Step 3: Momentum correction
                                                 n+1                −1       n+1
                                         [M p ]{u 1 }  = [M p ]{˜ u 1 }− CO  [G p1 ]{p}
                                                 n+1                −1       n+1
                                         [M p ]{u 2 }  = [M p ]{˜ u 2 }− CO  [G p2 ]{p}     (8.65)
                        The quasi-implicit (QI) form is very similar to that of the above scheme but now the
                        viscous, second-order terms are also treated implicitly (θ 2 = 1) (Nithiarasu et al. 1997).
                        The important difference, however, is that the quasi-implicit scheme does not benefit from
                        mass lumping when solving for the intermediate velocity values. A simultaneous solution
                        of the LHS matrices is essential here. It has been proven that both the QI and SI schemes
                        generally perform well (Nithiarasu 2001).
   256   257   258   259   260   261   262   263   264   265   266