Page 319 - Computational Fluid Dynamics for Engineers
P. 319

10.7  Boundary  Conditions  for the Euler  Equations                  309


                     2    2
         where  q = Vu  + v ,  which  can also be written as
                                        2
                                                 2
                                  (a 2  -  q )cp^  + a (f vr]  = 0         (10.6.4)
         with
                                     2                 2
                                  =  u <p xx  +  2uv(p xy  +  v <Py Xy     CIO 6 5)

         and
                                     2
                                                        2
                                  _  V <Pxx  ~  2uV(f xy  +  U (p yxy      .     ,
                              (fr]r] —          2                          ^lU.O.DJ
         When  a 2  > q 2  (subsonic  flow)  is present,  central  differences  are used  just  as
                                                 2
         in the elliptic  TSD case.  However,  if a 2  < q , then  upwinding  is more  complex
         than in the hyperbolic TSD case. Introducing the forward,  backward  and central
         differencing  operators  A,  V,  <5, respectively, the discretized  form  of Eq.  (10.6.4)
         can  be written as

                                                      2
                             2
                  ,  2 _  2\  (U V XV X  +  2uvV xVy  +  V \7yVy\
                  K a   Q ) \              2               I  Wj
                          /                            \                   (10.6.7)
                             2
                          /  ^ V x 4  +      +  2       \     _
                        2             2uv8 x8 y  u V yA y
         where the first  term  represent  upwinding, and the second term  central  differenc-
         ing. When  u < 0, then  upwinding  is achieved  with  forward-differencing  (A XA X).
         The  same  applies to the ^-derivatives  if v < 0.
            The solution  of the conservative  full potential equation  requires the introduc-
         tion  of explicit  artificial  dissipation,  as discussed  by Jameson  [4]. Alternatively,
         the  lagged  density  approach  developed  by Hoist  [5] provides the necessary up-
         winding  in the supersonic  region  while  simplifying  the  implementation.  The
         reader  is referred  to the original  papers  for details  on the subject.



         10.7  Boundary    Conditions    for the Euler   Equations

         The  mathematical  formulation  of the boundary  conditions  for the Euler  equa-
         tions  is more  complex  than  the scalar  wave  equation  or even  the TSD or  po-
         tential  equation,  since  there are now more than  one wave  travelling  along  char-
         acteristics.  The eigenvalues  Ai, A2, A3 of the one-dimensional  Euler  equations
         are  respectively  u,  u +  c and u  — c,  as  discussed  in  Section  5.1. For  super-
         sonic  flows  (u > c), all waves  travel  in the same  direction  (u > 0,  u +  c > 0,
         u  — c > 0) whereas  for subsonic  flows,  two waves  are travelling  in the flow di-
         rection  (u > 0, u + c > 0) while the third  one travels  against  the flow  direction
         (u-c<0)    (Fig. 10.11)
            Consistent  boundary  conditions  must  be  imposed  for  both  types  of  con-
         ditions.  If the supersonic  waves  of Fig.  10.11a  travel  into  the  computational
   314   315   316   317   318   319   320   321   322   323   324