Hi,I cam writing a code for compressible flow. In my code,I use the pressure-based method such as SIMPLE which has been extended to the compressible flow by including the density corrections. The code is orginally written for viscous flow by solving the N-S equations, while I can get the invisid results by setting the viscosity equal zero. Untill now I have finished the validation computations such as the flow in a nozzle and flow in a tunnel with an arc bump, and the results are in good agreement with the results of other authors. But when I turn to the flow around an airfoil, I find that it doesn't work.In the computation, I use the first-order up-wind algorithm for the convetion terms. After carefully debuging, I find that the coefficients of the discretized momentum equations are zero in some nodes near the wall in the staganation region which terminated the computation. I try to solve this problem by set the coefficient to be 1e-30 when it is zero, but it doesn't work. Because there is no diffusion term in the invisd solution,so in the stagnation region there is the possiblity for the coeefficents to be zero when the flow is reverse to the wall at the begining of the computation. I don't know how to deal with this problem, so I wonder if anybody can give me some suggestions. I will be very grateful for any suggestions. Thanks!!

Computer arithmetic is not the same as human arithmetic, when you set the viscosity coefficient equal to zero, in a Navier-Stokes solver, you assume that all viscous dissipation terms will be equal to zero, but the computer programme, specially when using FORTRAN, still computes these terms and at the end of a computation cycle, these terms that you assumed equal to zero, are not. Either you use Double precision math or higher (If using the C language) or modify your code to solve Eulers equation. If you have access to a commercial programme, repeat your calculations setting mu equal to zero and compare the results. Feed back on this point is important