Dear Sebastien,

continue to the last question, On SHEN's paper(J.of Comp.Phys.1998) the scheme which using the ( 3*u^(n+1)-4*n^n+u^(n-1))/2dt for the time derivative ,he said that the linear parabolic operator(Laplace operator)is approximated by a second-order backward scheme which appears to be more stable than the crank-nicholson scheme ,I donot know how he deal with the viscocity term,I didnot mean that the viscosity is varable,in generally,the central difference is used for the viscosity term.but on his paper,it seems not,

Thanks

1) If you post somethong here, than don't say TO Sebastien. This a question open to everybody.

2) In this article, Shen use a spectral method for "space discratisation".

3) You told me you used a FDM method. Than a central difference would be appropriate for the spacial discretisation of the second order terms (viscosity terms).

4) I don't want to be rude here. But, your questions seem to show a lack of experienced with numerical methods. Therefore, I don't think it is a good idea to construct your own NS solver. Since your project is to simulate the flow around a bridge and I beleive NOT to develop a new NS solver, I suggest using an existing code. As an exemple, you could use the Featflow code available at http://www.featflow.de . The source code is available and it is very well documented. Than you could modify the code to fit your own needs (i.e introduce the movement of the bridge).

Good luck.