A better way to compute drag?
 

i've tried using a 2d structured ns solver to simulate flow past an airfoil. lift and pressure coefficient compare well with other ppl's results. however, the drag values are not too good.

i've computed my drag by resolving the velocities normal & tangential to the surface, then tau(nt)= miu(du(t)/dn + du(n)/dt)

Just to confirm one thing though ... du(n)/dt, tau(nn) & tau(tt) are all 0, right?.

du(t)/dn is then approximated using 1 sided finite differnce.

I've read some paper which talks abt computing drag from the fluid flow perspective (as compared to body). Is that more accurate? Any papers?

Someone also suggested using least square mtd to compute the derivatives. However, the one my lecturer suggested is not accurate when the nodes are all above or below the point of interest. Does anyone has a better paper to recommend?

Thanks

Re: A better way to compute drag?
 

Hi!

I had tried to compute drag of flow over a circular cylinder and had found the problems about sign convention in polar coordinate and accuracy. The procedures were similar to yours. I didn't sure that my old computations are right or wrong.

But I have found the fact that 'the derivatives computed from difference formula have low accuracy'.

I suggest you to estimate the accuracy of your difference formula by (1) finding the exact solution of 2D heat transfer of a rectangular plate with heat generation from textbook or handbook and (2) balancing the thermal energy by using heat flux obtained from difference formula. You may find something surprised.

Now, I use weighted residual FEM to solve laminar convection problems. In the processes of deriving element equations, the Gauss' theorem is performing. This invokes the physical quantities at the boundaries into equations and I can find the physical quantities such as heat flux as a part of solutions or substituting the solutions into the relevant formula.

Drag would be similar. Accuracy may be improved.

I think most of authors concentrate to the methods of discretization too much. Sometimes, they try to explain the physics of phenomena but they never reach the better way.