Thread: How to compute this weird derivative?

Hello,

I am working on numerical solution of incompressible fluid dynamics problem. I need to compute the Jacobian of a viscous flux. I have to deduct some derivatives like this



U, X are independent to each other.

Some people say the result is absolutely not zero, but I still could not find an analytical answer. I wonder this will turn out into a derivative operator rather than an expression of fixed value. If an analytical expression is not possible, how to handle this in a finite difference method?


Thank you and have a good day!

Bearcat

Originally Posted by bearcat

Hello,

I am working on numerical solution of incompressible fluid dynamics problem. I need to compute the Jacobian of a viscous flux. I have to deduct some derivatives like this





Some people say the result is absolutely not zero, but I still could not find an analytical answer. I wonder this will turn out into a derivative operator rather than an expression of fixed value. If an analytical expression is not possible, how to handle this in a finite difference method?


Thank you and have a good day!

Bearcat

I've seen something like this before...

What you want to do is:

(1)Take the partial of U w.r.t. x. Denote the result by L.

(2)Transform L so that it is in terms of U. (This is the hard part)

(3)Take the partial of Transform(L) w.r.t. U.

Can you post what U is?

In my case, U is the unknown velocity in N-S equation, x is the coordinate. they are independent to each other. What I am looking for is an analytical expression for my coding. I am sure it's solvable in some way as I am not the first to encounter this problem.

I suspect it's possibly .