CFD Online Logo CFD Online URL
www.cfd-online.com
[Sponsors]
Home > Forums > Main CFD Forum

help!

Register Blogs Members List Search Today's Posts Mark Forums Read

Reply
 
LinkBack Thread Tools Display Modes
Old   March 9, 2004, 23:45
Default help!
  #1
lsm
Guest
 
Posts: n/a
who can tell me why the "middle point rule" in FVM is 2th precision?

  Reply With Quote

Old   March 10, 2004, 01:49
Default Re: help!
  #2
sunil
Guest
 
Posts: n/a
jgjkagdxjagj
  Reply With Quote

Old   March 10, 2004, 05:15
Default Re: help!
  #3
P. Birken
Guest
 
Posts: n/a
Any textbook or lecture notes on numerical methods for ODEs. Or a Taylor-expansion.
  Reply With Quote

Old   March 10, 2004, 06:16
Default Re: help!
  #4
lsm
Guest
 
Posts: n/a
sorry,middle value rule and middle point rule are popular,but i can't see any analysis about precision.
  Reply With Quote

Old   March 10, 2004, 07:33
Default Re: help!
  #5
Tom
Guest
 
Posts: n/a
It follows from the mean-value theorem (=> Taylors theorem) in any school (A-level in the UK) maths text book.

Alternatively consider two points x_l and x_r with midpoint x_m = ( x_l + x_r )/2 and spacing h = x_r - x_l then the trapezoidal rule gives, for a function f,

integral f = h( f_r + f_l )/2 + O(h^3)

and by Taylors theorem f_m = ( f_r + f_l )/2 + O(h^2) which imples the result (assuming you accept the Trapezoidal rule is second order!)

  Reply With Quote

Old   March 10, 2004, 09:38
Default Re: help!
  #6
lsm
Guest
 
Posts: n/a
yes integral f = f(average value at one point between x_l and x_r)*h=?f_middle-pint*h+o(h^3)
  Reply With Quote

Old   March 10, 2004, 23:42
Default Re: help!
  #7
lsm
Guest
 
Posts: n/a
my puzzle is that f = f(average value at one point between x_l and x_r) by middle value rule is strict, but f(average value at one point between x_l and x_r)=?f_middle-pint+o(h^1) is 1th precise. how to explain it?
  Reply With Quote

Old   March 11, 2004, 04:40
Default Re: help!
  #8
Tom
Guest
 
Posts: n/a
Basically the O(h) error term vanishes at the midpoint - look at a book on finite difference methods.
  Reply With Quote

Old   March 11, 2004, 11:47
Default Re: help!
  #9
Hrvoje Jasak
Guest
 
Posts: n/a
Do integral of phi over a ceontrol volume and approximate it with Taylor series expansion.

int_V phi(x) dV

phi(x) = phi_P + (x - x_P) . grad(phi)_P + hot. (1)

When you integrate you get

int_V phi(x) dV = phi_P V + (int_V (x-x_P) dV) . grad(phi)_P + hot.

The gradient comes out of the second integral because it is a constant. So, if the integral

int_V (x-x_P) dV = 0 (3)

the method will be second order, which is true if the point of expansion you are expressing phi in eq(1) around in the centroid of the cell. Fortunately, eq(3) is indeed the definition of the centroid so all is well.

"... and there you are!" (quote from My Big Fat Greek Wedding)

Hrv
  Reply With Quote

Old   March 11, 2004, 23:59
Default Re: help!
  #10
lsm
Guest
 
Posts: n/a
thanks,but excuse me, let's do 1d case for 1d case your (3) equation int_V (x-x_P) dV = 0 (3) become int_V (x-x_P) dx = 1/2(x-xp)^2 it is o(2), not o(3)
  Reply With Quote

Reply

Thread Tools
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are On
Pingbacks are On
Refbacks are On



All times are GMT -4. The time now is 11:51.