|
[Sponsors] |
July 18, 2002, 06:27 |
L1 and L-infinite error
|
#1 |
Guest
Posts: n/a
|
Hi everyone. I am trying to test the accuracy of several schemes with a linear advection test problem. And I want to get eL1 errorf and eL-infinite errorf by comparing numerical solution with analytical one. Can anyone teach me the definition of L1 and L-infinite error. Thanks.
|
|
July 18, 2002, 14:39 |
Re: L1 and L-infinite error
|
#2 |
Guest
Posts: n/a
|
L_1 = sum |exact - numerical| from i=1 to N L_infinity = max |exact - numerical| from i=1 to N
If you want to compare with different grids, then multiply grid size in the front and take of space dimentions. Kim |
|
July 19, 2002, 12:56 |
Re: L1 and L-infinity error
|
#3 |
Guest
Posts: n/a
|
Thanks for your reply. What I want to do is convergence study of high order schemes. All of such papers, which I have, show that L_1 error is always smaller than that of L_infinity. But the definition that you wrote seems to be opposite to the results. Maybe L_1 = (sum |exact - numerical|) / (gridsize) is right ??? Thanks
Kenshi |
|
July 19, 2002, 13:19 |
Re: L1 and L-infinity error
|
#4 |
Guest
Posts: n/a
|
Yes, you are right.
Some authors divide sum by grid size, some are not becuase if you want to calculate error on single grid size, it doesn't matter, but as you told before, if you want to calculate convergence rate, you need to divide the sum by mesh size so that you can compare the error on the different mesh sizes. Kim |
|
August 7, 2002, 07:37 |
Re: L1 and L-infinity error
|
#5 |
Guest
Posts: n/a
|
L1=sum(abs(Ri)) L-infinity=max(Ri) ; i=1,2,3,.. where R=residual
|
|
|
|