home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: sci.math
- Path: sparky!uunet!elroy.jpl.nasa.gov!sdd.hp.com!caen!destroyer!ubc-cs!unixg.ubc.ca!unixg.ubc.ca!jjpan
- From: jjpan@unixg.ubc.ca (James Jun Pan)
- Subject: Help! Numerical Analysis? ???
- Message-ID: <jjpan.716596103@unixg.ubc.ca>
- Summary: Looking for a solution to a numerical analysis probelm. Find
- Keywords: Numerical analysis, approximation, fit, software, routine
- Sender: news@unixg.ubc.ca (Usenet News Maintenance)
- Nntp-Posting-Host: unixg.ubc.ca
- Organization: University of British Columbia, Vancouver, B.C., Canada
- Date: Tue, 15 Sep 1992 22:28:23 GMT
- Lines: 48
-
- Hi, there. I have come across the following problem in my work. I am
- a chemist so please be patient if it is too simple for you guys.
-
- In general, for an equation ,
-
- y + a/y + b*LN(y) == u + w*LN(x) (1)
-
- where a,b,,u, and w are constants, we want to find a approximate solution
- of y in the form
-
- y = A + B*LN(x) (2)
-
- for a given range of x, where x > c, u+w*LN(c)=0.
-
- This is mentioned in a paper and it is assumed to be trival so no explanation
- is given. I could only figure out one way to do this. The above problem is
- the same as to find the parameters A'and B' for
-
- A'*y + B'* = y + a/y +b*LN(y) (3)
-
- i.e. use a straight line to approximate the curve. And I had to do it
- by first plotting out the curve, fit the points to a straight line and
- find out the slope and intercept. The reults are not satisfactory and
- it heavily depends on how big the range of the ploted curve is.
-
- Is there a general way to solve the probelm? Is there a way to get a pair
- of A and B for a very large interval where the curve is "good" enough?
-
- In particular, though, I need a solution to
-
- y - m/y - n*LN(y) = l - p*LN(x) (4)
-
- where m,n,l,p are all positive real numbers in the range 0 to 10.
-
- Is there an established computation routine for this kind problem?
-
- E-mail to either of my e-mail address is fine.
-
- Thanks a lot.
-
- James Pan
-
-
- --
- ===========================================================================
- James Jun Pan E-mail: JJPAN@unixg.ubc.ca
- UBC & TRIUMF JJPAN@erich.triumf.ca
- Vancouver, BC Phone: (604)- 222-1047 (o)
-