I found very useful experimental curve in literature, and I want to digitize it and feed into my CFD code. Does anybody know how to do it? Please give me some info and hint on it1

Thanks ahead!

Try to read some (x,y) coordinates manually from the curve. Once you have the coordinates, fit an orthogonal polynomial (i.e. discrete fourier, Chebyshev (rescaled to [-1,1], or other polynomial) using a collocation (gridpoint) or spectral coefficient approach. This is low tech, but may be your most efficient method.

thanks, Tom.

Can you give me more detail on it. I mean the tools, such as what software (matlab, tecplot, origin) is best to fit the selected data. And what is the collocation (gridpoint) or spectral coefficient approach.

Thanks a lot for your reply1

I always use the Engague Digitizer - a free tool which allows you to digitize a scanned plot. You can download it from here:

http://digitizer.sourceforge.net/

It works really well and has lost of nice features.

thanks a lot, Jonas, that's really helpful.

I use the same one too.

Hi, dear all,

After digitizing the curve, a set of data is obtained. If I want to fit them, such as lnear fit, polynomial fit, even a distribution function, what software do you recommend? I used Origin before, and the fitting function is too simple. Can you give me any suggestion?

Thanks !

The best I have used is Xact (www.scilab.de) but it is not really cheap.

Origin Lab is very Good. There are lots of Curve fitting options from simple to complex. You have to experiment which curve fits best. Dont just try the default linear ones. Try the non-linear ones and there are hundred and hundreds of options there. Also you can use TableCurve 2D.

Matlab is my tool of choice.

Then, could you please tell me more on how to do it. I only know the polyfit function for polynomial fit. Is there any other functions, in which toolbox? Is it flexible to use?

Thanks again!

This is difficult for me to answer, because it is problem dependent. If it is 2-D data I will first us the plot function to see what it looks like. After this step I will decide whether I need a hyperbolic, exponential, polynomial, or some other fit.

Matlab has a section on regression and curve fitting in addition to their curve fitting, statistic, and optimization tool boxes that are very helpful for curve fitting tasks. Good luck.

I like PlotDigitizer better than Engauge, but either program will work. PlotDigitizer is a Java program and works on various platforms.

http://plotdigitizer.sourceforge.net/

Nonlinear curve fitting is very powerful, as you can fit your data to any function that you specify (with a number of free coefficients). I like using xmgrace for this purpose. Like most other good programs, it's open source.

http://plasma-gate.weizmann.ac.il/Grace/