LabVIEW
how to do a curve fit
All you are showing us is a prehistoric VI that does a way to complicated read from a binary file (which you have not attached yet!).
This VI is mostly obsolete! All that song and dance with FOR loops, type casting, string reversals, etc can be done with atomic operations in the LabVIEW version (9.0) that you have. Simply use read from binary fine, set the datatype to U16, and select the byte order (little vs big endian).
In any case, none of the code has anything to do with fitting.
Please do the following:
- run the program and read your binary file so the graph on the front panel contains the data you want to fit.
- make the current graph data the default.
- Delete everything but the graph, it is irrelevant for fitting.
- Save the VI under a new name.
- Attach the VI here.
@luminita wrote:
I can't make all the values as default as the data file is so big...
I attached a image of the two graphs.
Your data does not look like something suitable for a polynomial fit. Do you have a better model?
In any case, to show the general process of fitting, we don't need a file with so many points. Attach a data file where the data is decimated to about 2000 points, for example.
Sorry, I don't know what a masine is, but lets look at the two datasets you attached (image).
@luminita wrote:
I am a material scientist... Unfortunatelly I need just the data to interpret it. The data is from a masine. They have one program to aquier it from the device and this one is ploting it. Unfortunatelly I can't compare the stages if I don't do the curve fit.
OK, let's do the scientific approach.
On the left we have some heavily distorted square wave fragment with very few features. In any case, you have way too many very noisy data points to describe the curve. Clearly, it could be filtered and decimated to a few hundred points without loss in real information.
Maybe you can tell us what kind of information you want to extract using a fit.
A polynomial fit is certainly inappropriate and you need quite a high order to approximate the curve. Still, the polynomial coefficient would have very little scientific meaning (or no meaning at all).
All I can say is that the data was recorded with way too many data points and way too little filtering.
The data on the right is highly oscillatory and again I see no reasonable way to fit this unless you have an appropriate mathematical model. Maybe an FFT would show some characteristic frequencies that could mean something.. or not.
So, again what does a "stage comparison" include? Can you explain in more details what you actually need to do with all that? What kind of parameters are you trying to extract from the raw data? What is the relevance of the two images?
