Skip to main content
P
ptc-2252261-Visitor
1-Visitor
March 15, 2013
Question

Genfit Function for Data Set

  • March 15, 2013
  • 4 replies
  • 2547 views

I'm trying to find a decay equation that I can use with the genfit function but can't seem to find anything that fits. I'm try to generate a function of alternating flux density vs incremental permeability for various DC Magnetization levels.

My question is twofold:

How can I generate a function that will fit the data for a single DC magnetization (Hdc) level?

Can I generate a function that will find incremental permeability at any Hdc and any alternating flux density?

I've included a file for review.

Thank you in advance for your help

Sergio Kraljic, Jr.

-

4 replies

F
Fred_Kohlhepp23-Emerald I
23-Emerald I
March 15, 2013

Curve Expert came back with a Gaussian, not great.

Might find a Weibull function.

P
ptc-2252261-VisitorAuthor
1-Visitor
March 15, 2013

Never heard of a Weibull function, but I'll look into it -- I'm not a mathemetician, just an engineer

Interestingly enough, the curve looks similar to Planck's radiation law, but unfortunately I couldn't adapt it to my particular application -- the constants were crazy.

Thanks for looking into this.

H
HarveyHensley12-Amethyst
12-Amethyst
March 16, 2013

I think you might try the gamma probability (not the gamma function alone) and the log-normal probability distribution functions as well. You can get good starting estimates of the two parameters in the distributions by calculating the mean and variance of the data. The horizontal and vertical values will have to be scaled to your data, so that adds two parameters.

Message was edited by: Harvey Hensley

A
AlanStevens19-Tanzanite
19-Tanzanite
March 16, 2013

The attached shows reasonable (?) fits to individual curves, but I haven't attempted to develop the fits as a function of Hdc.

Alan

    P
    ptc-2252261-VisitorAuthor
    1-Visitor
    March 18, 2013

    Alan --

    The curves look great! Thank you...btw how & where did you come up with this function? I looked everywhere in my math references and couldn't find anything that would work as well as what you provided.

    A
    AlanStevens19-Tanzanite
    19-Tanzanite
    March 18, 2013

    Sergio Kraljic Jr. wrote:

    Alan --

    The curves look great! Thank you...btw how & where did you come up with this function? I looked everywhere in my math references and couldn't find anything that would work as well as what you provided.

    It's just the sum of two generalised Weibull-type curves - generalised by flinging six fitting constants at them! Not that clever really! I was hoping the constants would each be a simple function of Hdc, but I was unable to find one. This means the fits aren't much use for interpolating on Hdc. They're probably ok for interpolating on B, but, given their arbitrary nature, I suggest you don't extrapolate with them.

    BTW what is the physical system they represent?

    Alan

    H
    HarveyHensley12-Amethyst
    12-Amethyst
    March 17, 2013

    I tried the Gamma and Weibull distributions and they didn't work well. I also tried a polynomial with 6 terms (5th order) and it worked as well as the function Alan provided...in the region of the data. However, as expected, the polynomial didn't continue to decay for B > 2. Thus, if prediction beyond B = 2 is needed, Alan's function is the best.

    Terms & ConditionsAccessibility statement