I have collected data points for a system that can be modeled by a third order transfer function (G= 1/(ax^3+ bx^2+cx^1+d)), at resonance. I would like to try and curve fit the data so I can analyze the experimental with the model but I cant figure out how to do it. I have tried the pwrfit and genfit commands but without luck. The plot of experimental data is included.
Solved! Go to Solution.
Please attach to the message (by using "Use advanced editor") your Mathcad worksheet.
This is easy to do. Genfit is one approach, but minerr is a better one. Please post a worksheet with the data and the equation for the model and I (or someone else if they beat me to it) will show you how to do it.
I have attached the data set and form of the equation.
Thank you, this helps immensely. How did you come up with your initial guesses though (vg)?
Anthony Roberts wrote:
Thank you, this helps immensely. How did you come up with your initial guesses though (vg)?
I do not know the physical meaning of your problem, so first I took the initial guesses for all parameters equal to 1. After that was the result of using the genfit again and used them as initial guesses.
Hello,
I tried using the results of the fitting function as coefficients, using the same fitting function, but the results do not match those of the original genfit command. The frequency responses are the same however the magnitude is markedly reduced andit is not obvious what could be wrong. I appreciate your help. The sheet is attached. I would have thought that since these were the extractged coefficients then I should be able to plug them directly into an equation of the same form and get the same results.
Mathcad computes with full precision and displays with adjustable precision.
If you raise the precision in V(s) the graph improves.
Thank you. This is very surprising to me about the function and that its Q is so large/resonant frequency is so specific.
Try please Mathcad server with this task (genfit):