curve fitting a charging capacitor type curve

[mcgradys]

Hi all,

Newbie to this forum, but need some excel help. I have collected a
bunch of data from an object as it heats up. The curve is very similar
to a capacitor charging current curve. In my case, manually trying to
approximate this curve, I come up with a formula of y=
(rise)*(1-exp(-time))+ambient. Is there a way in excel to fit a curve
like this or do I need an aftermarket solution? I need something that
will use the data points and perform calculations to determine the
curve.

Thanks,

Sean

 

[Bill Martin]

Hi all,

Newbie to this forum, but need some excel help. I have collected a
bunch of data from an object as it heats up. The curve is very similar
to a capacitor charging current curve. In my case, manually trying to
approximate this curve, I come up with a formula of y=
(rise)*(1-exp(-time))+ambient. Is there a way in excel to fit a curve
like this or do I need an aftermarket solution? I need something that
will use the data points and perform calculations to determine the
curve.

Thanks,

Sean

Have you tried to use the Solver? Set up a table with a column of empirical
data, a column of calculated data using your equation, and a column with the
square of the error between the two with a sum at the bottom. The calculated
data needs to be calculated using spreadsheet cells to store each of the
parameters you're trying to fit.

Then ask Tools/Solver to minimize the sum of the column of squared errors by
varying the paramater value cells.

I suspect Solver will work pretty well with this kind of problem.

Good luck...

Bill

 

[Guest]

Sean, check out Chartool & Smooth_Curve_Bezier the smart guys at XLRotor have
developed, it leverages Excel's native chart smoothing capability to fit the
curve.
It's free too.
http://xlrotor.com/resources/files.shtml
Cheers...Terry

 

[Herbert Seidenberg]

Assuming your data is in xvalu and yvalu
xvalu yvalu yvalu2 yvalu3 k_ast k_diff k_rc k_amp
0 0.0104 2.5896 0.0000 2.6000 -0.0273 -0.2057 2.6273
1 0.4459 2.1541 0.4885
2 0.8011 1.7989 0.8862
3 1.1477 1.4523 1.2100
4 1.3927 1.2073 1.4735
5 1.6954 0.9046 1.6881
6 1.9086 0.6914 1.8627
7 1.9831 0.6169 2.0049
8 2.0567 0.5433 2.1206
9 2.2272 0.3728 2.2148
10 2.2512 0.3488 2.2915
11 2.3140 0.2860 2.3540

Name all columns.
Guess what your asymptote is and enter into k_ast.
Enter into the yvalu2 column
=k_ast-yvalu
Enter into k_rc, k_amp, k_diff respectively
=LN(INDEX(LOGEST(yvalu2,xvalu),1))
=INDEX(LOGEST(yvalu2,xvalu),2)
=k_ast-k_amp
Do a Goal Seek to set k_diff to zero by changing k_ast
Your best fit curve is in yvalu3
=k_amp*(1-EXP(k_rc*xvalu))

 

[Jerry W. Lewis]

if you know time and are trying to fit rise and ambient, then the
function is linear in the unknowns and you can simply use SLOPE and
INTERCEPT (or LINEST, if you prefer). In that case, your X column for
fitting would be the calculated value =1-exp(-time)

Jerry