How can I improve the attached code to reduce the gap between fitted curve and the original ones?

Hello friends

I have a code that I want to know the amplitude of Vabc between t=0.1 and 0.2

But the fitted curve is a bit different from the original ones.

Can you advise?

[Mathieu NOE]

hi

why the need to do a fit to measure a sine wave amplitude? this seems overkill IMHO

[Atefeh]

Hi Mathieu

I need the amplitude value in order to use in a machine learning process, the output waveform is extracted from a simulink model.

[Mathieu NOE]

hello

why not simply use max / min of your data ?

Answer 1

[Sandeep Mishra]

Hi Atefeh,

I ran the provided code in MATLAB R2024b and identified the same disparity between the curve plotted using the data and the curve obtained through curve fitting.

Upon debugging, I noticed that the code uses the ‘fit’ function across three time ranges: ‘ta’, ‘tb’, and ‘tc’. The RMSE for the ‘tb’ range is approximately 0.2, whereas for ‘ta’ and ‘tc’, it is significantly lower at 1e-6.

The discrepancy arises due to the high RMSE for the ‘tb’ time range. This occurs because the ‘fit’ function applies the ‘sin1’ ‘fittype’ parameter to data comprising a linear segment (from 0.1s to 0.2s) followed by a sine wave (from 0.2s to 0.21s), leading to inconsistent results.

To resolve the issue, you can update the time range of ‘tb’ and ‘tc’ interval to focus on a single curve type at a time.

Below is an example modification:

% Split data for straight line 
tb = Vtime(1002:2000); 
Vb = Vdat1(1002:2000); 

% Split data for sine wave 
tc = Vtime(2001:3001); 
Vc = Vdat1(2001:3001); 

% Updated time interval for curve fitting 
t  = linspace(0, 0.3, 3001); 
t1 = 0.1; 
t2 = 0.2; 

Refer to the following MathWorks Documentation to learn about ‘fit’ function:

https://www.mathworks.com/help/releases/R2024b/curvefit/fit.html

I hope this helps you in resolving the query