Skip to main content
R
RantEng14-Alexandrite
14-Alexandrite
April 17, 2024
Question

Derivative of Vector data with smoothing/fitting

  • April 17, 2024
  • 1 reply
  • 1676 views

I have been performing drop testing with a product.  Collected data shows displacement during the freefall and deceleration to a stop.  I've tried several methods of data smoothing and curve fitting.  The best I have achieved gives incorrect initial velocity and acceleration. 

 

In the image below, after smoothing (blue line), it shows an initial velocity rather an acceleration from zero.  Any ideas on how to accomplish my goal would be appreciated.  See the attached Mathcad Prime 9 document.

RantEng_0-1713364455148.png

 

Thanks,

 

Bob

 

1 reply

W
Werner_E25-Diamond I
25-Diamond I
April 17, 2024

How should supersmooth know that you want a curve with a zero slope at t=0?

From the given data (red) it really does not look that way:

Werner_E_0-1713365796255.png

 

If you know the type of equation you expect for drop(t),  you could try a non-linear regression to find its best fitting parameters using the measured data.

 

Just a question because I am confused: When I look at the Excel component

Werner_E_1-1713365844118.png

I can't see that you export column A to a variable "Time", Nonetheless variable "Time" with the correct unit seconds is defined after the Excel component. I wonder where an assignment like

Werner_E_2-1713365954239.png

which obviously must be present would hide ?? I was not able to find it even after resizing the component.

 

 

 

R
RantEng14-AlexandriteAuthor
14-Alexandrite
April 17, 2024

Hi Werner,  I was certain you'd respond.

 

Time is in the output of the spreadsheet.  Not sure why you don't see it.

 

RantEng_0-1713366348508.png

 

It is always a challenge to make recorded data reflect reality.  In reality, the drop starts from zero distance and proceeds.  Noisy data makes it a challenge.  With a pencil and paper, it is easy to draw lines over the data to make the curves.  The challenge is getting the computer to understand what you eye sees intuitively.  

 

Supersmooth does a nice job of making the line my eye sees.  Unfortunately, I'm not being successful in taking that to the next step of deriving velocity and acceleration cleanly.  I don't know if there is a better method of normalizing and fitting the data.

 

 

W
Werner_E25-Diamond I
25-Diamond I
April 17, 2024

Crazy, I knew that that definition for Time in the Output area must be somewhere, but I am still not able to make it visible.

 

According smoothing I can't think of a suitable solution.

 

Best approach may be a curve fit but this would require you to be able to specify a function type to which the process corresponds.

 

Terms & ConditionsAccessibility statement