Smooth a graph in Processing

cygig

I am making a program to generate a battery discharge graph for accessing the capacity of batteries (because Excel is too slow and troublesome).

I average a group of values together to make my graph smoother, and I am wondering if there are other methods to draw smoother graphs in Processing. I heard of methods like regression and best-fit curves, but I never knew how they can be implemented in Processing, and to be honest, I am math ability is abysmal.

Any input and advices will be welcome!

Here is my graph:

Here is the one generated by the electronic dummy load software:

I would prefer to generate my own graph to control the design and be able to output in a vector file.

And here is a part of my code:

String fname = "Trustfire 14500 900mAh 0.5A.csv";
CellData cellData = new CellData();
float[] voltage;
float[] capacity;

void setup(){
  size(800, 600);
  background(255);
  cellData.readCSV(fname);

  voltage = cellData.toMappedArray("voltage", 550, 50);
  voltage = bandAvg(voltage, 50);

  capacity = cellData.toMappedArray("capacity", 50, 750);
  capacity = bandAvg(capacity, 50);


  beginShape();
  for (int i=0; i<voltage.length; i++){
    vertex(capacity[i],voltage[i]);    
  }
  endShape();


}

void draw(){
}

kfrajer

For drawing a line, you can check bezier curves:

You can use check
https://forum.processing.org/two/search?Search=bezier

For example:
https://forum.processing.org/two/discussion/21355/fixed-make-bezier-curve-pass-through-control-points/p1
https://forum.processing.org/two/discussion/comment/91900/#Comment_91900
https://forum.processing.org/two/discussion/comment/92127/#Comment_92127

However for smoothing, you can try any option available in the web. For example: https://en.wikipedia.org/wiki/Smoothing

I am not sure if Processing has any library that would do something like that. For best fit, you need to know beforehand the behavior of the discharge of your battery. As you see, your battery discharges almost linearly from 0 to 660 and then afterward it drops faster. For a best fit, you need to know a function what will have that shape and that you can controlled with a finite number of parameters. In theory, it is possible to try to characterize the discharge of a battery. In practice, it depends on the age of the battery, charge and discharge cycles it went through, quality of battery materials, etc. etc. etc. It is lots of work and for smoothing I will just focus on a windows averaging algorithm which can be implemented easily in Java.

Kf

jeremydouglass

@cygig -- what kind of data resolution are you smoothing? How many sample values along the x axis are you trying to fit a curve to?

My first recommendation would be to look at the Curves Tutorial.

• https://processing.org/tutorials/curves/

Option 1: Spline. In particular look at its example of splines drawn with curveVertex() -- these can be your data points.

noFill();
beginShape();
curveVertex(10, 10); // the first control point
curveVertex(10, 10); // is also the start point of curve
curveVertex(40, 40); // ...point
curveVertex(80, 60); // ...point
curveVertex(90, 90); // the last point of curve
curveVertex(90, 90); // is also the last control point
endShape();

Option 2: Bézier. Depending on your data resolution, you could also choose to render bezier curves through sets of your points, although this is trickier -- you need to find the control points to fit the curve to your data points. There is a recent discussion of how to do this for four points here:

• https://forum.processing.org/two/discussion/21355/fixed-make-bezier-curve-pass-through-control-points

...and I think the method could be extended to n points. If you are interested in digging in to these kinds of approaches then check out the very interesting ArcLengthParameterization example sketch.

kfrajer

Only if you are interested, example analysis on Li-ion batteries for example

http://www-personal.umich.edu/~hpeng/DSCC2013_Weng.pdf
http://www.mdpi.com/1996-1073/8/4/2889/pdf

Kf

cygig

Awesome guys! Good to have such a helpful community for beginners like me.

prince_polka

I would prefer to generate my own graph to control the design and be able to output in a vector file
If by vector file you mean svg then why don't you generate them in inkscape directly?
Otherwise if you need a formula you could enter a few strategic coords into mycurvefit.com and try out which one works best. Also I saw this related question trending on reddit just now.
https://reddit.com/r/askscience/comments/63suuf/is_there_a_way_to_find_the_equation_of_a_random/ It should ofcourse be possible to implement any of those algorithms in processing, and has probably already been done for a few.
A phamtasmagorical solution I just thought of is turning the points into a sound. (don't ask how) and feed that sound to the FFT analysis tool of the sound library, apply lowpass and convert into an appropriate format... I know it's dumb in practice, but in principle it should work right?