Entire nominal stress/strain curve

I've made a calibration curve for stress vs displacement and found the relationship between them (ie how to calculate the stress given only displacement data).

I had originally planned to plot the strain data from the first test up to the last valid load measurement, then continue that curve at that load value using the calculated strain from the second test data. However I noticed that the Load vs Displacement data is very different between samples, thus I can't use data from the second test to continue the curve found from the first test.

Q: So are we supposed to get the entire stress strain curve from the second test data, where the first test was only needed for calibration and precision error (when compared to other tests)?

Nope, the first cylinder is what you are using to build the full curve (remember, we didn't take any strain data on the second cylinder).

If you have the function to get strain from displacement you are halfway to completing the last portion of the stress-strain curve. What you'll need to do is find the displacement where the strain equals 30% from your cylinder #1 30% strain file. Then, you can match that displacement up to the closest value of displacement you can find in the cylinder #2 file (so you will be starting at nearly the same point) and using your handy relationship, calculate the strain all the way out to failure (to the end of the cylinder #2 file). Then, if you use the first four files plus that last piece, you'll have the full stress-strain curve.

I plotted this yesterday using data we took in lab and it actually makes a rather nice looking curve without any filtering or other manipulation.

Oh okay that makes sense. I was trying to match the force instead of displacement, thanks.

Also a high frequency filter was mentioned in the lecture slides for smoothing the curve. Is there a resource on how to implement this?

I don't think that the professors uploaded any sort of resource for everyone to use, but I can make some personal suggestions:

There's the obvious 'brute force' approach of just removing some of the data points to smooth out the curve a bit, but that is probably going to be your worst option in terms of time consumption.

With Matlab, you have a number of digital filtering options. You can use the fast Fourier transform and then inverse Fourier transform to filter out frequency peaks of noise. You could also use a Butterworth filter to filter out high frequency noise (oftentimes digital signals will have noise around 50 or 60 Hz due to AC electromagnetic interference from devices powered through wall outlets and such).

If you aren't confident with designing a digital filter, then you could probably just use the Matlab 'smooth' command, or for the final curve that you put in the document, you could even do a piecewise curve-fit for each part just to get them looking nice and piece them all together (obviously, you would still use your actual measured data to determine all the brass properties you need, this would just be for looks for the 'client').