# Should I discard part of the data (steel traction test)

I have to deal with an old hydraulic machine, recovered from a deposit. I have no knowledge on mechanical engineering. Only civil engineering.

The machine is for doing traction tests on steel bars (for example ribbed bars). The machine is old, from the 70’s or 80’s I guess.

It has a system for plotting the stress-strain/deformation chart on paper, but also has digital instrumentation added, which is being read by an old PC still running Windows XP, and it only provides the raw data from the instruments (as a text file).

The digital system seems to be added, I guess, in the 90’s, and comes with a handbook with some instructions (as for example, how to calibrate the machine), but no explanation on the internal workings.

I need to algorithmically process the raw data produced by the tests, to extract various parameters.

As I had been taught theoretically, the type of chart I should get, must look like this:

But off course, this is a real world physical test, so I get this kind of chart

I had been told that it starts with an accommodation phase which I should ignore, and I need to extract various information from the rest of the chart.

Now, here is a zoom of the “elastic phase”

I have no knowledge on why it behaves that way, but I conjecture (guess) that the traction force is increased in discrete steps, and between each force increase, the machine waits some amount of time for the steel bar slowly elongating until it reaches a stress-strain/deformation static equilibrium.

If that conjecture is correct, then I should isolate only the “static equilibrium” points to do calculations, and discard the other, because the other points are transient states, and the theoretical applications of the test are only for static behaviors (civil engineering applications). If the conjecture is correct, I should discard the transient states and retain only the static data for purposes of static mechanic applications of the traction test, which are the only theoretical applications of the test in civil engineering.

But off course, I have no knowledge on dynamic mechanics, and I may be making some huge mistake. I’m not a mechanical engineer.

So, the question is: Is the conjecture correct? Should I make some additional considerations?

 I would welcome if you can give me a book title, for reference, on the dynamical behavior of this type of machine.

• If you hope for details of the type of machine, then you need to provide all the data you can : make, model, serial numbers, capacity etc etc. Commented Apr 4, 2018 at 16:04
• @Solar Mike There is no data on the manufacturer, but it looks like an ad-hoc machine made in an university. Commented Apr 4, 2018 at 16:13
• how about some labels and units on the plots Commented Apr 5, 2018 at 3:18
• you appear to have swapped the axes in the zoomed plots..why? Commented Apr 5, 2018 at 3:30
• @agentp, thanks for pointing that out. I haven't noticed, now I'll need to amend my answer, perhaps scrap it
– ChP
Commented Apr 5, 2018 at 5:36

Edit after AgentP's comment:

As AgentP pointed out, you have swapped your stress-strain axes in the zoomed image. This fact alone refutes your conjecture... Your graph should look like this:

You see vertical increases in stress, and sudden jumps in strain.

A reason the steps might occur:

The steps in your data occur at very specific displacement/strain intervals, this says to me that the displacement read-out has a too low resolution compared to the force read-out.

What to do with the data:

First of all, this data and machine won't cut it if you would like to certify any materials.

If the resolution of the displacement read-out is too low, you would be better off using the points indicating a change in displacement. Judging by the spacing of the force read-out points, their resolution seems much higher, therefor, if you read a change in the displacement read-out, it is more likely that that point will be the most representative of the stress-strain relationship of the specimen at that point. Below is a picture illustrating:

(Pardon my Paint skills, I'm more of a CAD guy)

As for the accommodation zone, as niels nielsen suggests, extrapolate the elastic zone to the x-axis and disregard the data to the left of that line. Below is an image illustrating:

P.S. Why your conjecture would be wrong even if you hypothetically did not swap your axes:

The machine wouldn't apply a force on the bar and wait for the bar to elongate to "catch up".

Firstly, you would have to apply the force very quickly to see any significant effect, I don't think it would be possible in the elastic zone. And if that was even possible, your graph won't have steps with a horizontal plateau, but it will be a series of bumps, going up in force and elongation, and when the force is stopped being applied, you will see a decrease in stress as the bar elongates to catch up.

Properly built traction testers are built to be strain-controlled, i.e. the machine would have a PID controller or something similar that controls the force applied to ensure that the strain increases at a linear rate w.r.t. time. This reduces the effects of potentially catastrophic failure, especially when testing brittle materials and prevents the jacks from "running away" once the specimen yielded.

• Thank you all. I swapped the axis just because I'm retarded. It is obvious that the problem is a quantization error and has nothing to do with dynamic effects. I think that the real curve should cross closer to the middle between changes in deformation, but the half steps in deformation jumps seem to challenge that. Anyways, thank you all for your help. Have a reward: youtube.com/watch?v=875WSKDmVfc Commented Apr 5, 2018 at 12:50
• @zexot You're not retarded, mistakes happen, and after a while of working on something you tend to overlook them more and more. At first glance I haven't noticed it either and it was only brought to my attention after agentp mentioned it
– ChP
Commented Apr 5, 2018 at 12:59
• I have to disagree with the comment that this data is unusable. This is typical even on modern well calibrated equipment, if you "zoom in" tight enough you will see such data acquisition artifacts. Commented Apr 5, 2018 at 16:00
• @agentp, I might have worded it a bit strongly, I don't mean that the data is useless, but to certify materials, you'll need a calibration certificate for your testing equipment as well, which, judging by OP's post and comments doesn't have that. Also, if you have a look at the accommodation zone, it seems to only really straighten out at 40% of the yield strength, which is a bit excessive in my opinion. But I'm not a lab technician or materials tester, so please correct me if I am wrong
– ChP
Commented Apr 5, 2018 at 17:36

all tensile testers exhibit this accomodation range to a slight degree, but in your case the large accomodation range is most likely caused by worn parts in the drive mechanism that pulls on the bar- most likely the mechanical "grabbers" that dig into and engage the ends of the bar under test. this might be the reason that this machine was discarded in the first place. when analyzing the test data, the accomodation "foot" is cut from the data by shooting the straight portion of the elastic slope down to the x-axis and using the resulting intercept as the true "beginning" of the test.

But I recommend inspecting the gears inside the machinery and the teeth in the grabbing fixtures to determine if repair is possible.