3
\$\begingroup\$

I'm testing a load cell where the result drifts over time.

By weighing it down with a known weight over night, the result crept up by 10% in 18 hours. It's a completely linear increase, with consistent variations due to noise.

I'm using a HBM QuantumX MX840B and two pairs of strain gauges that form a full bridge. The cable has been perfectly still during the test. The strain gauges are 120 ohm and it's been set at a 5V exitation voltage. The load cell is of stainless steel and is quite massive, so I would assume heat dissipation shouldn't be too big of an issue, but I don't know.

Temperature is reasonably stable (it's indoor) and the full wheatstrone bridge would compensate for change, something we also tested with a heat gun.

I'll re-mount the strain gauges on the load cell and see if it gets better, but I'm not sure as to why it drifts, and especially in such a linear way, and advice would be helpful.

Edit: This is what I'm going to do in order to re-mount it properly, please give me input as to where your experience differs from this.

Physical installation of strain gauge:

  • Remove all strain gauges

  • Clean with isopropanol

  • Prepare the area with rough grinder if need be

  • Sand down to 320 grit with electric sander or grinder flappy disks

  • Slightly roughen the texture with a courser paper to increase adhesion

  • Clean with isopropanol again Put a small amount of HBM Z70 cyanoacrylate glue

  • Push the strain gauge down with my thumb Cover with suitable material (silicone or similar)

Results from load cell overnight testing:

This is from testing the first cell over night, it dips considerably. Initial test, first load cell

This is from the second cell, it has a much smaller variation. It makes a huge dip in the morning, something we have seen before. This is because the test lab has eight huge halogen lamps in the ceiling that draws a big current when starting, I'm certain those affect the results. What is concerning is that it does not recover back to it's previous state.

Latest test, second load cell

\$\endgroup\$
  • 1
    \$\begingroup\$ 208 mW may not seem like much but unless the gauges are temperature compensated then you will get errors. Show the data for the gauges used. \$\endgroup\$ – Andy aka Aug 8 '19 at 11:35
  • 1
    \$\begingroup\$ You presume it acts as temperature compensation? Rule out the possibility that there is insufficient compensation. \$\endgroup\$ – Andy aka Aug 8 '19 at 11:50
  • 2
    \$\begingroup\$ it also might be your ADC... I never figured it out, but I had done some work with HX711 boards, and they exhibited behavior like you describe under ostensibly static loading. \$\endgroup\$ – vicatcu Aug 8 '19 at 12:28
  • 1
    \$\begingroup\$ what is the temperature coefficient of the used steel and strain gauge? Did you use the same coeff? \$\endgroup\$ – Marko Buršič Aug 8 '19 at 12:35
  • 1
    \$\begingroup\$ @Marko that’s a good question, but wouldn’t such a discrepancy only affect the linear accuracy of the measurement, rather than make it drift while at a constant load? As it happens I think these old cells are simply calibrated with a tension machine and given a linear scale based on the actual measurements. \$\endgroup\$ – Vincent Vega Aug 9 '19 at 16:55
2
\$\begingroup\$

Mounting strain gauges is an art. Your long term variation seems to me that the gauges are mounted in a way that the adhesive is not keeping them in the same position over an extended period of time. At least this is what I would consider after making sure to rule out temperature and voltage variations.

| improve this answer | |
\$\endgroup\$
  • \$\begingroup\$ I agree that seems to be the most likely point, but I'd like to go through any sort of potential issue that could cause this problem. What I'm going to do is the following: Remove all strain gauges Clean with isopropanol Prepare the area with rough grinder if need be Sand down to 320 grit with electric sander or grinder flappy disks Slightly roughen the texture with a courser paper to increase adhesion Clean with isopropanol again Put a small amount of HBM Z70 cyanoacrylate glue Push the strain gauge down with my thumb Cover with suitable material (silicone or similar) \$\endgroup\$ – Vincent Vega Aug 8 '19 at 11:24
  • 1
    \$\begingroup\$ The easiest thing here is to repeat the tests multiple times. Adhesive creep under the strain gauges is most likely to occur with the tension direction and not in the compression direction....but who knows. I would repeat long term tests alternately with load and no load and cycle three or four time to see if some stability appears or not. \$\endgroup\$ – Michael Karas Aug 8 '19 at 11:30
  • \$\begingroup\$ Thanks, I'll try that, also with different load cells, in order to provide some consistent results. Right now I'm using a forklift, lifting the load cell and hanging a giant swivel under it as weight, with the forks being propped up by a beam in order to resist the creep of the hydraulic cylinder of the fork lifter. So I can only do one at a time, so this will take some weeks (and the fork lift). \$\endgroup\$ – Vincent Vega Aug 8 '19 at 11:34
  • 1
    \$\begingroup\$ @scott thanks, that was a very to-the-point manual. I will make sure to do it like that. \$\endgroup\$ – Vincent Vega Aug 9 '19 at 16:48
  • 1
    \$\begingroup\$ @KalleMP you mean hang them after each other? I haven’t really thought of that. I tried again with a new cable, and on my most recent test, the strain actually decreased in an inverse parabolic shape for some hours before it gradually started to creep up again quite linearly over night. I am running a new test over the course of the weekend to see if it just keeps going. \$\endgroup\$ – Vincent Vega Aug 9 '19 at 16:49
0
\$\begingroup\$

Might be heating due to current, though the full bridge should mostly null that out. Could be creep. https://en.wikipedia.org/wiki/Creep_(deformation). Could be drift of your excitation voltage.

| improve this answer | |
\$\endgroup\$
  • \$\begingroup\$ Well it does say that material creep occurs at 35% of the melting temperature of the material, which for ANSI316 SS would be rather high. When these things are in production they are actually under water, so cooled down by seawater, but still giving us loads of problems. \$\endgroup\$ – Vincent Vega Aug 8 '19 at 11:36
  • \$\begingroup\$ Could be creep in the bond, in addition to the substrate. \$\endgroup\$ – Scott Seidman Aug 8 '19 at 11:38
  • \$\begingroup\$ Also see strainblog.com/content/introduction-gage-creep \$\endgroup\$ – Scott Seidman Aug 8 '19 at 11:41
  • \$\begingroup\$ Sorry, English isn't my first language, is the substrate the surface of the measured material? Yes, the bond might be the problem, I'm not sure how much glue they used or which kind. As far as I've learned, cyanoacrylate is the go-to adhesive for strain gauges in normal temperatures, but some are much more thin-flowing than others. I know we've had a goopy kind before, which might cause problems. \$\endgroup\$ – Vincent Vega Aug 8 '19 at 11:43
0
\$\begingroup\$

I‘d suspect the type of glue. We use M-Bond 610, which is a special two-component hot glue. This also needs conditioning and neutralizing with the proper chemicals, after sanding just the right way. Applying strain gauges really is an art and requires careful process design and loads of know-how. If you need professional help, contact me. The company I work at specialises in one-off and custom designed load cells for new installations and/or replacement of unobtainable parts. We would be happy to help!

Edit: We can laser-weld for IP68 rating as well.

| improve this answer | |
\$\endgroup\$
  • \$\begingroup\$ Thanks, do you have a website? We've nailed the process down a lot better thanks to this thread. It's still a lot to learn though, I guess experience helps. \$\endgroup\$ – Vincent Vega Sep 19 '19 at 11:40

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.