Does anyone know why a Strain Gage measurement drifts over time?

[Paul V]

I am building a test platform to measure total stress placed on various metal supports in a large industrial machine. This will comprise of sampling 2-3 strain gages and making the required calculations from the measured data.

I have the majority of the issues resolved, but for some reason beyond my comprehension the measured strain when nothing is happening slowly decreases over time. I initially suspected it was due to temperature changes of the strain gage heating up, but I have not been able to find ANY stability, even after running for an extended period of time. The most recent test created a change of -600psi over 3 minutes. Often a desired test will run for 20 minutes or longer, so this amount of drift will seriously skew any data found at the far end.

Has anyone else experienced this?

Any suggestions how to minimize this drift?

Any other random useful thoughts that would help the process in general? (ok, I know that one is open for interpretation... but what the h*ll.)

I could post a sample vi, but LV seems to be working fine. I think its external to the software.

Thank you for your help, and for all the other posts... I learn something new and useful everytime I browse this place.

Paul

running LV 6.0.2

[becktho]

Just another point you should check because you didn't mention it:

Is the measurement hardware configured correctly? Means you acquire the value of the strain gauge using some hardware (i.e. NI DAQ-Card, Fieldpoint, 3rd-party Card, ...) - is this hardware setup properly?

In addition it will be helpful if you could post information to that topic too (what device are you using, how are you using it, ...).

Thomas

[Henrik_Volkers]

Is this drift permanent?  Say, you have balanced your bridge and you continiusly have to decrease your balance dial for zero?

What type of strain sensor are you using?

My suggestion is the sensor itself, to be more specific the way it is fixed. However please feel free to provide more information.

[johnsold]

If you connect a voltmeter to the strain gauge output, do you see the same drift?

Lynn

[Paul V]

First... thanks for all the suggestions... to answer in the order which they were recieved...

1) Hardware Configuration:To the best of my knowledge, it is correct. I am using a DAQCard-6062E for data acquiusition, running the incoming data through a 5B01 backplane with 5B38 isolation/ amplification modules to better see the signal and prevent frying the card in the event something stupid is done. I do not believe this is an issue due since I have other channels sampled which stay constant over time which should drift if it was a configuration problem.

2) Additional Information: Currently using a 350 ohm strain sensor from Micro-Measurements across a wheatstone excited with 10Volts in a quarter bridge configuration. Gage  

3) Does the same drift occur via voltmeter... Good Question... didn't even thnk to check that part directly, as LV has treated me well otherwise. Unfortunately, Yes, the VM output also drifts over time.

[Paul V]

Sorry, sometimes my premature posting problem gets out of control... 

2) continued...

Gage information: (Micro-Measurements) CEA-06-125UW-350

Bridge information: (Micro-Measurements) MR1-350-130

Isolation Module: (Analog Devices) 5B38 Isolated Strain Gage Input (wide bandwidth)

Bridge Balancing: This I am performing in LV using a shift register to return the previous loop value to the 'Vinit' terminal of the strain gage conditioning vi. Thus the bridge is balanced using an offset in LV, not on the bridge itself. (I'm performing this project for someone that knows little about such things... and would rather have him press a button on the front panel to initialize the bridge... than to fiddle with a balancing resistor.... though if I can map the drift with no load and reliably replicate it, I can set up a program to scale the data once it is catalogued. Would be a messy way of doing things... but might work. (Brainstorming with the keyboard.))

Ok, I think that is about all there is for now... Again, Many thanks for all suggestions.

Paul

--- LV 6.0.2

[johnsold]

Paul,

Some strain gauges have a significant temperature drift, especially in the quarter bridge configuration. 10 V on a 350 ohm strain gauge is 290 mW. If it is mounted on a poor thermal conductor, self heating could raise the temperature enough to cause quite bit of drift. I am not familiar with your hardware, so this suggestion amy not work: If you have enough sensitivity, try reducing the excitation voltage. At 1 V you have less than 3 mW dissipated in the sensor. The heat is reduced by a factor of 100 while your signal only goes down by a factor fo 10. Other options are half or full bridges with the other elements arranged to compensate for the temperature effects.

Lynn

[Paul V]

Lynn,

Thanks for your thoughts and suggestions. Will have to determine how to reduce the voltage. A previous person on this project bought everything (3 years ago) and I recently inherited it. The 5B38 modules I spoke of not only perform signal isolation/ amplification prior to hitting the DAQCard, but also provide the excitation voltage for the bridge. I  could probably place a resistor in series with the rest of the bridge to drop some voltage..... and the resistance change on the bridge would not be sufficient to change the overall excitation voltage enough to tweak the calculation..... plus the overall change in voltage on the bridge would be consistant, as long as I get the correct excitation voltage into the Strain.vi all the other numbers should follow..... I already have a calibration resistor in the circuit to simulate a specific amount of strain..... if I achieve the same result, it would verify the calculations and modified circuit. (<<-- rambling with the keyboard again.)

yeah, I think I will give it a go. Many thanks,

Paul