Calibration of Strain Gauges
Calibration of Strain Gauges
Following a work discussion today:
If one wishes to use strain gauge data as part of certification data, how does one calibrate the strain gauge?
If the data involves airspeed, the certification authority would expect to see some form of calibration of the airspeed recording against an established standard. Similarly for masses, pressures and so on. So how is a strain gauge calibrated?
My thoughts were:
- statistically, from the factors supplied by the manufacturer
- some form of additional resistance added temporarily across the terminals (shunt?)
- similar installation technique / bridge circuit on a dogbone specimen
Or are they accepted "as-is"?
Does anyone have some advice on this issue?
Thanks in advance
If one wishes to use strain gauge data as part of certification data, how does one calibrate the strain gauge?
If the data involves airspeed, the certification authority would expect to see some form of calibration of the airspeed recording against an established standard. Similarly for masses, pressures and so on. So how is a strain gauge calibrated?
My thoughts were:
- statistically, from the factors supplied by the manufacturer
- some form of additional resistance added temporarily across the terminals (shunt?)
- similar installation technique / bridge circuit on a dogbone specimen
Or are they accepted "as-is"?
Does anyone have some advice on this issue?
Thanks in advance
Join Date: Apr 2003
Location: UK
Posts: 233
Likes: 0
Received 0 Likes
on
0 Posts
They should come with a calibration certificate from the manufacturer?
If not, you probably need to calbrate in the lab against a known good standard - most likely on a piece of material of known characteristics on a standard test machine in current calibration.
If they're use once stick-on units, you need to test enough to give you a statistically valid sample, identify the (probably) mean and SD characteristics and put 3-sigma error bars on your in situe test data. Better still, the strain gauge supplier did that and gives you the data.
P
If not, you probably need to calbrate in the lab against a known good standard - most likely on a piece of material of known characteristics on a standard test machine in current calibration.
If they're use once stick-on units, you need to test enough to give you a statistically valid sample, identify the (probably) mean and SD characteristics and put 3-sigma error bars on your in situe test data. Better still, the strain gauge supplier did that and gives you the data.
P
Thanks for that.
The strain gauge manufcaturer supplies a gauge factor, which defines the gauge characteristics. Where I am unsure is how the overall system is calibrated, ie gauge plus bridge and DAQ.
For example, if I have a gauge with a known factor and I put it on a wing, how can I be sure that the reading on my data acquisition is accurate? I'm not sure how one would apply a known loads / measured outcome to a structure with multiple load paths.
The strain gauge manufcaturer supplies a gauge factor, which defines the gauge characteristics. Where I am unsure is how the overall system is calibrated, ie gauge plus bridge and DAQ.
For example, if I have a gauge with a known factor and I put it on a wing, how can I be sure that the reading on my data acquisition is accurate? I'm not sure how one would apply a known loads / measured outcome to a structure with multiple load paths.
Join Date: Apr 2003
Location: UK
Posts: 233
Likes: 0
Received 0 Likes
on
0 Posts
The structure may be complex with multiple load paths, but that's not true of the strain gauge. The strain gauge will only only measure strain in a single axis, and similarly that's exactly what you need to calibrate it for.
A good laboratory test rig will measure strain as a matter of course, so if you instrument up a testpiece with a strain gauge on it and a facsimile of your in-flight recording hardware you can come up with characteristics near-identical to what you'll have mounted in flight.
That'll give you discrete strain values in the gauge axes where they've been mounted. Translating that into stress clearly you need to know the cross section and material cross-section of the material it's mounted onto. If it's a complex structure, then you'll certainly need multiple gauges attached to it athough hopefully FEA (or just some competent back of envelope calculations) will indicate the critical bits to instrument.
P
A good laboratory test rig will measure strain as a matter of course, so if you instrument up a testpiece with a strain gauge on it and a facsimile of your in-flight recording hardware you can come up with characteristics near-identical to what you'll have mounted in flight.
That'll give you discrete strain values in the gauge axes where they've been mounted. Translating that into stress clearly you need to know the cross section and material cross-section of the material it's mounted onto. If it's a complex structure, then you'll certainly need multiple gauges attached to it athough hopefully FEA (or just some competent back of envelope calculations) will indicate the critical bits to instrument.
P
strain gauge calibration...
the strain sensor reads strain, temperature effects and poisson ratio as a single output... so your configuration is important to know before saying anything of note. 1/4, 1/2, full bridge, temp compensated, etc.
For a single axis 1/4br strain sensor, a triple wire connection nominally gets rid of temperature affects at least in the leads. The sensor element will have the temperature/strain and gauge values as well as the resistance of the element. The temp curve is pretty much a 2nd order polynomial, and more or less a straight line for small temperature ranges.
Frequency response is not simple, and if measuring strain on a rotating component or one subject to oscillatory loads, then the freq response of the sensor/DAU is necessary. We use a variable speed motor acting on an arm that has a fixed deflection. The outcome is a 5th order polynomial, with an initial linear response becoming constrained at higher frequencies, and then stabilising as a linear response. The curve is sample rate dependent, and is not an artefact of sampling frequency, but a limitation of the DAU, which is supposed to exhibit linear response. I would be wanting to have at least an order of magnitude between freq of interest and sampling freq, particularly if looking at an FFT of the signal.
On temperature, while it is more or less true that the triple lead or full bridge compensates for temp, that is not completely true, if the local flow over the full bridge sensors are not insulated from the boundary layer, and they are attached to a foil with circulation... in such a case, you are also measuring to an extent the pressure differential and therefore temperature differential (Charles, Gay-Lussac).
cheers
the strain sensor reads strain, temperature effects and poisson ratio as a single output... so your configuration is important to know before saying anything of note. 1/4, 1/2, full bridge, temp compensated, etc.
For a single axis 1/4br strain sensor, a triple wire connection nominally gets rid of temperature affects at least in the leads. The sensor element will have the temperature/strain and gauge values as well as the resistance of the element. The temp curve is pretty much a 2nd order polynomial, and more or less a straight line for small temperature ranges.
Frequency response is not simple, and if measuring strain on a rotating component or one subject to oscillatory loads, then the freq response of the sensor/DAU is necessary. We use a variable speed motor acting on an arm that has a fixed deflection. The outcome is a 5th order polynomial, with an initial linear response becoming constrained at higher frequencies, and then stabilising as a linear response. The curve is sample rate dependent, and is not an artefact of sampling frequency, but a limitation of the DAU, which is supposed to exhibit linear response. I would be wanting to have at least an order of magnitude between freq of interest and sampling freq, particularly if looking at an FFT of the signal.
On temperature, while it is more or less true that the triple lead or full bridge compensates for temp, that is not completely true, if the local flow over the full bridge sensors are not insulated from the boundary layer, and they are attached to a foil with circulation... in such a case, you are also measuring to an extent the pressure differential and therefore temperature differential (Charles, Gay-Lussac).
cheers
Join Date: Aug 2009
Location: UK
Posts: 18
Likes: 0
Received 0 Likes
on
0 Posts
The standard way to calibrate strain gauges is with a 'four point bending rig'. This puts a section of a beam in constant bending with no shear. An alternative is to use a cantilever beam, with the beam tapered towards the point where the load is applied. Although the bending moment is not constant in the latter case, the surface strain is approximately so. The curvature of the beam is measured with displacement transducers and the surface strain calculated. The strain is then traceable to length standards via the calibrations of the displacement transducers.
Samples from a batch of gauges have to be tested and a statistical approach taken in applying the results to the rest of the batch.
The above takes considerable effort and is only justified if traceability of the results is of paramount importance. If this is the case, there are some existing facilities that can be utitilised. For example, the National Physical Laboratory in the UK has a four point bending rig and might be persuaded to carry out the tests (for a fee).
Samples from a batch of gauges have to be tested and a statistical approach taken in applying the results to the rest of the batch.
The above takes considerable effort and is only justified if traceability of the results is of paramount importance. If this is the case, there are some existing facilities that can be utitilised. For example, the National Physical Laboratory in the UK has a four point bending rig and might be persuaded to carry out the tests (for a fee).
Last edited by Flash131; 21st Nov 2012 at 08:17. Reason: Typo fixed
