Surface Temperature Measurements using Thermocouples--Best Practice? 2

[Jimmy770]

Hello Folks,

I am looking to measure surface temperature on the inside of a flat duct through which air is flowing. I plan to put holes in the surface to stick the thermocouple through, but my question is what should I do with the sensing end to get an accurate reading of the surface temperature without disturbing the air flow?

Maximum temperature will be approximately 50 Celsius (122 F). All of the "surface" thermocouple mounts I have seen have are relatively obtrusive. I would have to wire through the duct from the inside and then curl the surface probe over (bending the wire at 90 degrees) to connect it to the surface.

Other thoughts have included wiring the thermocouple and making the solder bead such that the thermocouple will not pull all the way through, or soldering the TC to a copper plate and placing the copper plate across the hole so the sides rest on the surface, which is aluminum, and the copper should heat up to the same temperature.

Lastly, if I use more than one thermocouple, as long as they touch the aluminum surface will they interfere with each other? I should think it would be acceptable as long as the wires are isolated from each other when connected to my instrumentation.

Sorry for the long post, Thanks for reading and if anyone has any suggestions or a yea/nay for the ideas I've presented I would love to hear it!.

Many Thanks

 

[IRstuff]

Aluminum is a conductor, and was even used for house wiring. You should insulate the thermocouples from each other

TTFN

FAQ731-376
Chinese prisoner wins Nobel Peace Prize

 

[MintJulep]

http://www.rdfcorp.com/products/surface_therm/t-f_01.shtml

 

[Compositepro]

It all depends on how accurate you have to be. You also have to be fully aware of what YOU mean by accuracy. Your case does not sound difficult, though. Be aware that thermocouple wires are fairly good conductors of heat so heat transfer along the wire will affect your measurement. This is usually minimized by having the wires leaving the sensing junction in a direction where the temperature gradient is zero for significant L/d where d is the wire diameter.

 

[Jimmy770]

IRstuff,

Okay, I thought maybe it didnt matter, but I guess it does. I will find some paste.

MintJulep,

Thanks, I have been looking for something like that, just need to see if I can bend them 90 degrees.

Compositepro

I would prefer to be as accurate as the thermocouple type will allow (supposedly +/- 0.5 C). The wires will run vertically through 0.5" of aluminum, which is the shortest distance out so-to-speak. The temperature gradient will theoretically be at max 25 degrees C and I was going to use the smallest insulated thermocouple for my range I could find, which I believe was on the order of 0.05 inches. Am I correct in assuming that is significant? (L/d ~ 10). If the wires are long enough will that lessen the problem? Are there any good ways to compensate for it? Lastly, is it possible a thermistor would be a better fit for my application?

Many Thanks for all the replies!

 

[racookpe1978]

It might seem an odd question, or even a trite question, but are you trying to measure the "air" flowing right at the duct wall (on the air side of the wall-air transition), or the wall temperature of the duct right at its intersection with the air (on the wall side of the air-duct transition)? Or are you interested in the (average) temperature of the air as it flows through the duct in the middle of the duct?

What is your Reynold's number of the air? What is the delta T of the air going through the duct?

 

[hacksaw]

50 DegC does not give you much of a signal with T/C's, you need to consider thin film rtd's or thermistors of some kind.

your temperature gradient is a concern as you have such a position dependent temperature that it cannot be measured meaningfully unless you are doing profile measurements.

the L/d of the sensor is largely irrelevant given that the gradient near wall is so great

 

[davefitz]

Disturbing the airflow on the inside of the duct may cause an inaccuracy. Maybe the better configuration is to solder a pad thermocouple on the outside surface of teh duct and to insulate the "adiabatic side" of teh outer duct and thermocuople assy. Of course the thermocouple will need to be calibrated first.

 

[moltenmetal]

If you want duct metal rather than air temperature: use a type T thermocouple, get good thermal coupling to the duct metal, attach to the outside of the duct and insulate as davefitz has suggested- otherwise you may have radiant effects or disturb the airflow. This assumes that the "duct" is thin, and metal. But the temperature you measure will not be the same as the temperature that duct actually runs at. Best to verify with a pyrometer, properly calibrated and compensated for emissivity.

Want to know the air temperature? Insert a thermcouple but make SURE it has a radiant shield on it. Air flow will be disturbed but that will help you get the measurement you want.

 

[Jimmy770]

racookpe1978, I want to measure the temperature of the wall at the air-wall interface. The DeltaT of the air will be 25 degrees C, and the reynolds number will vary between 400 and 2300.

Hacksaw, 50 degrees is on the lower side of thermocouple sensing range but I wouldn't think that would cause a significant problem. Now do you mean a temperature gradient through the duct axially or through the wall?

Davefitz, I will be measuring temp on the wall outside the duct as well, so for all intensive purposes that will not work.

Molten, As mentioned, attaching the thermocouple to the outside of the duct is at this point unacceptable. The duct material itself will have really low emissivity to minimize radiation interference, unfortunately that also means pyrometers are out of the question.

Thank you all for your replies, I really appreciate your time and input.

 

[btrueblood]

"Lastly, if I use more than one thermocouple, as long as they touch the aluminum surface will they interfere with each other? I should think it would be acceptable as long as the wires are isolated from each other when connected to my instrumentation."

I have used multiple thermocouples directly spotwelded or otherwise in electrical contact with the measured device, in a number of applications over the years. The thermocouples don't care. Your measurement circuit might, but in general if the circuit is isolated you will be fine.

Conduction loss throught the T/C wire can be estimated, and a thermal analysis could get you back to a "true" (i.e. undisturbed by the presence of the sensor) wall temperature. Not always a trivial excersize, but it can be done. Routing the wire for some distance downstream before penetration of the wall is the method I think CP is suggesting, similarly you can fun the lead some distance under insulation, both methods move the point of conduction heat loss away from the measurement point of the sensor.

Ribbon style thermocouples are available, where the "wire" is only .002 in. thick. See

http://www.nanmac.com/documents/b4.pdf

 

[hacksaw]

at 50DegC your thermocouple will be producing microvolts and measurement error will be driven by a lot of factors including the extension wires used and the precision of the cold-junction compensation.

once you define what you are measuring (metal or fluid), the accuracy you need, then you can sort out the sensing methods.


at the Reynolds numbers you describe, stratification of the flow will be an issue depending on the geometry. Radiation is always a factor with walls so much cooler than the fluid, but it all depends on the specific details of what you are trying to do

 

[moltenmetal]

RTDs will give you higher measurement accuracy in a dry block calibrator, but they'll also be physically larger and harder to get a reading without disturbing the flow. So it depends on how important an undisturbed flow is to you.

50 C can be measured to quite a reasonable accuracy with a very physically small type T thermoccouple.

Running the leadwires downstream for a distance is good practice and will typically eliminate the lead wire thermal conduction error.

I don't know what you're doing, but would a heat flux monitor not be better?

 

[hacksaw]

interesting set of comparisons, notice the effect of insulating the duct work albeit for a steel duct

the calculation does not include temperature gradients in the fluid

 

[btrueblood]

Can't see what that chart says, can you post a sketch of configuration, hacksaw?

Regarding T/C errors, any thermal sensor can have errors from multiple sources. Most errors can be accounted for. The quoted 2 deg. C accuracy for typical thermocouples and wires is easily improved by direct calibration in ice point and boiling point baths.

 

[Jimmy770]

Molten, It cannot be a heat flux sensor, I need to know the surface temperature.

Hacksaw, That is an interesting graph. Is it assumed that the "length" is the length of the whole wire or is it just the length of the "tip" of the thermocouple? Does thermocouple error really improve for increasing speed? For some reason that seems counter-intuitive. Is the graph referring to insulated ducts or an insulated thermocouple wire?

Btrue, So if I calibrate all of my thermocouples then I can get better than the 0.5 degree accuracy that is the "special limit of accuracy" for the T type thermocouple? And thanks for the link, I will see if I can use something like that.

Thank you all very much

 

[btrueblood]

Jimmy,

The basic 2C accuracy of thermocouple wire is "run of the mill" for typical type K, i.e. it is at best just calibrated in batch form. All TC wire loses accuracy due to oxidation damage over time, and/or due to defects associated with fatigue damage. There's a lot of detail on this and other subjects surrounding accuracy, in the white papers on the site I linked to (nanmac.com).

Best methods for thermocouples, or for any sensor for that matter, is to calibrate the sensor in end-to-end fashion, with leads and measurement circuitry in place, just prior to connecting the sensor to the test article. And yes, technically, you can get the accuracy of your sensor to better than the 0.5 degrees if you are careful in choice of calibration standards. It's not rocket science (even though that is where I cut my teeth and learned this stuff) but it IS time consuming, and can be a pain to implement, especially if a boss/manager type is breathing down your neck wanting to see his test up and running.

 

[hacksaw]

the calculation is for a slender sensor in thermal contact with the conduit wall (in this case a 8" pipe) perpendicular to the inside pipe wall and extending into the flow. The lengths assumed (0.75" & 1.00") represent the length of the sensor penetrating past the pipe wall with the base of the sensor fixed at the pipe wall temperature.

The pipe wall temperature assumes ambient temperature of 70 DegF, for a horizontal pipe with normal convection for both un-insulated pipe and pipe with 1" insulation. The wall temp is 108 DegF without insulation and 119 DegF without- that too will be flow dependent. The pipe wall emissivity is taken to be unity for a worst case estimate.

@Jimmy770
The tip error is reduced as the flow increases because of improved heat transfer between the fluid and the sensor. Radiation, conduction and, convection effects are included.

 

[HDS]

http://www.omega.com/Temperature/pdf/5TC.pdf

Stick on bare wire thermocouples.

 

[EdStainless]

Great comments, it just so happens that I have been playing with a similar issue.
1. insulate TC sheath from duct
2. Use non conducting sheath (I managed with high temp thermoplastic)
3. Use as thin of TC wires as possible, I was using 0.006" wire to minimize thermal conductivity losses
4. TC junction needs to be shrouded, open for flow direction but blocking radiant transfer in all other directions
5. If the distance isn't ridiculous run the TC wire clear back to the instrument (no lead wire or junction)

If I had the instrumentation I would use RDTs. Microvolts are never your friend.

= = = = = = = = = = = = = = = = = = = =
Plymouth Tube