This is an issue from a previous thread...In the typical Thermal report, we use a reference point and develop a Delta "T" between the anomaly and the reference point.

I have found with many anomalies, however, that it is quite common in post repair thermographs to find that the entire device had an elevated temperature, particularly in enclosures. What does this say about the reliability of our reference points?

I would think that for severe anomalies, we usu. recommend immediate action regardless..i.e. plus or minus 10c is not going to make a difference in our decision making. However for moderate anomalies, this could have an effect of making an anomaly look less severe than it really is.

Have any of you ever considered this, if so what do you do about it?

That question can take us in philosophical directions pretty quickly.

One thing I would say is that no simple rule such as a delta-T limit is going to cover all situations. But comparing delta-T between phases does a pretty good job of identifying degrading connections. (Also in the distant past on this forum, I proposed a slightly different method which I believe does a better job at identifying differences effectively, but adds some complexity.)

Now if you should go past identification and try to judge severity, the questions can get a lot murkier and more complex. If you want to look at how close you are to a material temperature limit, then I agree with you that the rise is not the only factor, you have to look at the absolute temperature. There is certainly a lot more that can be said on the subject of judging severity. John Snell has published some good stuff on that.

What do I do about it?
I try not to spend to get dragged too far into the business of evaluating severity. If it's a non-critical circuit, we will repair it on a routine basis. If it's a critical circuit, we usually will schedule repair pretty quickly regardless of severity. If it's a critical circuit that can't be deenergized for repair, I get pushed to evaluate severity so the risks/costs associated with repair can be weighed against risks/costs of not repairing for awhile. Absolute surface temperature indication as compared to material or design limits seems to carry a lot of weight with my audience in such an evaluation, but I try to emphasize to them that failure is much less predictable than is implied by comparing a single number to a limit.

We haven't seen any cubicles where temperature of the cubicle is identified as a problem in itself. But there are certainly many situations that don't fit the mold of a simple visible connection to be evaluated using deltaT.

Don't know if I missed your point.

This message has been edited. Last edited by: electricpete,

You definately got my point Pete. It seems you've tried to greatly simplify by avoiding severity judgements, and that is good. As we all know, we do get asked these kind of questions and as you said it can get tricky pretty quickly.
My interest is not really on what we say to the customer, but our internal decision as to whether we elevate to "ASAP" or just "Plan".
Obviously at some point you make those decisions, and they are based largely on Delta T. And of course they need to be reasonably defensable if questions do come up.
Have you ever underestimated a pending failure?

We have never had a component fail between the time it was identified and repaired. For all I know, that is just luck. Have you ever had that?

Since I have been involved in the program (which is only two years), we have had four different electrical "failures" in the plant due to overheating and none of them were anything we had identified. Three were non-critical equipment which we don't survey. One was critical equipment (an inverter) which we survey and the component that failed was an auxiliary transformer...didn't fail due to hot connection but did fail due to general overheating (not a condition we were looking for)

I have heard two occasions at other plants where a hotspot associated with a fuse was identified (I'm not sure if it was at the clip or the ferrule or what), and the fuse blew before the hotspot was repaired. I have the written report for one of those buried in my files somewhere.

Pete:
If its not too much trouble, that example could be a really great lesson/example to those of us who, like you, have yet to see an identified anomaly fail before repair.

The disadvantage I have is experience.. for ex: in Vibration, I have lots of information and examples, data and other stuff, but for therm, there isn't as much data, the percentage of anomalies/nominal is very low, so it takes much more time to accumulate experience with these kind of things.

I did report on a fault with a 7 Degree C delta. The customer decided it was of minor significance and delayed the repair. 5 months later it still had not been repaired and it then failed in a big way and took out one production line for a couple of days. The fault was a connection problem on a contactor. A couple of things could have been done, firstly it could have been repaired/replaced. Secondly they could have prepared for failure (sometimes the best option), if they had done this it would have been back up and running after an hour or two instead of the couple of days. Naturally when you do nothing at all for five months it will fail, and as in this case it will do so on a Friday evening at 17:30, when their supplier was closed and they cannot locate the spares required until Monday morning.

Reading your original post, your delta T can be off due to an overall increase in the component temperature, experiance should help us know when this is the case. You should then use another simular component under simular load as your reference and NOT the component with the elevated temperature.

To follow up: I have not yet had a failure after identification on therm, but I haven't done a huge amount of work in this area either.
Thanks guys for the notes, it is valuable to me.
Bob: Regarding your suggestion about using like device comparisons in the case where a good reference point cannot be found: This is one of the reasons why I have been compiling data/charts of nominal wiregauge/load/ambient and trying to develop a mean Temp for given conditions. Of course it doesn't apply to more complex devices.

Martin,

I personally would'nt bother trying to develop mean temperatures for given conditions. It seems like a lot of work with a lot of potential for error, and I doubt you would end up with something that can be universally applied. Simular components under simular conditions work quite well as references and there are usually plenty around. Of course the best reference is historical data, and if you want to devote time to your thermography then this is where it is best spent.