No it was actually two anomalies. It is a VFD, but that had nothing to do with it. We did ask those questions though.
We have the wonderful privlege of have VFDs on belt driven systems.

More than once I have seen connections heat to the point they welded themselves together at which point they cooled. I'm not sure if you saw evidence of this or not.

Post mortem images would be nice... do you have any that you can share?

I knew you would ask that...

Sorry for the huge image, I guess I should have put it into a word doc.

Based on your visual it is possible that welding occurred temporarily. When that happens, tempertures can drop but, obviously, not for long. I've watched this sort of thing in real time.

Welding is a possibility; I haven't removed the carbon from those contacts to see what is underneath. This was an indirect Thermograph. The numbers I used on the spreadsheet came from surface temperatures, but there was a crevice on the device where I could get peaks of 125C, and that was still an indirect reading. So there was definately some very high temps at the point of contact on this one.

John -

You mentioned watching this real time - was that an experiment? Were you able to watch it to failure? Anything you can add to how these types of problems progress towards failure?


Thanks for the great case history Martin!

Given the moderate to large thermal gradient involved (temperature difference between the surface seen and the source of heating),it would not surprise me if the measured temperatures were not close to weld temperatures. Even with fairly direct views, such as of a bolted connection, welding can occur at the threads while to bolted connection only appears warmish.

The instances where I have seen this in the field were not planned experiments. In both cases we deemed it not useful to stick around and find out what would happen next.

From the reading I've done, problems I've seen from others, and a great deal of analysis, the increasing temperature trend, if one can be generalized, is that it is slow, even years. Once it gets to a "runaway" condition, failure is quite rapid. Two major factors are involved: (1) heating results in more oxidation which results in more heating and (2) heating results in more RESISTANCE (10C temperature increase results in a 4% increase in resistance) and this leads to more heating. The cycle begins to accelerate quickly at some point and failure then typically comes soon after.

I used a 30x magnifier to look at the connection point and it "appears" that there was molten copper, however when our department metalurgist looked at it he said he would have to do additional testing to confirm it. I told him it's not worth the time, besides from a common sense point of view, it certainly appeared to have melted some copper....
shiny, mottled, pitted surface bordered by brittle black carbon areas.
Regarding your description of the progression to eventual failure: It seems to me that a variety of scenarios are likely:

1. Molten copper flows onto the connection like solder, improving the connection.

2. Molten copper flows away from the connection, reducing surface area of available on the conductor and increasing resistance.

3. Either of these scenarios happening repeatedly, randomly, establishing a cycle of random variation in detectable heat.

I suppose what happens in these regions is a bit of a mystery, and kind of chaotic and therefore risky from a reliability perspective.