How long are the cables?
Are the 6 conductors in a jacket, without other conductors, or are the discrete wires lashed up with all sorts of other conductors in a bundle or ... ?
Are these fine-stranded wires or just the usual in the range of about 7 to 20 strands?
If all the strands are broken and only fully separate on flexing, is it safe to run the robot or will it break things? If a wire fault were detected on a running robot, is there any easy way to use a signal from the detector to stop the 'bot before it goes berserk?
Can you tell us any details of the nature of the signals in the wires? What magnitude of currents? Voltages?
Would a box that you plugged in at one end between the cable and ??? for testing be acceptable? i.e. the box installed in-line but designed so it doesn't interfere with normal operation

I want to try to calculate what sort of resistance difference you would get for a wire with some broken strands. The wires are fairly small and an added fraction of an ohm caused by broken but touching strand ends might be a very small percentage of the normal total end-to-end resistance.
I'm also thinking in terms of some sort of way to detect transient interruption of the signal by sensing current in the wires, or something like that. You would have to cause the flexing that would cause loss of continuity and if that can't be done manually (presumably not, or you'd be doing that way already), then the robot would have to go through the motions.

 

Thanks for your questions/input!

The feedback cable is 6 separate insulated wires, all 6 are run in a shield. I'd say they are fairly typical wire, maybe 24ga, 10-15 strand? I'm not exactly positive from here at home. Longest run is probably 4-6 feet?

So, 2 wires are + and - to an exciter coil. 2 are Y channel, 2 are X channel. 6 in all.

This goes from each axis servo's resolver to a serial measurement board. Basically a 10v, 10k Hz AC signal is sent from the serial measurement board to the resolver's exciter coil, and then a X and Y channel coil sends voltages back to the board for positioning info.

The board monitors these returns, so a loss angers it and the robot errors out for a motion error. No worries of damage or robots smashing people. Yes, for testing purposes, A test box could be temporarily installed I suppose. If stretching the robot out to try to move the cabling to mimic whats happening in auto is the goal yea.

I can somewhat manually flex some of the cabling, but depending on the model, the majority of the cables are run internally, so much of the cable isn't visible.

If the robot exhibits errors on a specific axis- that are determined to be feedback related, Thats when I want to use a defined test procedure to look for damages in those 6 wires for that specific axis.

 

Can the signals be measured relative to a signal ground that could be found near each end of the cable? I suspect the shield would be able to provide the ground.

Since the signals are continuous, it should be easy to "sample" each of them through a high-value resistor or low-value capacitor (a few picofarads), amplify if necessary and use the signal to continuously retrigger a timer with a period somewhat longer than the period of the signal - say 200 µs. With good signals, the timer would never run down. If the timer runs down, it means the signal has been lost for at least a couple of cycles, and I strongly suspect the time would be much longer than that with flexing a cable with a broken conductor with ends touching when not flexed. A simple manually-resettable latch could capture the event and light up a LED. Probably the right way to do this is to look at each of the three pairs with a differential amp so we aren't reliant on a ground reference for them. This would mean we couldn't detect which wire in a pair was broken, but it isn't like you're going to try to replace one broken wire in a jacketed cable. Each pair would get its own amp, timer, latch & LED. I suspect all of the signals are relatively low impedance so loading them at microamp levels would not cause any problem. I need to dig out my old Analog Devices resolver signal processing book.

A fault in the resolvers might cause transient loss of signals that couldn't be distinguished from a conductor fault - presumably you can speak to that.

What do others think of this notion? I have almost no experience with resolvers, so I may be proposing something that is not really doable.