We have fitted some 4-20ma accelerometers to our plant. They give a dual output. One is to the plant process controller the other allows us to tap in with our data collectors. They are connected to bearings via screwed mounts. Some of them have been working ok then suddenly started to give a maximum flat line 21mm/sec reading. It isn't a true reading because we've double checked with our hand held equipment. I suspect electrical interference from somewhere. I've taken some of them off the screw mounts and fitted them onto magnetic mounts and they've started working again. This was meant to be a temporary fix but we haven't progressed any further. The accelerometers we are using are Monitran 1186CQ-20. Has anyone else experienced this or got any ideas?

I have not experienced something like that, therefore it is going to be just a wild guess.

If the drilled and tapped hole is not square with the contact surface of the accelerometer the piezzo cristall may experience constantly applyed mechanical force messing up the output.
I do realize though that by making this assumption it is hard to explain the fact that they worked OK before.

Dave,
You may have suggested a fix for a problem I have with another IMI sensor mounting. Although I used a spot facer and intergal pilot tap drill, the hole was tapped using a hand tap. There are 4 holes in two bearings, three work normally, the 4th gives me erratic response. I change out the accelerometer with no difference, then moved one of the "good" ones to this spot. Still got a bad response.

A quick look at the data sheet suggests two possiblities:

1. These accelerometers have a mounted resonance of 5 KHz. What type of machines are they mounted on? Any high frequency noise may excite the natural frequency and cause overloading. If the machines are turbines, have gear boxes or ceramic bearings, this could be the problem. Even a motor with a lot of rotor bars could excite the natural frequency with this result. Putting the accel on a magnet lowers the natural frequency, which could explain why the work on a magnet.

2. The data sheet says they are "base-isolated." I'm not certain exactly how they acheive this. If the isolation is not effective, a ground loop will result. If the ground loop potential reaches a high enough potention, an overload again results. Mounting on a magnet could provide enough elelctrical isolation to cure this problem.

Jon
Spintelligent Labs

Jon knows more about this than I do, but if there are impacting events in the area it is possible that the impacts are causeing the integrater to go nuts. What is the machine? What is the process? If it is like a rock crusher or a recip compressor that might mess with the integration circuit.

Jon,
If I put grease or some other insulating material under the accelerometer, that should remove the ground loop? If you think this would be a solution, what would your preference for the insulating material be?
Steve,
This is a fairly smooth machine with no normal impacts.

Isolated mounting pads used to be common, but the trend is towards case-isolated accelerometers - the case does not make electrical contact with the sensor elements at all.

IMI used to have one that was an hard-coat anodized aluminum piece with a threaded, tapped hole, which was adhesively mounted to the machine. The anodization prevented current flow from accelerometer to machine. A quick check at IMI, CTC and Wilcoxon makes me think these are a thing of the past.

Grease or even adhesive will not provide reliable ground isolation. You might try an adhesive base with a thin layer of phenolic or other insulating material sandwiched between the machine and mounting pad. Make the layer as thin and rigid as possible.

You could test for a ground loop problem by connecting everything as you normally would, except with the accelerometer not installed on the machine (but data collector and monitoring system hooked up normally. Using a DVM, measure the AC and DC voltage between the accelerometer base and the machine mounting surface. Anything over a few millivolts would indicate a problem.

Jon
Spintelligent Labs