Eddy Current Non Contact Displacement Drift 1

[suddenDEBT]

Hey guys,
I have an application with a sensor(see thread title) mounted .040-.080" away from a rotating shaft in a splash lubricated gearbox. The sensor is not submerged but may be splashed, the wire is wrapped in silicone tape at the top of the body to ensure no oil can follow the wire down into the sensor. The goal is to measure the dynamic deflection of said shaft under load. We have tried sensors from two different manufacturers and neither has held up in this environment. Both have threaded bodies with some sort of epoxy doping that retains the internal goodies.

http://www.omega.com/pptst/LD701.html

http://www.micro-epsilon.com/displa.../eddy-current-sensor/eddyNCDT_3300/index.html

We seem to be having a problem with sensor robustness, we are on our third sensor. Two from ME and our first from Omega looks like its starting to fail also.

The problem arises when we do a series of runs to collect data, the system experiences drift from run to run, and from day to day. We will recalibrate between runs, and experience changes during the run until the sensor fails to report reasonable changes in distance. The most recent sensor has "temperature compensation" up to 140oF. This is not really even at operating temperature so we performed our tests until the oil temperature reached 140F. The temperature may have continued to rise past 140 after shut off due to heat soaking. Upon processing the data it is evident that the data points drifted down, (voltage is analog to distance) as temperature went up. We let the unit cool overnight and repeated the runs today, only to find a drift even lower than the previous days hot runs. The amount of drift is not in the realm of correction via signal processing, in the attached image the DC offset is approximately 20% of the signal amplitude. The overall amplitude and wave shapes appear to be correct but as the sensor fails, the drift becomes unreasonable compared to the calibration. This is the same behavior we saw with previous sensors.

Inspection of the sensor(s) reveals no contact, no apparent oil ingress or damage otherwise. We have submitted the following questions to the manufacturer(s), but I am curious what the community here thinks as well:
Is it possible that the thermal stresses have yield the sensor's internals and change its behavior?
Is sulfurized gear oil (EP 80-90w) corrosive to the epoxy or could it be making it swell? (EP oil is corrosive to copper, the OMEGA body is made from nickel plated brass)
Who makes something sturdy enough for this type of application?

 

[ScottyUK]

Bently Nevada (now part of GE) and CSI (now part of Emerson) both make probes designed for this type of application. In fact they both make an entire range of hardware for machinery diagnostics. 140°F should not trouble the sensor. Is it possible that the electronics is letting you down? Older B-N gear appears on eBay now and then - new it is damned expensive, but the industries who use high end equipment tend to have large, high capital value assets like generators and turbines to protect so the cost is relative.

You might get some help in forum666 if you don't get any joy in this one, but give it a few days before double-posting.

 

[suddenDEBT]

Thanks Scotty. Taking a look at the B-N stuff now.
We've taken a pretty hard look at the data AQ setup and have checked for noise, bad grounds, differential voltage comparison, and are still being let down. We performed our own low budget temperature compensation test yesterday and got a pretty nasty reaction from the Omega sensor. We are going to try both the ME and Omega in a heated oil bath tomorrow to see if the temperature compensation curves are at least repeatable. Starting to feel like Groundhog Day.

I think we may have chosen our sensor(s) poorly. The Bentley's look like I would expect a hermetically sealed, temp independent sensor to look.