"We have a 600-liter sump that supplies 150 liters per minute to eight turbine bearings from one reservoir. The oil flows from the reservoir by a single pump and filter, then separates to the eight bearings. We take the oil samples from the return lines before they join for the return to the reservoir. None of the samples show much wear. On the last sampling when there was a sign of bearing failure, the sample for that bearing showed only 4 ppm iron, 2 ppm tin, 1 ppm aluminum, 2 ppm silicon, 2 ppm sodium, 2 ppm magnesium, and everything else 0. Viscosity was 66.7 on Regal 68 at 1,000 hours (6,200 on the unit). Why would samples not show that there was a problem?"
The answer to your question can likely be summed up in one word — dilution! Assuming, as you state, that you are sampling on the return lines from each bearing, the reason for the low wear metal levels is likely due to the comparatively high oil volume in the return lines. In oil analysis, wear debris is measured in parts per million. When you state "2 ppm tin," what you are actually saying is "2 mg of tin for every kilogram of oil."
So the same amount of wear debris distributed in a large volume of oil, such as a circulating turbine, will generate a much lower ppm than the same amount of wear in a small wet sump system where the volume of oil is typically much smaller.
To minimize the effects, try to ensure the sample is taken as close to the bearings as possible. In addition, ensure that the sampling method is precisely controlled with the same method used every time, including flushing volumes. You might also need to tighten your alarms considerably to the extent where "normal" really means 0 ppm of tin (and other key elements), and any slight increase (even just 1 to 2 ppm) is considered "cautionary."
Other possible causes of the low wear metal limits could be a failure mechanism, which really doesn't generate significant amounts of wear debris, or one that creates larger sized particles (in excess of 5 microns), which do not show up in conventional elemental spectroscopy techniques. However, given the nature of this application, dilution is the most likely cause.
While most lubrication teams grasp the importance of good oil sampling, they are not exactly sure how to address the specifics. The good news is that instituting a quality oil sampling program in a plant is generally a relatively inexpensive exercise with high, short-term paybacks. If you haven’t instituted a world-class oil sampling program, now is a good time to start. If a sampling program is in place, maybe now is the right time to review it for compatibility with reliability goals. Don’t let a poor oil sampling program shoot your oil analysis program from the sky.