Bearing Monitoring using ultrasound 

As a vibration analyst, we have pretty much an established experience and reference using vibration (seeing BFF in gE then in velocity plots) as when to optimally call a bearing failure (4"ish" stages). For Ultrasound, when do you tell your maintenance guys that the bearing had failed or about to fail? By experience even detecting BFF (bearing fault frequency) in enveloped acceleration is too early a call. I think using ultrasound is much much earlier and relying on it alone may give a case of like "the boy who cried wolf". What do you do when you find ultrasonic indications of a bearing fault besides adding lubrication to see the signal disappear? Would you recommend a bearing change even the vibration signals (gE, vel) are way way too low? How long does an ultrasound signal go without correction before it can be detected by existing vibration HFD (env acc) techniques? A practical experiment result will be interesting to see, pls post link. Thanks