I have recently been looking at a vibration monitoring system both off- and online that is developed for a specific type of machinery. Special is that it communicates with the user when it think it´s time to do something like bad bearings or balancing and such and you can look at the data but there are no trending as it is. What are your opinions? Is trending a required tool? I myself try to compare trends and similar cases as much as possible. Do I have to do that? Are you using trends to evaluate things? Olov

I think that trend data is one of our most useful pieces of information. I always refer to any available trend information when trying to assess a machines current condition.

Machines are a little like people in that no two are exactly alike. Some machines are "noisy" all the time. Others deteriorate in a nice, predictable fashion. I can walk up to most machines I've never seen before and make a call on it with about an 80% gut-feeling confidence level. But that requires more than just collecting a spectrum/waveform... I also check temperatures, run my hand around the machine, listen to it with headphones, listen to it with a yellow-handled-analyzer, etc.

But to be really be sure (say 95% confident), I need trend data. I wouldn't even consider trying to do "condition monitoring" without trend data. It might be possible, but I wouldn't associate my name with such a system.

I like trending on contineously monitored machines; normal route monitored machines and the trend isn't so great in my opinion. I'm using a CSI 2120 in a metals plant is trending is almost useless on regular routes. But is a tool none-the-less and useful at times.

Where does the machine fail; that is everything. But I see tending all over the planet. I like spectral analyses along w/TWF.

Just to get some more to argue, would you use a route collector for that purpose, route collecting data, that can´t display the data you are collecting? eg. no FFT graphics? Olov

I think OA meters are widely used and you can easily log and track the OA level reading and perform math on it basically better than any data collector I've seen as they don't do the curves generally or curve fitting to reality; only linear curves as a rule or mean taken. So if trending were adequate, we'd be looking for a job.

Sam,

Why can't you trend with the 2120? Don't you use Mastertrend or RBMWare?

Danny

RBMware 4.9v I believe and yes it presents a tend but isn't worth a whole lot. Full time monitoring trends are very good I think but route trends or 5 sec/mo is lacking. So my unit shows Pos AH up, Pos AV down and maybe BH up; who knows what it'll be next.

I personaly would take a five minute a month trend as an indicator of how far a machines condition has changed along with the spectrum and timewaveform that comes with it. Sure it will vary from one month to the next but to show untrained upper management who don't understand the squiglies that the machine is on its way out is very vauable. You don't make calls from trends but its just one indicator.

Sam, how many 100's of examples would you like to see of how well monthly trend data works (taken with a 2120 and MasterTrend)? Most machines trend very well, drawing a flat-then-exponential curve with the 'trigger' being a sharp upswing in several parameters.