I was sitting at my desk today looking at some data and it got me to thinking.

How do most folks view data and make calls? Are the mass majority of folks here at this forum the type that looks at all spectrums and waveforms for all points where they collect data? Are there any who use just trends? Maybe a combination of both. Of course, we all use our most valuable tools - sight, sound, smell, etc....

I personally am an every spectrum/waveform kinda guy. Normally, I scan the spectrums and waveforms for anything that seems out of the ordinary. It is very time consuming and when I first started with this job often found that I could easily get thrown off course by spending too much time trying to figure out what a peak was being caused by when the amplitude was so low that it didn't really matter anyway. I found that setting a constant amplitude scale for the spectrums helped to cull the high nails a little more quicker, but I still find myself getting caught up in the unusual peaks that usually end up not leading anywhere anyways. Also, making notes for my findings on all the equipment I check in a spreadsheet seems to help me keep an eye on the potential problems that haven't reached a point where I feel it is necessary to make a call. I have always been guilty of getting caught up in details, hence the inquiry.

A search for Autostat and Inspector (and other vendor's versions) shows that there are some users partial to these technologies. I've heard the pros/cons of Autostat and have researched this option in the past. Only problem with me is that we don't have a gold version of CSI's software.

Really, I was just curious about the course of events that some of you folks used after dumping the data back to a PC. I thought if I started a post on the subject that it may help all of us to hear different points of view on the subject.

Well, I'm an exception-only kind of analyzer. My version of plotdata has these nifty test tube looking icons next to the equipment--they're different colors for different alarms. I can go right to the bad equipment. I use the tab key to cruise through the points (usually FFTs first) then switch over to the waveform and shift+tab backwards through the points. I very rarely use the mouse. If something pops up or if I have a note on something, I'll bring up the trend (ctl+9) to see what it's been doing, look at the multiple spectra screen (ctl+2), maybe the multiple waveform (ctl+6), all the while switching between velocity units and acceleration--depends on amplitudes and what I'm looking for. That's how I review the data.

As far as what is good, bad or ugly, that depends on what I saw in the field, what I may have heard from the maintenance guys and what I see in the data. When I write something up, I have just a few categories: Watch, Caution, Alarm I and Alarm II. Watch is for my information only, just so I can keep tabs on it. Caution is when I think it's significant enough to bring to the customer's attention, not necessarily actionable. Alarm I and II require attention. I've altered RBMview to use just these. I'll save the graphs to the problem, explain why there's a problem and recommend a solution. My final report is a combo of Problem Report A from RBMview and a summary report I created myself in Access. If CSI made the notes from the route data accessible from their ODBC driver, I'd have 80% of my report done in the field. It's a little pet peeve I have with them. I mean, really, with all the useless info available through ODBC, how hard is it to include the one or two tables that contain info we could actually use? But I digress.

I, too, fall into the trap of seeing something odd in the spectrum that doesn't seem to match up with anything and then pick apart the waveform and try to figure it out just to conclude, "Well, it's not the bad. I'll keep an eye on it."

First let me say, I'm in house, so I see the same equipment every time. With this, I mean I have a feel for how the machine should sound or feel when I walk up to it. I do about 75% of my (complete) analyzing out in the field. If while taking readings and the machine is smooth and vib levels are steady from the last reading (overall shown on the screen for each measurement point), I usually won't spend any time on it in the office. The ones I do look at in the office, I usually start out with viewing the single parameter trend for overall level. Then kick over to multiple spectra to see what frequencies have changed, check out the multiple waveforms from here, then go on in depth and find what frequencies are causing the problem.

I'm similar to Don. In house and familiar with my equipment. I do most of my analysis at the machine. My data collector is set up to show me the last overall then the current spectrum when the reading is finished. I will review all the data at the machine and take off route data if warranted. Sometimes I save it and make a note to look at it in my office. I generally look for high 1x, 2x or harmonics. Also non-synchronous patterns. And ofcourse my eyes, ears, nose and hands are at work at the same time. I also carry a 9" flat screwdriver to have a listen. This can be a good reality check.

I'm a view everything guy. I use Entek stuff and software lets you build your own plot views. I've built one called alldata that lets me see everything at a glance and it is just a matter of paging through the points, looking for anything out of the ordinary.

Hey Ed. On your waterfall you can go to options; Z axis and select "Evenly spaced data" it will clean it up for you quite a bit. I just learned about that little feature last year.

I usually look at the overall trend, spectrum and waveform plots for every single point in my databases. For the trend data, I tend to look at the overall, HFD, PK-PK waveform, VHFD and crest factor mostly. Then I like to look at the historical spectrum and waveform plots for increases or changes. I take alot of extra data out in the field and make notes in my data collector as to what I should observe more carefully.

That's a great tip Ron. Usually I have to limit how far back in time I go so my waterfall doesn't get overcrowded, but your way is much better.

Hey, way cool Ron! Thanks for the tip.

I am also a look at everything guy. My approach is very similar to Ed's and I also use Entek.

I use to feel that I knew if a machine was good or bad when I took the data. However, I now find that I am taking less and less of my own data. This is not what I recommend but it seems I have less and less choice in the matter. I feel the alarm approach can work for most catches. But to catch the small stuff you either have to look at it or spend so much time customizing your alarms that you might as well look at the data.

I am a quick look at everything guy too. I usually bring up multiple spectra, usually the last 4. Look at the first point for any change, tab, tab, tab, go through all of the points quickly. I know all of my equipment pretty good so if I didn't have a problem before and there is no change I go to the next machine. I usually depend on changes more than amplitudes. If I see any change I tab through the last waveforms just like I did with the spectrums. If anything has changed on a point I go to single spectrum and hit the X key and toggle through the garmonic families and try to identify all of the peaks. I take a lot of notes in the field and if I entered a note on a machine the little box on top of the screen is red, then I know I have something I wanted to look at. I have worn the headphones for a few years when I am taking the data so I usually hear anything that needs attention. I could probably use the headphones and just hit "field alert" on data I need to look at and do ok.

I have the data collector show me the, overall and compare it to last, spectrum and waveform while taking the measurement, being familiar with the machines I know what the data should look like, if I see anything untoward I collect extra data if required then hit the mark key this makes a note against the point to ”review data” , I then use the autoplot “review data” notes filter to quickly go through points of concern, this method allows me cover more machines with better accuracy than if just relying on alarms which I can also use at the machine to trigger my use of the mark key. I find this a lot quicker than trying to make notes with the data collector onsite; most of my notes are made back at the office in the software

There is nothing wrong looking att all data all the time, but your mental capacity expediting measurements has a limit. If you train the software to do the green (the good ones) base of the population pyramide of machines and focus to look at the upper small number red and then the orange and spend the time left over on actions to cure the warning and bad ones, you can do a more efficient and rewarding job.

This is all very much depending on how your company has organized the work. I see, generally, that people tend to stay longer, grow experience and enjoy their work more the more of the ladder from routine measurement to detection to cured sound machine they can be involved with.

A fresh example: One customer I discussed this with during Christmas said he did spend some quarter of his work on routine green machines and the rest tuning the database (alarm levels and points to take to routes) and actively fighting bad machines the first 2 years. Now, after 7 years, he spends 10 minutes a week on trimming the alarms in the database and only reacts on trend alarms of trend warning and "red". He covers 1672 aggregates being two paper machines including basement and fans on top and an old paper intake process. We ran the vibindex report and I could see clearly that the main heap was well high and red during the startup years. Now there are some 10 red and some 30+ in warning status and he claimed to have a reason for each of them. He is very satified and also mentioned that the company had recognized his efforts and also shown that in the salary.

My aim was to focus the issue about a company using your expertise. Are you basically a watchdog or are you a part of the curing process that ensures a timely repair and a change of a machine that reduces or eliminates a repeat. I think the more you can cover it broad, the more your company can use your capacity and experience and the more you like the work. Just an observation here, I can only speak for the fellas I meet, there can be a lot of other approaches just as valid.

Very good reply Arne, definitely food for thought. I was on a kick some time back to get my company to purchase Autostat for the purpose that you speak of, but because of the fact that I'm looking at all the data, I really don't have time to put together a proposal for this purchase (kinda ironic isn't it). Perhaps it's time to raise this idea from the dead and try to breath some life into it again.

Very Cafefully.

I am a look at everything person. Never had much faith in artificial intelligence looking at paper machine data. Guess on a fixed speed pump or fan with few process variables it would be OK, but with our 60 year old machines that are running 3X faster than when installed, you will go up in flames!! Trending on the machines varies so much with speed changes that it is only used as a guide. I look at patterns in the spectra and use that to trigger more indepth analysis. Being stuck in a mill, I do not have to spend extra time collecting analyze data; I can always go back and do the indepth stuff later. Over the past couple years, I have been working to pass the data collection and analysis off to a younger group of techs, while providing guidance, some insight, and all the tricks I learned over the years. I now have the opportunity to pick machines/issues that affect reliability (the PITA machines), determine root causes, and develop actions/tasks to make them run faster, jump higher, live longer and make more paper. Being able to go out in the field, drag up a chair, and have a heart to heart with a rogue machine is what Arne speaks of. It makes for a good day!!

Gary B

I’m another in house guy who collects his own data, and have the luxury of being a 5 minute walk away from any machine on site, so that I can collect any extra data that I feel I might need. It’s common for me to make a follow up collection before I make a call, and might carry a number of tools to help me; ultrasonic gun, strobe, IR camera. It makes the decision making process easier.

As far as viewing data goes, I am an exception only analyst. That is to say, that if no alarms are tripped, then I don’t look at any data on the machine at all. If one or more alarms are tripped, I may look at the whole set of readings and not just those that are in alarm. I’m of the opinion that if someone who is a damned sight cleverer than I am is going to design a software that will filter out the good from the bad, then it would be rude not to use it

Of course the rules can change on variable speed machines and I may look at the data differently, and yes I do commit a certain amount of time to trimming alarm levels, but I feel it’s worth it and saves me time later on analysis.

I use both magnitude (trend) alarms and spectral alarms. I’ve also got a few TWF alarms on certain applications. I’ve used RBMware and Odyssey and experience has shown that they can both give me the confidence that I haven’t missed anything if the alarms don’t trip.

My current method of analysis (which Odyssey allows because of it’s customisable views) is to display the current spectrum, waterfall plot and trend plot simultaneously. I have the archive filter set to 12 months, so that I can toggle between all historical data and the last 12 months of data at a click of the mouse. I figure if the current spectrum hasn’t changed much in the last 12 months, then there’s a good chance I’m safe until the next data collection. It helps me to walk away from those machines where you know something isn’t quite right, but you’re not ready to make a call. The remote control toolbar allows me to keep this view on the screen and toggle between readings.

I think in that in this game there are a few black and white rules, and then a huge grey area of personal preference. My colleague who sits next to me covers the same plant using the same hardware and software and we don’t seem to do anything in exactly the same manner. After reading what everyone else does on this forum, it only reinforces that opinion.