Sign up Calendar Latest Topics Donate
 
 
 


Reply
  Author   Comment  
BaartCM

Member
Registered:
Posts: 15
Reply with quote  #1 
Hi guys,

I read a paper talking about new parameters for bearing defect indication, Thikat, Talaf, Insar and Sianna. Unfortunately I can’t post a link as the forum thinks it’s spam but if anyone is interested, I can email them a copy.

I’m wondering if anyone has any experience in actually using these parameters in a ‘real’ application, not just a test rig in a university lab. What about the other ones they mention? Eg shape factor, impulse factor etc!

I’m writing an app which collects a 2048 point twf at 2560sps every 5 minutes. FFT would give an 800line spectra with fMax 1000hz. I want to calculate and save all the trendable parameters I can from this twf before it is discarded

Basically, my system uses several four channel modules connected and powered by Ethernet cables (Power over Ethernet) I have written a comms protocol which allows me to set the sample rate and can collect anything upto 8196 samples (3200 line spectra)at up to about 50,000 samples per second. Once per day, a full twf is collected and stored but regarding the trends, only the results are stored otherwise I’d need a massive amount of storage. It is being tested on two paint lines at a major manufacturer in NE England but I’m still tweaking the PC software and I’d appreciate any suggestions, advice etc. regarding what best to trend.

Thanks guys, Steve.

Attached Images
jpeg 01BD12FC-2B80-4AFC-8122-2B7B7D164614.jpeg (539.57 KB, 25 views)
jpeg 9BA352CA-FCE9-46AD-B108-FA3C8C0233D2.jpeg (225.16 KB, 25 views)

OLi

Sr. Member
Registered:
Posts: 1,914
Reply with quote  #2 
1000Hz is way to low for bearing condition IMHO. We use 12kHz and are now considering 8kHz unless your machine speeds are really low and even then.
__________________
Good Vibrations since early 1950's, first patented vibrometer 1956 in the US.
http://www.vtab.se
BaartCM

Member
Registered:
Posts: 15
Reply with quote  #3 
Oli, the 1khz spectrum is only for the trends. I have amended the post to show that the module is capable of sampling rates upto 50ksps - actually it can do faster but I’m not sure of the accuracy at higher rates and don’t see the need to concern myself with higher rates.

I do wonder if I should take a second, higher rate twf to calculate additional parameters. I’m here for advice.

Steve.

OLi

Sr. Member
Registered:
Posts: 1,914
Reply with quote  #4 

Yes I do that. Currently we sample about 30kHz and filter so we use up to about 12kHz and that has worked for nearly 15 years with std. accels and I do look at reduce to 8kHz due to sensors available now.
If you monitor electrical motors it is good to have a FFT up to 1500 or 2000Hz, all is not according to ISO IRL. If you monitor average gbx's you need more depending on speeds etc.

 

 


__________________
Good Vibrations since early 1950's, first patented vibrometer 1956 in the US.
http://www.vtab.se
John from PA

Sr. Member
Registered:
Posts: 963
Reply with quote  #5 
Quote:
Originally Posted by BaartCM
Hi guys,

I read a paper talking about new parameters for bearing defect indication, Thikat, Talaf, Insar and Sianna. Unfortunately I can’t post a link as the forum thinks it’s spam but if anyone is interested, I can email them a copy.


Interesting. I also got a spam message relative to a link to the paper titled “TALAF” AND “THIKAT” AS INNOVATIVE TIME DOMAIN INDICATORS FOR TRACKING BALL BEARINGS.

Could be we both are referencing the same paper?  In any event if someone wants to read the paper I mention insert the full title into a search engine.
BaartCM

Member
Registered:
Posts: 15
Reply with quote  #6 
John, that’s an older paper before siana and insar were added.

The full title of the more recent paper is


Time-Domain Based Quantification of Surface Degradation for Better Monitoring of the Health Condition of Ball Bearings

And it includes the formulae for calculating said parameters.

Attached Images
png 727E4419-91B9-4C04-908B-92CEF9DF44E8.png (413.70 KB, 22 views)

John from PA

Sr. Member
Registered:
Posts: 963
Reply with quote  #7 
Same paper I found...let’s see if this direct link to MDPI will work.

https://res.mdpi.com/d_attachment/vibration/vibration-01-00013/article_deploy/vibration-01-00013.pdf
OLi

Sr. Member
Registered:
Posts: 1,914
Reply with quote  #8 
Interesting, I can agree on some findings on the old parameters like RMS naturally have a later detection and Peak an earlier but what I and some 3-letter company also found is that Peak trend is more irregular as time pass and RMS give a more regular trend so if you like a early detection for some reason go for peak, if you like a stable trend and a indication when to replace brg. go for RMS it is obvious due to the rectification applied. Seems the new parameters premiere even earlier detection as it is like Peak on steroids. I personally don't need that, if I IRL can give a warning 12 months before "you really need to change the bearing now", it is fine for me and give me plenty nagging time but it is interesting. I have seen the resulting system that used the data from the Red Army running down many thousands bearing w/o simulating faults and that gave a system that never indicated a normal functional bearing as I guess in normal use there are none but that was not useful either as I just want a indication in reasonable time when the bearing is finished.
__________________
Good Vibrations since early 1950's, first patented vibrometer 1956 in the US.
http://www.vtab.se
Walt Strong

Sr. Member
Registered:
Posts: 889
Reply with quote  #9 
Here is a link to the first paper:
http://profs.etsmtl.ca/mthomas/Publications/Publications/A19-sassi.pdf

I haven't "digested" any of this!

Walt
Curran919

Avatar / Picture

Sr. Member / Supporter
Registered:
Posts: 599
Reply with quote  #10 
I also read the old THIKAT/TALAF paper and was pretty intrigued. I implemented it into our software, but haven't had a test case that really shows its usefulness yet. I'll have to look at these other parameters. Like you say, the more scalar parameters you can trend, the better. You can always figure out what to do with them later.

The problem is taking all of these parameters and appropriately weighing them against eachother to put into a useful single bearing parameter based on the user's requirement for pre-warning time and appetite for risk. This is the million dollar question, and in this regard, we are heading into a much more competitive topic I think. [rolleyes]
Previous Topic | Next Topic
Print
Reply

Quick Navigation:

Easily create a Forum Website with Website Toolbox.