Sign up Calendar Latest Topics Donate
 
 
 


Reply
  Author   Comment   Page 2 of 3      Prev   1   2   3   Next
fburgos

Sr. Member
Registered:
Posts: 644
Reply with quote  #16 
Quote:
Originally Posted by OLi
Well, I built one "expert", 15 years ago and it can help newbies to dare to start but making them more complex is not easy. I used NspectrII quite a lot as maybe the only one? I also used the Russian version that Commtest tried and I think they sold like 2 licenses me and a Japanese....  


I belive rules are nice to have
how does nspectre work, complex machines, gearboxes? you need to create all the machines and asociated components in dbase? then its al ruled based
Alex

Avatar / Picture

Sr. Member
Registered:
Posts: 184
Reply with quote  #17 
Quote:
Originally Posted by Walt Strong
Perhaps the new Elephant in the room!

Press Release

https://www.symphonyai.com/symphonyai-group-acquires-azima-global/

 

https://www.symphonyindustrial.ai/

 

Symphony Industrial AI

https://www.symphonyindustrial.ai/

 

https://symphonyazimaai.com/


Only 23 miles away.






I see that a lot of companies are calling themselves  as leading, the best, No.1 ........

It sounds funny isn't it 
MarkL

Avatar / Picture

Sr. Member
Registered:
Posts: 1,100
Reply with quote  #18 
Funnily enough, I was listening to a podcast this evening with a guy from Texas works for Symphony azimadli. seems that symphony are more into automated DCS and process optimisation, the AzimaDli bit is planned to be an add on to that.
The 'skilled people getting replaced with computers' debate came up and thankfully he was saying they want to augment what guys have, and still have a human element in the mix, he mentioned a few disasters with efforts previously attempted to have it completely automatic.

https://accendoreliability.com/podcast/rrp/rrp-067-using-ai-optimize-processes-don-doan/

The link to the episode is at the bottom of the page.



I have no problem with expert systems to save time and cherry-pick 'low hanging fruit' and let us concentrate on the more complex issues.
OLi

Sr. Member
Registered:
Posts: 1,861
Reply with quote  #19 

Yes, you cover the 80% standard problems with less than 65 rules w/o effort a midsummer eve, at least that was what I did.

Nspectr and at the time competing Mr. Moore solution from Palomar what to became SKF were straight rule based no AI there in my book, you could sneak peak in the rules w/o problem and in Nspectr you could even get them to look at to see what was triggered by different problems like the problem it have to always indicate "locked coupling" it was thus correct once. Azima are similar in tech in my view I looked at all of them CSI,SKF,Azima,Russian out of S:t Petersburg...... I really wanted to see how they did it.
CSI and SKF are pretty similar, CSI was part built by users sending in data and ev. rules so it became messy and very specialized like specific gasturbines and stuff, nothing wrong with that.' Yes you need to describe machinery so it does take a extended effort. In Azima you need every nitty gritty detail or you get nothing in my virew. If you do have a template done it is quicker.
SKF got a bit more personal input by Mr. Moore I thought at the time and maybe Canadian coast guard in some way?
Russian was built from Red Army destroying 20 000 bearings or something, it never declared a bearing to be ok.....
and all of them had hundreds of rules. "Inference" machine as the examining body are called would be a lot of IF and THEN likely exactly like mine, but I stick to lower number rules like 65, and don't say I catch all I say I catch the basic and get a consultant for the rest... So in periods I did receive data to do that.
So I say KISS and you get a support tool for users that think they need that. I have delivered like 384 pcs that are happy w. that :-)
Real Soon in a Android near you.....


__________________
Good Vibrations since early 1950's, first patented vibrometer 1956 in the US.
http://www.vtab.se
Vibe-Rater

Avatar / Picture

Sr. Member / Moderator / Supporter
Registered:
Posts: 1,380
Reply with quote  #20 
I do have a problem with automation. I prefer total manual analysis. Keeps us all practiced up. Pattern recognition. Can't go wrong then. Fuzzy logic comes into all of that.

Vibe-Rater

Avatar / Picture

Sr. Member / Moderator / Supporter
Registered:
Posts: 1,380
Reply with quote  #21 
I have no problem with expert systems to save time and cherry-pick 'low hanging fruit' and let us concentrate on the more complex issues.

All good when you are on  a refinery where you have pump a and b and you can swap relatively easily.  Not so good when on a paper machine where there is minimal redundancy.
OLi

Sr. Member
Registered:
Posts: 1,861
Reply with quote  #22 
Never saw any fuzzy or neural work IRL just in research... We haven't got consistent and good data in 1 place for the neural to train, it may be something to actually work on but I still think it is to marginally extend the basic capability, the more special cases you get into the more error evaluations you get and that sucks and can't be used. I still think it's funny to try to mimic brain work in software:-) I had a 2001 experience with UPS yesterday my package was at a UPS point in Germany tracking said not in Sweden and do you know how many times you need to tell a very limited 3 answer phone bot that it's stupid in their system until it hands over to a humanoid? 5 times in their system quite unacceptable. In a global courier, can the office in US handle packages in Europe when y call them. No only US packages not China-Sweden... In reading the tracking number to robots and humans more than 10 times, who wins understanding my Swinglish number reading? Robots like 100% better. Oh my this brand new world....... Never a robot courier again.... for me. And my easterly friend gave me wrong tracking....... Package came delivered to the door by the Swedish mail service after lunch today...... The world is wonderful!
__________________
Good Vibrations since early 1950's, first patented vibrometer 1956 in the US.
http://www.vtab.se
Curran919

Avatar / Picture

Sr. Member / Supporter
Registered:
Posts: 589
Reply with quote  #23 
Quote:
Originally Posted by OLi
We haven't got consistent and good data in 1 place for the neural to train


Precisely. I work closely with the AI/ML guys in my company. Its not all hogwash, but their outputs can only be as good as their data, which is always shit. Unfortunately, they need to sell the products to get better data to improve their algorithms. Its a chicken-egg problem, so any AI to this point has obviously been shit. The biggest problem is the marketing, but once you can see behind the marketing curtain, you see there is some real benefit to it (in the future). Because lets be honest, in 50 years, most of our tasks will be automated. Just not today...
Sinski

Avatar / Picture

Sr. Member
Registered:
Posts: 393
Reply with quote  #24 
Quote:
Originally Posted by MarkL
Funnily enough, I was listening to a podcast this evening with a guy from Texas works for Symphony azimadli. seems that symphony are more into automated DCS and process optimisation, the AzimaDli bit is planned to be an add on to that.
The 'skilled people getting replaced with computers' debate came up and thankfully he was saying they want to augment what guys have, and still have a human element in the mix, he mentioned a few disasters with efforts previously attempted to have it completely automatic.

https://accendoreliability.com/podcast/rrp/rrp-067-using-ai-optimize-processes-don-doan/

The link to the episode is at the bottom of the page.



I have no problem with expert systems to save time and cherry-pick 'low hanging fruit' and let us concentrate on the more complex issues.


I think this is pretty much what Emerson want to do with CSi and I think also what Rockwell wanted to do with Entek. 

The next version of MHM/AMS was meant to be integrated in with Emersons systems but it looks like that was not successful so the next version is still standalone.
OLi

Sr. Member
Registered:
Posts: 1,861
Reply with quote  #25 
Lucky you :-)
__________________
Good Vibrations since early 1950's, first patented vibrometer 1956 in the US.
http://www.vtab.se
RustyCas

Avatar / Picture

Admin
Registered:
Posts: 1,791
Reply with quote  #26 
It’s called “a complex waveform” for a reason. But even simple rules are not so simple. Is that change in 1X a change in balance, a fastener that is loosening, a defect in the frame that is moving Fn towards the running speed? In my opinion, those all look the same, but that is my “gut feeling” - I suppose I’ve never really studied the differences in the data, because I’m always standing right there when the data is collected, and I have the answer to those questions before I walk away.
__________________
"The trend is your friend"
OLi

Sr. Member
Registered:
Posts: 1,861
Reply with quote  #27 
Yep, so does the rule based system  you may hold in your hand also, as long as you provide the speed........ and it is within the basic simple problems that it normally is.... ;-)
__________________
Good Vibrations since early 1950's, first patented vibrometer 1956 in the US.
http://www.vtab.se
MarkL

Avatar / Picture

Sr. Member
Registered:
Posts: 1,100
Reply with quote  #28 
Quote:
Originally Posted by RustyCas
It’s called “a complex waveform” for a reason. But even simple rules are not so simple. Is that change in 1X a change in balance, a fastener that is loosening, a defect in the frame that is moving Fn towards the running speed? In my opinion, those all look the same, but that is my “gut feeling” - I suppose I’ve never really studied the differences in the data, because I’m always standing right there when the data is collected, and I have the answer to those questions before I walk away.


Rusty, Don't get me wrong, I know most of what I'm going to write for the machine before I walk away from it, but I see what you mean if a less experienced guy was thrust such a meter in his hands and went at it, could come back with a lot of mis diagnosis.

OLi, any Idea why the locked coupling diagnosis was such a 'popular' result on the nspectre system?
fburgos

Sr. Member
Registered:
Posts: 644
Reply with quote  #29 
Wish I could diagnose a locked coupling only by the spectrum
Vibe-Rater

Avatar / Picture

Sr. Member / Moderator / Supporter
Registered:
Posts: 1,380
Reply with quote  #30 
I'm sort of thinking a locked coupling would produce a high 1x. And mayby a 2x etc.
Previous Topic | Next Topic
Print
Reply

Quick Navigation:

Easily create a Forum Website with Website Toolbox.