Experience with IIoT Condition Based Monitoring and Predictive Maintenance Tools?

Question

I am really interested to hear from others on here about their thoughts and experience with any IIoT Condition Based Monitoring or Predictive Maintenance Tools?

It is an area that seems to be gathering pace in the industrial world with a number of acronyms and buzz words floating around such as:

Industry 4.0
AI (Artificial Inteligence)
ANNs (Artificial Neural Networks)
Deep Learning
Big Data Analytics
IIoT (Industrial Internet of Things)
Machine Learning
Digital Twins
Anomoly Detection

And I am sure there are many more! With all these buzz words floating about, it would be great to hear about how this technology is actually being used and what members thoughts and ideas are in relation to it?

I have thought of a few specific areas and questions that could be used for a basis of some comments......... (Please don't restrict comments to only these points below though!)

What tools are out there (hardware and software) - both in terms of:
- Development kits such as the Brainium /SMARTEDGE AGILE that is currently available for RoadTesting
- Or commercial solutions such as GE's Predix Platform or the Honeywell Connected Plant portfolio of products
Have you used any of these tools in industry?

How have you or would you use some of the cloud based machine learning tools offered by some of the big cloud providers such as Amazon or Microsoft Azure?

Does anyone have any 'Success Stories' where they have used some of these technologies to optimise machine maintenance and reduce unplanned breakdowns or even improve production yields and quality?

What security concerns are there or should be considered around using Internet connected devices in a production environment?

What are the advantages of using an IIoT based system vs an 'in house' system - eg. Do the AI tools available on the cloud help predict issues earlier and more accurately than a non cloud connected system?

How easy is it to set up an IIoT based solution? How easy is it to utilise the large volumes of data that can be collected and how do you go about setting up good machine learning models?

I look forward to reading the comments others have on this!

michaelkellett · Answer

According to Gartner (Hype Cycle), Deep Learning and Digital Twins are at the peak of inflated expectations. Edge AI is on its way up the curve. They don't specifically mention the other tags you give .

I would lump (for your purposes as outlined in your first line) all this stuff into Industry 4, leave it there and put my money on a horse (or better still keep it in my pocket)

Predictive maintenance is an old idea and perfectly sound, people have been linking up sensors into small local nets and bridging those to bigger nets for decades. This will continue. Every now and again the marketing muppets will come up with some new buzz words and the media will hype them to the stratosphere and then we'll all forget them again. This will continue as well.

Meanwhile proper engineers will go on putting one brick on top of another (or whatever analogy for common sense you prefer) and steadily improve things.

MK

Jan Cumps · Answer

I've designed a project around predictive maintenance - as Michael says without any of the inflated expectation options . https://www.hackster.io/jancumps/rolling-material-monitoring-51b6fc

When I showed it in Denmark on an industrial design seminar where I was invited, the experts that worked on predictive maintenance and industrial data collection found it under par. Good as a design - not good as industrial. Because many valid reasons. Some of them tracking back to PLC related logic I broke, and naive implementation in several corners. I learned a lot and made new friends.

There is a strong IIoT market though - not to be laughed away, but most of the partners don't show up in the maker universe. In particular companies that provide area-wide common heating (east Europe has many of those projects, several rooting from the communist times) and the likes use IIoT and wireless communication to collect data from the heating power plant to the distinct flats that they heat.

shabaz · Answer

Hi Tony! I can partially put colour on some of this stuff but not all; I'm not familiar with the products you mention, but have seen similar ones in use. Sometimes they are used for monitoring or information gathering, e.g. usage of areas of a building. Regarding improving machine maintenance and so on, this can be low-hanging fruit in some industries. One example is on ships, where apparently stuff like oil gets recycled (filtered and reused) and so there's important needs like checking that the filtering is functioning well. In that case, local processing is used, until a connection is available. Many firms that are rolling their own systems rely on cloud services to build their solution. Regarding cloud v non-cloud, local processing is important, to reduce the volume of data that needs to be send to a cloud, and speed things up. It's got different names, like edge or fog, and gives the best of both worlds. So, some problems are resolved by having more computing power locally. The device that's responsible at each site for connecting to the network ends up having just as much power as any other network element, and can be connected and managed securely, like any other business device. Some technologies are very cool (albeit a bit hard to use currently) like Amazon Greengrass, where in theory the same app can run in their cloud or locally, so no need for different code either; so with some approaches, code or functions run at the best place, with no need to consider where that is. As for setting up an IIoT system, some businesses use more PaaS and SaaS than others, it really depends. But that's not the whole story, since there is quite a lot of overhead configuring and deploying at scale if you try to go it alone. Whereas, a lot of the cloud providers will offer their own solutions to bring up systems and manage/update them (i.e. day-zero onward). The goal (which is possible with some systems) is to have something shipped to sites that anyone can plug in and it auto-connects, i.e. no trained person needing to install at thousands of machines or locations, and no time wasted manually configuring.

tonydbeck · Answer

Hi Michael, Thanks for your comment - you make some really interesting points! This is something I have wondered about......... there certainly seems to be a lot of emphasis on the buzz words I have mentioned in marketing material from some pretty big vendors - but from what I have seen, there is a lot of information and statistics on what IIoT kit can do but I have not seen a great deal of real use cases. I guess a lot of this is pretty new and it will take time to become more widely adopted in industry. Personally, I do see additional benefits with cloud connected systems - such as having more powerful tools available in the Cloud to create faster and more accurate predictions and perhaps detect process anomolies that could affect quality of whatever is being produced. But, it is equally important that the results of the data analytics done in the cloud are acted upon locally. It does certainly seem the case though that some of this tech is at the peak of inflated expectations! Will be interesting to see how industry uses Cloud connected systems over the next few years. Great analogy by the way of engineers will go on putting one brick on top of another and steadily improving things. Cheers ----------------------- Tony

tonydbeck · Answer

Hi Shabaz, Thanks for your comment on this. The only experiences I have had with condition monitoring systems up until now are systems with only local processing which do work really well, but it definitely sounds like there is a lot of potential for Cloud connected systems to bring additional benefits. Amazon Greengrass sounds really interesting - I had come across this before but have never really looked into it in any depth. The idea sounds great though - especially in a production environemnt where reliance on a cloud connection may not be ideal! Cheers ------------------- Tony

clem57 · Answer

tonydbeck Digital Twins applies to cloud and IoT... See Azure Sphere.

Experience with IIoT Condition Based Monitoring and Predictive Maintenance Tools?

Top Replies