Deploying AI within Industrial Operations

Deploying AI within Industrial Operations should be fast, lean and valuable. For most, it’s not. The key may be the need to think differently when deploying it. 

In future posts I’ll go deeper into each aspect, but here are a few thoughts to get started… 

IT’S UNDENIABLE – DATA IMPROVES INDUSTRIAL OPERATIONS 

There are insights buried everywhere in the data from Industrial Operations.  Unlocking those insights and using them to mentor teams to drive sustainable improvement is at the heart of Industry X.0.   

For decades we’ve been breaking supply networks into small, digestible siloes of disparate data.  It seems completely logical that we’d need to reconstitute those networks and harmonize them in order to apply advanced tools like AI that pour thru the data from end to end and top to bottom. We’ve never been closer to unlocking what my friend Marco Annunziata theorized was ten to fifteen trillion dollars in benefit from exploiting the IIoT.  But that prediction was from over 7 years ago, and in some cases we seem as far from that goal as ever.  

“Innovation has always been the single most powerful ingredient to help us create more with less…”  Link 

IF WE ONLY HAD ONE MORE STANDARD…  

Maybe we need just one more universal standard for data connectivity?  Or a common asset hierarchy, or a digital twin of a process or machine all visualized in a new type of dashboard? 

“It’s like déjà vu all over again.” Yogi Berra 

Think of it this way. What if, in order to stem the advance of a new virus, we took all the travelers coming thru the world’s largest airports, and applied the same approach used to deploy AI within industrial operations.  First, we’d establish a connection to each human/device, determining their language and dialect. Then employ a team of translators to interrogate them.  How do you feel overall today? Warm? More or less tired than usual? You’re out of breath, is that typical?  I’ve now got a large body of data, but sometimes feeling “warm” is an emotional state as opposed to physiological, so we’d have to clean up the data before proceeding. 

Of course, this approach sounds absurd. Why not simply check body temperature and heart rate – baseline data – to profile those in distress.  The approach I described above is precisely what we do when we accumulate signals from an endless variety of what look like unique devices, machines and systems in our facilities.  

WHAT IF THE TRUTH IS HIDING IN PLAIN SIGHT 

Discrete, batch or continuous – manufacturing and industrial operations have patterns and rhythms within the data at the lowest level.  Those patterns tell us everything from cycle times and wear patterns to product families, but most importantly what healthy and unhealthy behavior looks like.  When properly applied, analytics and AI will detect the emergence of behavior that if left unchecked will manifest itself as downtime, quality failures or loss of performance.   

“THINK DIFFERENT”  Source

• The ability to start with the data you already have, as-is, where-it-is, reduces risk and cost, and accelerates time to value. 
• Insights from your existing data should guide the next best action – for everyone from an operator, to an engineer deploying new sensors or tech, to the COO. 

• Tools like AI should allow you to see your operations in a new light. 

You already have the data. Use it. Today. 

TAKE ACTION

If your industrial AI efforts aren’t scaling, don’t produce valuable insights within thirty to ninety days – with a lean, low risk deployment – you might want to see how Praemo’s technology, Razor™ can help. 

For more on transforming your industrial operations and driving new levels of performance, follow me on LinkedIn. 

To learn more about Praemo reach out or follow us on the social media channels below.