Don’t Rely on Artificial Intelligence for Accurate Condition Monitoring

by | Nov 10, 2023 | Condition Monitoring

No computer software developed to date can beat hands-on machinery domain expertise when it comes to condition monitoring and evaluation.

Artificial Intelligence and Machine Learning have been heavily promoted as the next generation of general condition monitoring technology.  But like the dot-com bubble that burst 20 plus years ago, it’s starting to look like AI in the condition monitoring space is soon to follow that same path.  Hundreds of millions of dollars have been invested, with no clear benefit from the resulting products at a price point that would lead to profitability any time soon.  (sound familiar, dot-com investors?)  AI in the general machinery condition monitoring space has yet to prove itself to be better than an experienced, human domain expert at identifying and diagnosing problems and recommending the appropriate corrective action.

The dirty little secret that few of the AI providers in the space will tell you is that the widely boasted success that comes from their trial periods – the ones that claim 100+% ROI within the first 3-4 months of the trial period – rely on human data evaluation during the trial period and not their AI solution!  The typical equipment selected for the trial is most often machinery with known – but yet unsolved – problem histories.  This makes perfect sense because a proof-of-value trial on machines with no current problems would likely not result in the needed ROI to justify further investment.  Any decent machinery domain/condition monitoring expert should be able diagnose the problem and recommend appropriate corrective action on-site within a couple of data collections and not need to take three to four months to remotely figure out the problem and its solution.

With a program reliant on cloud-based AI, especially one where its data center and primary team are located in lower cost international regions, how does the case where Operations wants to make a phone call to a condition monitoring analyst because in their mind, “The XYZ machine sounds funny so you need to come out here and check it!” get addressed?  How will “the cloud” ask these questions that a machine domain expert would, to help understand why the machine “sounds funny:”

  • What did you just do to it;
  • Was any maintenance or inspection recently performed; and
  • Have there been any operational changes with this system, such as:
    • Recent speed change;
    • Recent product, load or process flow change; and
    • Recent cooling or temperature control change?

AI has been the new favorite marketing buzz-phrase for the last few years, with just about every industry jumping on the bandwagon.  AI is promoted as being an element of every part of our lives, from dating Apps through our refrigerators automatically ordering more milk from the on-line home delivery grocery store and even to lawyers looking to spend less time writing their closing arguments.

AI in the condition monitoring space typically focuses on a single technology as though that one technology can provide all of the answers to the machine’s operating condition.  Those of us who have been in this industry for more than a few years know that this just isn’t true.  No single measurement or observation technology can provide an indication of all of the possible failure mechanisms that a machine can undergo at the earliest possible opportunity to provide the maximum amount of time to plan and schedule the corrective action.  Further, no single technology can always differentiate between true physical defects and changes in the data that result from operational variations or recent maintenance activities.

Nearly all of the condition monitoring software providers are developing and promoting AI/ML applications, with spins unique to the markets they sell to. But these systems are appropriately narrow and limited in their scope, as they should be.

This is why ATS employs multi-technology condition monitoring analysts who have machinery and system domain expertise26..  Our Team fully understands how operational factors and recent maintenance activities can challenge machinery and how to recognize in the data when machines are at risk of developing faults that if unaddressed would lead to functional failure and unexpected downtime.

Don’t be fooled by the marketing hype promoting Artificial Intelligence in the condition monitoring and evaluation space.  Contact ATS, at 1-800-845-5168 or to discuss how we can apply our machinery domain and condition monitoring expertise to help you meet your reliability improvement objectives.


Subscribe to Our Blog

We regularly write a blog about Predictive Maintenance and the technologies used in PdM programs. Give us your email address and we’ll make sure you get a copy in your inbox each time we publish a new post.


We also post each blog article on Twitter and LinkedIn.

 Follow our LinkedIn Company Page      

Related Blogs


Submit a Comment

Your email address will not be published. Required fields are marked *

The reCAPTCHA verification period has expired. Please reload the page.