Predictive monitoring is a game changing concept in almost every industry.  The ability to accurately predict is so valuable — to so many — that entire businesses are built around helping other industries, especially industrial manufacturing, with everything from remote asset monitoring, to predictive maintenance, to asset management.  But what can be predicted? How can it work within existing infrastructures? And what will it take to make a reality?

Predictive monitoring: using cognitive capabilities to “see” the impossible

Predictive monitoring for industrial manufacturing provides a distinction from another sought-after goal — which we will address shortly — predictive maintenance. In the former, using remote asset monitoring, systems allow a company to monitor and observe the behavior of data points over time.  The goal is to determine whether something needs to be done and what that task is —  not when to do a prescribed task.

Predictive monitoring uses AI to notice what would otherwise be impossible to see. It does not perform any correlation or offer any insight. The user is merely alerted to the behavioral change and applies their knowledge of the process and their own skillset to determine what action, if any, is needed. The critical difference is that the user prescribes the action.

Predictive maintenance: more than business as usual

In contrast to predictive monitoring, predictive maintenance is an approach that uses specific data and certain algorithms to determine the best time to perform maintenance. The goal is to perform the maintenance when it is necessary, not just simply at regularly timed intervals.

Think of it in the context of changing the oil in your car. Some manufacturers suggest that you change the oil in your car every 3,000 miles because that is easy to understand and easy to measure. Based on driving habits, that could mean every three months to every year.  However, that might not sufficiently account for driving conditions that could affect the quality of the oil.  Some people choose to change their oil every few months rather than use miles as an indicator. Again, based on driving habits and driving conditions, that could be too often or not enough.

Sticking with our car analogy, predictive maintenance methods would track information about a car’s engine condition, the driver’s habits, and driving conditions. That information is then used to determine the best time to change a vehicle’s oil. The benefit is that the driver spends just enough money to achieve optimal maintenance.

In manufacturing, predictive maintenance algorithms utilize historical performance, together with continuously updated factors such as age and frequency of maintenance. The application correlates all of the information to drive better control of the maintenance process, rather than simply adhering to predetermined schedules.  There are pros and cons to this approach, however. In general, it makes for better decision making, reduced costs, and more reliable assets.

Applying predictive monitoring and maintenance

When it comes to plant reliability, remote asset monitoring and asset management capabilities are critical because equipment ages, processes change, technology changes, people move on, and machinery keeps moving. As a result, things fail.

That brings us back to predictive monitoring designed to identify an event about to happen while also utilizing expert skillsets. Although this approach requires a human intermediary, the breadth and depth of visibility via AI yields a massive amount of insight. Predicting with AI models is a valuable thing, once you’ve trained the model and accounted for every possible variable. Until then, humans using AI-level insight can see things coming in a more truly predictive way.

Even if an event occurs quickly, or suddenly, predictive monitoring tools and resources allow support teams to isolate root causes faster than they would with other approaches. Even if predictive monitoring does not prevent a downtime incident, the approach expedites the return to uptime.  When an hour of downtime can cost well over $100,000, being able to respond and resolve incidents faster can dramatically decrease costs.

Through leveraging what AI has to offer, manufacturers can take advantage of remote asset monitoring at scale even though the platform does not understand all the pre-defined set points. Human analysis of AI-driven insights makes it valuable and is critical to helping AI learn the when, what, why and how. Predictive monitoring has tremendous potential to deliver production reliability results in ways distinct from current SCADA technologies or other implementation-intensive software solutions.