Process control systems have been evolving for centuries, from the ancient Greeks through the Enlightenment and the industrial revolution, and they are still under improvement today. Recent innovations in technology have dramatically speeded up this evolution, and manufacturing and process control systems have become far more efficient than ever before. The AI era brings a new and more powerful control layer that enhances existing control systems without replacing the current systems.

The history of manufacturing control

For a long time, process and manufacturing plants relied on maintenance schedules to periodically check on equipment and machinery in order to prevent expensive breakdowns. Drawing up a maintenance schedule required a combination of science and advanced guesswork. Engineers would combine observations from monitoring devices and historical data that showed how often breakdowns took place, and used those to make informed estimates about maintenance periods.

Although this kind of predictive maintenance used hard data, it was often gathered manually in the field and then brought to a control room to be collated and analyzed. It was an inefficient process. Eventually, the distributed control room arrived, using graphic display dashboards which aggregated and analyzed data from multiple different sensors and controllers.

The arrival of IoT sensors provided a new wealth of data which allows for far greater refinement in maintenance scheduling. It became the norm to set alerts when data exceeds a safe threshold, giving control teams advance notice about when a breakdown could be imminent. With these insights, maintenance schedules can be less rigid and more dynamic.

The drawbacks of static thresholds

These alerts are highly valuable, but they still leave much to be desired. They rely on static thresholds, which create almost as many problems as they solve. They tend to be set either too low, in which case you’ll be flooded with false alarms that quickly create “alert fatigue” and lead your team to ignore them, or too high, so by the time you discover that there is a problem, it’s already too late to prevent damage to the item.

Static threshold alerts are very effective when you have a mostly static environment where little changes from day to day or month to month. But in a plant, there’s no real “normal” and conditions are constantly changing. Finding the sweet spot for a static threshold that alerts you early enough to be of use without sparking false alarms is close to impossible. Even if it were possible, it would require a great deal of knowledge, and until that knowledge is acquired, the thresholds will invariably be wrong.

Enter the next phase: AI-driven control layers.

Seeing the future: AI control layers

An AI layer on top of the traditional control room system can constantly gather, process, and analyze data. Machine learning (ML) software uses this data to record behavior patterns and spot error patterns, triggering alerts whenever it detects anomalies. AI-powered alerts adapt to match changes in the business and production cycle, allowing for seasonality, variations in raw materials, changing external temperatures, and more. SAM GUARD’s unique ML modeling adds clustering to understand events on a deeper level that allows it to reduce the number of false alarms while also delivering advance warning about true anomalies in the system.

AI alerts can crunch data faster and more effectively than a human-powered system, giving your control system extra sensitivity to spot the slight yet significant deviations in plant conditions that herald potential part failure. AI proactive alerts notify you about glitches before you have any sense that something is wrong.

AI is the logical next step in the evolution of manufacturing control systems. Plants can build upon their existing process controls by adding an AI layer on top to generate early, accurate alerts that can flag potential breakdowns or other issues before they become major.

For example, a phosphate plant started using SAM GUARD’s AI-powered predictive analytics as a layer on top of its existing control system.

SAM GUARD’s AI-powered predictive

The solution picked up unusually high temperatures in a pump oil bearing tank. While the temperature was high enough to justify an alert on its own, SAM GUARD’s noted that the oil level in the secondary oil reservoir was higher than usual and understood that this was relevant to the first alert.

The relation between the two alerts enabled the engineer to immediately understand the problem and take action. The engineer went directly to check the tank and found that the secondary oil tank had been overfilled, making it impossible for oil to flow from the pump’s bearing back into the oil reservoir.

If the high temperatures had continued, it could have caused a bearing failure that would have damaged the pump. Thanks to SAM GUARD’s targeted alert with contextual information about pump conditions, the engineer was able to instantly spot the cause of the problem and act to correct it, preventing a serious breakdown.

AI can enhance your control systems to new heights

AI doesn’t replace current control room systems; it comes to enhance them. Together with traditional control room systems and their static thresholds, AI can lower costs and reduce downtime for manufacturing and process plants by giving greater advance warning before a potential part failure.

The advent of AI and ML brings a new layer that can enhance manufacturing controls and reduce costs for process plants, serving as the logical next step in manufacturing control.

To learn more, ask for a demo: sales@precog.co