In the world of manufacturing, predictive AI is being billed as the next big thing with new products promising the Earth launching every week. In reality, it is marketing dressed up as progress. In one recent example, a so-called new AI product was based on hardware that dated back to the early 90s, the only difference being the use of a router rather than a cable. In effect, this is white noise and it is causing confusion for food and drink factory owners who are under pressure to raise productivity and ensure reliable supply to meet the demands of a growing world population. They cannot afford to get it wrong and have someone else eat their lunch.

In this article, I will challenge the key selling points of this predictive AI in manufacturing and show that if it is lubrication that machines need to increase their output, their operators ought to steer clear of snake oil. Instead, the answer to increasing equipment effectiveness can be found much closer to home, hidden inside the plant and machinery of virtually every factory in the world. This is what we discovered in our journey as a company which began in Britain’s manufacturing heartland more than a decade ago with the founding of engineering consultancy 53North.

The basic proposition of predictive AI in manufacturing is the application of sensors to machines with the aim of increasing the ability to predict and improve their performance. Sounds simple, right? Such sensors though are not new. They have been around in one form or another since the 80s. Most critical assets in factories already have them, and the mitigation they provide, so where is the sense in adding even more? Salesmen claim they offer greater insights. But in fact, many sensors are unsuitable for industrial applications and incapable of monitoring the right parameters to gather meaningful data. Most just don’t cut it.

A favourite buzzword of predictive AI evangelists is the ‘data lake’, a repository that ingests, stores, and allows for processing of large volumes of data in raw form. The theory goes that factory owners can compare the performance of similar assets. But in the real world, operating conditions vary wildly from place to place and such comparisons are of little use. Two pumps might look and feel the same, but they operate at different speeds with different flow rates through different pipe lengths. Simply, they do not have the same profile. Any criteria for predicting failure applied to the same classification of pumps in a data lake would lead to false positives, which are no good to anyone.

How then can factory owners get the accurate insights they need to increase the efficiency of their assets? The answer is machine learning, through the harnessing of millions of data points that already exist within plant and machinery. This is the data that is hidden inside the programmable controllers that automate and monitor industrial processes that, with cutting edge technology, can be captured and put to productive use. We are layering vast amounts of contextual data, such as speed or SKU, alongside operational data, such as temperature or vibration, in a machine learning platform to create a unique fingerprint with dynamic alarming for every component.

This is only scratching the surface of what is possible. With enormous datasets being added to every 10 seconds, operators can predict and prevent wear-related events in their factories with increasing levels of success. They can better understand the errors leading up to stoppages previously written off as random. As a result, they can reduce waste, improve quality and boost throughput. This is the future of AI in manufacturing and the sooner factories see it, the better for everybody.

We are not jumping onto the AI bandwagon. Our capabilities are based on deep knowledge gained in intelligent asset management since we founded 53North in 2013 in Dinnington, South Yorkshire, an area rich in manufacturing expertise. We built a client base including many of the world’s top food and beverage groups, a number of which are already deploying our machine learning platform in their factories. This summer, we completed the flotation of IntelliAM AI on the Aquis Stock Exchange Growth Market to raise £5m to invest in people and technology, roll out the platform even further – and silence the white noise.