Predictive Maintenance: A Catalyst for the Next Industrial Revolution

Artificial intelligence’s transformative potential is a topic that sparks debate across industries. While its future impact is often discussed, the true scope of its influence remains complex and multifaceted.

“Artificial intelligence” can refer to anything from narrow, task‑specific algorithms to the speculative concept of artificial general intelligence (AGI)—machines that reason and act with human‑level intelligence. Hollywood has dramatized this vision in films like The Terminator, and public figures such as Elon Musk and Bill Gates have voiced concerns about its implications. Yet, when and whether AGI will materialize is uncertain. Gartner predicts decades may elapse before machines approximate human reasoning, even though narrow AI already outperforms humans in specialized arenas like chess and Go. Martin Ford, author of Architects of Intelligence, echoes this view, noting that AGI’s arrival remains unpredictable.

Similarly, the next industrial revolution remains elusive. While concepts such as Industrie 4.0 suggest that AI, IIoT, and related technologies could spark a new wave of productivity, Western productivity has stagnated for years. U.S. industrial production fell 0.1% in March—a dip that the Wall Street Journal described as evidence that manufacturing has entered a “soft patch.”

Despite these macro trends, numerous industrial firms are reaping tangible benefits by pairing IIoT with machine learning. FogHorn, a startup, helped Daihen, a Japanese industrial electronics company, eliminate 1,800 hours of manual data entry in a single plant. A leading beverage manufacturer saved the equivalent of one million beer cans by implementing Augury’s predictive‑maintenance platform, which fuses wireless vibration, ultrasonic, temperature, and magnetic sensors with ML to detect equipment faults. Augury co‑founder and CEO Saar Yoskovitz explained that identifying severe bearing wear on a filling machine allowed the brewery to perform a scheduled repair during downtime, preventing an eight‑hour outage that would have cost $200,000 in lost revenue.

Air Liquide’s digital arm, Alizent, demonstrates the power of Industrie 4.0 at scale. Leveraging OSIsoft’s PI software as the platform’s data engine, Air Liquide launched the SIO plant‑optimization platform a few years ago. According to Michael Kanellos, senior manager of corporate communications at OSIsoft, Air Liquide achieved a payback in just three months and a ten‑fold return in the first year. The company subsequently expanded the platform to manage “lights‑out” factories across France and Southeast Asia, eventually spinning Alizent into a separate unit to serve both Air Liquide and external clients.

White House Utility District, a leading Tennessee water and sewer utility, cut water leaks from 32% to 15% through digital transformation, yielding millions in savings. The district also reduced data‑management costs by $30,000 annually and postponed a $15 million plant upgrade by 11 years, boosting community reputation and credit‑rating agencies’ assessments.

Mark Willnerd, CEO and president of Toumetis—a machine‑learning firm for industrial applications—anticipates a productivity surge in the next five years. “Thanks to technologies like machine learning, we’re poised for significant gains,” he said, comparing the expected improvement to the productivity boom of the 1990s.

Toumetis is collaborating with an energy company on an oil field that could become one of the world’s top producers. Unreliable electrical submersible pumps have hampered output. By monitoring 1,500 wells and 100 data streams, Toumetis can predict pump failures up to 14 days in advance, enabling proactive repairs that maximize profit and production.

Implementing machine learning in industrial settings is rarely straightforward. Data inconsistencies—false sensor readings, missing values, and the need for extensive cleaning—can hinder analysis. Willnerd notes that we are still in the infancy stage of applying ML to industry.

A Bain & Co. study, Beyond Proofs of Concept: Scaling the Industrial IoT, surveyed 600 high‑tech executives and found that IIoT, especially predictive maintenance, is more challenging to deploy than expected. Nonetheless, the report concludes that industrial IoT remains a promising opportunity.

One core challenge is the requirement for a rare blend of domain expertise and data‑science acumen. Toumetis addresses this by hiring seasoned industrial experts who have worked in data analytics since the late 1990s and early 2000s. Willnerd emphasizes that success hinges on a clear understanding of the problem and its business value.

Yoskovitz echoes this sentiment, noting that talent shortages are pervasive. In one training session, the average age of attendees was 55, illustrating the generational gap. With older workers retiring and younger generations uninterested in maintenance roles, the U.S. risks losing invaluable legacy manufacturing knowledge over the next decade.

While it may be premature to declare whether IoT and AI will usher in a productivity era on par with the first industrial revolutions—or if Industrie 4.0 will resemble a software upgrade rather than a cyber‑physical systems revolution—organizations can still answer the practical question: how can these technologies deliver immediate value by ensuring the right technicians reach the right machines at the right time?