Here’s How Manufacturers Are Leveraging AI

In case you hadn’t noticed, AI has become a rather hot topic lately. But despite all the hype and endless stories about how it’s changing the business landscape, there are many people in manufacturing–especially those in the small and medium-sized business category–who are secretly scratching their heads and wondering just how it applies to them.

If you’re in that camp, don’t panic. First, you’re not alone. And second, it’s still early days. According to a Q3 2023 survey of more than 100 manufacturing and distribution executives by Sikich, a global company specializing in technology-enabled professional services, just 14% have current AI implementations on their factory floors. Only 19% have any plans for factory floor AI use at all.

Still, you should be studying how others in industry are already benefiting from this breakthrough technology, and working to see where you might apply it to your own benefit. Below are a few examples of how AI and machine learning are already delivering benefits in industry.

But first, here’s a quick primer on AI. In his recent book, The Cloud Revolution: How the Convergence of New Technologies Will Unleash the Next Economic Boom and a Roaring 2020s, physicist, Manhattan Institute senior fellow, and faculty fellow at Northwestern University Mark P. Mills says it’s “…a shift to a new class of logic. It’s the shift from binary logic to inference engines, or so-called artificial intelligence… The recent maturation and now rapid growth of silicon engines based on inference, or learning algorithms, rather than calculations, signals a deep structural change… The phase change in the means of discovery through AI will have the double effect of both assisting data interpretation and enhancing data acquisition… As economist Alexander Salter succinctly put it, ‘Data doesn’t interpret itself.’ The machines are amplifiers. They don’t replace imagination.”

One of the biggest companies leading the charge on industrial AI is ABB. With more than 105,000 employees, the Swiss-Swedish multinational company has been around for over 140 years and has been a proven leader in electrification, motion, process automation and robotics. Peter Terwiesch, president of ABB Process Automation and a member of the ABB Group Executive Committee, has been heavily involved in their AI efforts. “I’ve been focused on our quest towards autonomous operations,” he said. “I stress ‘quest’ because we’ve found you can operate certain areas autonomously, but others require the human touch.”

ABB’s efforts involve several different areas of industrial operations. “Sustainability is a big one, especially decarbonizing,” said Terwiesch. “That involves everything from LNG to reshoring of manufacturing facilities. There’s also the imperative to be safer and more efficient. And one big area is definitely data. Most places it just gets stored and never looked at by humans or machines. That’s a treasure trove, because data can drive better decisions.”

One example of that involves a new reality for some operations. “We’ve all been taught that manufacturing requires stable power,” Terwiesch explained. “But now, with the decarbonization imperative, often the lowest marginal cost power producer is renewables that can be unreliable. How do you reconcile that? We’ve focused for more than 10 years on the integrated control of the process with the power side. Certain things have to run all the time, like compressors. Others have a built-in buffer, such as process heaters. Digital solutions can allow you to shed non-critical load in milliseconds while you protect the critical load. That’s one area where we see a big opportunity.”

Another area of opportunity is in emissions monitoring. “We’ve offered methane analyzers for quite a while,” said Terwiesch. “In the past they’ve been used in static applications or have been handheld, like those used to check a wellhead. Now we can combine them with AI and other technologies to have drones that patrol pipelines and drilling areas ‘sniffing’ for methane. We can detect the size and intensity of a leak, and drive improvements in safety, sustainability and economics. Tech that used to be for labs is now in the field.”

Another well-known name in industry, Fluke Reliability, a subsidiary of Fortive Corporation, has also been increasingly involved in AI. The 75-year-old company is a mainstay in preventive maintenance with its temperature and vibration monitoring systems. The company was already working on incorporating AI into its offerings when, in August, it acquired Azima DLI, a market leader in AI-powered vibration analysis software and subscription-based remote condition monitoring. Ankush Malhotra, president of Fluke Reliability, isn’t surprised at the slow uptake of AI in industry. “It’s a little bit the elephant in the room. Everybody’s talking about it, but people are still wondering how to get into it. Our customers have a need for expertise. That led us to look at Azima.”

Fluke combined its long history of industrial monitoring with Azima’s AI expertise and is now able to offer off-the-shelf solutions. “We’ve been able to assess 60 to 70 trillion data points,” said Malhotra. “From there we built a rules-based engine, so we know when a machine has a risk of failure, the root cause, and preventive measures. We’ve got 18,000 unique machines covered. We’re able to train the model quickly enough to see results in a few months.”

An example of those results is the company’s work with food, agricultural and industrial giant Cargill, Incorporated. “We monitor 15,000 assets for them in one of their divisions,” Malhotra explained. “We’ve been able to reduce their maintenance by 10%, reduce their downtime, and increase their machine longevity. The ROI is very clear.”

On the far other end of the data set size from Fluke is Amatrium, Inc., a solutions provider that uses machine learning, a subset of AI, to help small and medium-sized manufacturers eliminate waste with its custom tools such as Amatrium Process, a quality control tool that aims for scrap reduction, and Amatrium Predict, which can foresee the properties of a metal alloy based on its component materials, saving time and expense in the development process. And Amatrium does its work with little input data.

“About 500 to the low thousands of lines of data is what we typically see,” said Andrew Halonen in technical sales and marketing for Amatrium. “Material results are so equipment-related and raw materials-related, it’s imperative that we use the customer’s data as opposed to random data from other sources. The beauty of ML is that there’s no bias. Why not leave it up to the tool to tell you where the biggest impact is? You want to drive the highest profitability. Scrap is money. If you can identify what’s driving scrap, that’s a big deal.” In its work with one global foundry, for example, Amatrium was able to drive a 10% scrap reduction.

Again, even if you haven’t begun with AI yet, you’re not too far behind. But that’s going to change fast. “In 10 years, it will be standard practice,” Halonen said. “Today, only the early adopters are taking advantage of it.”

“How do we democratize the technology?” asked Malhotra. “It provides a level playing field. A customer can start with 25 assets, the biggest pain points, and see results almost immediately.”

“I couldn’t think of a more fun time to be in this industry,” said Terwiesch. “There’s tremendous excitement around the opportunities and solutions.”