The effect that artificial intelligence (AI) will have on manufacturing will be immediate on two levels. The first is that events such as equipment going offline because of a part failure will be increasingly rare. The second is that the days when manufacturers ran short of a critical raw material, or conversely, had too much of that material on hand, are coming to an end.
There are already plenty of examples where manufacturers are employing machine learning algorithms and advanced analytics to more reliably predict when a specific piece of equipment is likely to fail. Instead of routinely replacing equipment every few years regardless of its condition, machine learning algorithms enable manufacturers to have a much better sense of the condition of, for example, an airplane engine.https://o1.qnsr.com/log/p.gif?;n=203;c=204663295;s=11915;x=7936;f=201904081034270;u=j;z=TIMESTAMP;a=20410779;e=i
That predictive maintenance capability, as it turns out, is one of the highest use cases for AI technologies as manufacturers seek to reduce downtime, says Sumit Gupta, vice president of AI and machine learning for IBM Cognitive Systems.
“They can reduce losses by applying predictive maintenance,” says Gupta.
Predictive maintenance applications now employ both visual and audio algorithms that watch and listen to machinery, says Gupta. The audio that is captured is then turned into WAV files that can be analyzed by algorithms to identify anomalies, says Gupta.
In fact, having that level of deep insight into the condition of equipment is leading many manufacturers to experiment with new business models. Instead of selling equipment to customers, they are essentially now renting that equipment for a fixed number of hours. The end customer incurs much less cost up front, while the supplier of a piece of equipment bets that recurring revenue generated over a multi-year service contract will wind up being more profitable.
The second major impact AI will have on manufacturing will take a little more time to unfold. Most manufacturers attempt to delicately balance the amount of raw material they have on hand versus projected demand. The number of times products are either out of stock or delayed suggests this process is an inexact science within most manufacturing companies.
AI models based on machine and deep learning algorithms will enable manufacturers to more precisely know not only how much of what material to order when, but also prioritize orders for their best customers whenever there is a projected shortage that might be caused by a geological event halfway around the world or political unrest in a country where a key supplier is located, says Monte Zweben, CEO of Splice Machine, a provider of a data warehouse platform that comes embedded with machine learning algorithms.
Zweben notes that it’s common for salespeople to make promises based on the perceived availability of product or what it has historically been. But all too often, events beyond the control of the manufacturer can have a major effect on the supply chain. Analytics powered by machine and deep learning algorithms make it possible to model the impact of those events on the manufacturing process, says Zweben.
“It’ll be like having a crystal ball,” says Zweben. “For the first time, manufacturers will be able to see around corners.”
It’s this capability that will ultimately drive a fourth industrial revolution that will result in manufacturers that fail to keep pace with AI innovations being either acquired by rivals or simply one day closing, says Zweben.
The issue that many organizations will need to contend with is that upfront costs associated with investing in AI technologies are considerable. Not only do organizations need to find some way to plug the right algorithms into their applications, those algorithms are not very useful when applied against a small amount of data. That latter issue will eventually push AI-manufacturing applications into the cloud, says Ramchand Raman, vice president of product development for Oracle. A new Oracle Adaptive Intelligent Analytics for Manufacturing cloud service is intended to enable manufacturers to take advantage of machine learning algorithms without having to make a massive upfront investment. Pricing for the service starts at about $3,000 per month.
“We want this service to be very accessible,” says Raman.
Specifically, the Oracle cloud service applies machine learning and AI models to layer advanced predictive analytics on top of all the data being collected to identify yield, defects, scrap, rework, cycle time and the costs associated with specific processes. The cost associated with trying to develop that capability from the ground up is beyond the financial reach of most manufacturers, notes Raman.
But not everybody thinks manufacturers will want to move core manufacturing process themselves in to the cloud, especially in the midmarket.
Scott Hays, vice president of product marketing for Epicor Software Corp., a provider of enterprise resource planning (ERP) applications for the midmarket, says that reluctance doesn’t necessary preclude smaller manufacturers from taking advantage of AI, but it typically requires someone in that organization to become a change agent.
“It’s more about that individual person than a specific job role,” says Hays.
In general, manufacturers around the globe have seen increased sales as the overall economy continues to slowly expand. Hays notes that expansion of manufacturing in the U.S. especially is much higher now than most anyone would have anticipated as little as two years ago. Some of that revenue will eventually get reinvested in the business. At the same time, IT vendors from IBM to Intel are focusing on making it simpler to invoke machine and deep learning algorithms within a broad spectrum of applications by embedding support for them directly within a wide variety of processors. As those capabilities become more widespread, the cost and complexity associated with building AI applications should continue to fall.
In the meantime, manufacturers would be well-advised to think about AI far beyond simply investing in additional robotics on the manufacturing shop floor. Companies that can turn data into actionable intelligence will enjoy a significant advantage over slower moving rivals. Whether that amounts to a fourth industrial revolution remains to be seen. But it is already apparent that AI is about to transform to a significant degree almost every aspect of manufacturing.