It’s Not Just Hulking Robot Welders
Sure, when you think of manufacturing and Artificial Intelligence (AI), it’s always easy to picture automotive assembly lines, replete with arc-flashing metal giants that Henry Ford certainly wouldn’t recognize. AI is so much more than that—if that even qualifies as AI!
Consider: Robot Welders and other automated tools have no innate intelligence. Ever since we invented PLCs (Programmable Logic Controllers), we have been automating processes, but it is by programming that we put our machines through their paces.
A canning plant (peaches, dog food, motor oil—it doesn’t matter) send thousands of cans per hour down the filling line. Most easily manage one completed product per second at the exit point (or more). One particular beer manufacturer has one plant capable of 60,000 units per hour, or 525.6 million cans per year, from a single plant. But it is not a smart process yet…
The Monkey Wrench
We’ve built some great machines called “descramblers” that are 99.99% accurate in placing all of the cans right side up on the conveyor, ready to receive product. For our example beer plant, that means 52,560 upside-down or sideways cans per year. That wastes product, production time, and directly causes downtime, increases cleanup time, and takes employees away from other tasks.
Relying on mechanical relays, electromagnetic sensors, laser detectors, or even computer-vision equipped with rudimentary pattern recognition makes this high level of accuracy possible. However, we will never reach 100% accuracy relying on analog systems.
AIs Learn on the Job
AIs, quite literally, cannot differentiate between reality and simulation. Using a video record of perfect operations for a particular manufacturing station, as well as providing every possible failure scenario, an AI can run those endlessly, isolating developing problems, pre-problem contributing factors, and problem resolutions.
Presented with a new error-scenario, it can interpret and integrate other experiences (cf. “think”) and find a new solution that solves the problem. That new solution is added to its database (experience memory) and can help to solve new problems, or even revise other procedures in light of the new information.
Human vs. Machine
Our brains are stunningly complex; AIs are comparatively simple. It may not seem so, but our brains are capable of trillions of interconnections and virtually unlimited storage space. AIs, on the other hand, have limited storage space depending on how much we’re willing to spend while we’re building them, and possess far fewer possible interconnections.
Our brains can handle such abstruse relationships that we might never replicate that in a machine. The machine, however, examines problems millions or trillions of times faster than we mere humans. Equipped with a massive, multidisciplinary database, an AI could examine and discard millions of possibilities in the time it would take us to reflect on a single instance.
Let’s say “baseball player” and “unscented deodorant”…and you think…what? You might think of celebrities promoting products. Perhaps it is that baseball players need more deodorant than the rest of us.
There might be just a handful of sports fans that would remember an ancient baseball pitcher who used so much of that product that it leaked through his shirt. He could then get it on his fingers to alter the flight characteristics of the thrown ball (a form of cheating).
The AI would find that relationship every single time—provided its database was comprehensive enough. AIs have the speed to look at every possibility, while we humans only look at the things in our own experience, and very slowly, too.
Using that Advantage
Defining artificial intelligence may be problematic, but using its capability certainly is not. Microsoft™ has a product called HoloLens™ which superimposes digital models in the real world. A useful device, in and of itself, but now they have incorporated AI into this product. You can zoom into your digital model on a 1:1 basis to examine it more thoroughly; you’ll be able to see interferences with moving parts; you’ll be able to identify structural faults that could result in unit failure.
You’ll be looking at the real thing before it is even built—but not just you—your entire engineering team can participate. Thanks to AI they don’t have to be in the same time zone, or even on the same continent. Being able to physically move through an enlarged model would let you spot potential defects in ways never possible before.
The AI would know where you’re facing and be able to duplicate what you’re looking at, so everyone else can see what you see. You can sketch in mid-air to illustrate your point, and that can be saved as a “layer” that others can add and remove at will.
It will change not just engineering, but the design of all sorts. You can teach an AI almost anything if you’re willing to commit the time and resources. Some can now write (bad) fiction, not-horrible poetry, and even paint some decent pictures. AI is going to insinuate itself into every aspect of our lives, including manufacturing.
Not Just Physical Manufacturing
Imagine an AI which is so clever that it:
• …schedules all your production, based on maximum performance of the machines, with the least impactful changeovers;
• …looks at which staff members are scheduled or present, with the greatest efficiency and efficacy for creating a particular product, and then making sure they are assigned properly;
• …automatically manages maintenance schedules so that downtime occurs when there is the least impact;
• …that handles material forecasting so that you don’t end up with a warehouse full of “stock” that takes space and ties up money;
• …that can communicate directly with your customers, telling them when their order is complete, and when it arrives;
• …that can coordinate all your deliveries, with real-time traffic awareness, knowing where your delivery trucks are at any moment, and being able to connect directly with the drivers to reroute for efficiency;
• …that can identify who your good customers are, and who is paying you a handsome $360,000 per year, but setups, changeovers, and slow payments mean you’re only making $18,000 (5% net) from them;
• …can even predict when a product line’s desirability is winding down, and steer you towards making more of what is selling, and even suggesting sales trends that haven’t even been recognized by humans yet.
All of this contributes to a competitive weapon that your peers (those who are slow to adopt AI) cannot match. But there is more…
ERP (Enterprise Resource Planning), MRP (Materials Requirements Planning), or any number of other programs, are designed to keep your warehouse empty. The whole idea with JiT (Just in Time) manufacturing is that your raw materials arrive when you need them, and don’t consume space that you have to pay for.
Humans will never be as good at predicting trends, understanding historical logistics, or accurately estimating needs. An AI can calculate actual needs, even applying a reasonable margin for waste or errors, and create strategies to reduce such waste.
Walmart™ has it down to a fine art. In a long narrow depot building, Receiving is on one side, and Shipping is on the other. When a truckload arrives, everything is offloaded and driven across the building to a waiting truck. Since everything is pre-labeled and identified by RFID tags, most material is in the building less than one minute before it is on its assigned truck. Properly implemented this can save you millions of dollars per year, depending on your volume of business.
A salesperson can send an order from the field via their tablet or smart device; they do that nowadays with regularity. They tell the customer that everything is fine and the order is scheduled…however…
What happens when Accounting gets a hold of the order? Do they notice that the customer is 90 days late on their payment? That they’ve exceeded their assigned credit limit? Maybe… But the odds are that they don’t know that there was a footnote in the financial pages about this company contemplating bankruptcy protection, or in danger of entering receivership.
AIs can know so much more about a situation than a human can track. Wouldn’t it be nice to know if a customer was a threat to your business before they did you any harm? The salesperson gets a red flag and must tell the customer that they have to speak to an account rep before the order can be processed.
It’s better for your bottom line if they catch up delinquent accounts before they can place an order, or that they pay in advance for any orders that exceed their current credit limit. Do you want to invest in materials and labor to make something that you’ll never be paid for, and that you cannot sell to someone else? And don’t forget two-way shipping costs, extra handling, repackaging, and more. Maybe you won’t lose all the money you invest in the product, but why lose any at all?
Early experience with AI has shown that manufacturing production capacity has increased by up to 25%, while material consumption rates have decreased by as much as 5%. These are not small or inconsequential numbers. They might be the difference between survival and failure of a business.
Implementing artificial intelligence should not be intimidating. Nowadays numerous companies are offering AIaaS (Artificial Intelligence as a Service) so that you don’t need to try to hire ever-so-rare Data Scientists. Our education systems haven’t produced enough—good luck trying to find one to hire—but AIaaS offers a solution because hundreds or thousands of companies can share the data scientists that already work for the AIaaS companies.
It’s cheaper than having an on-premises team; it’s as fast as if they were in-house; and, to top it off, there is no excuse to be still dawdling over this decision. The question you have to ask yourself is “Do you want to be the leader who sets the pace, or do you want to be the laggard who is struggling to catch up?”
If you want to learn more, we provide webinars to show you where the future is going and how you can get there. Don’t shortchange yourself. Join us today!