It's all about the data
“AI is used to augment controls in robotics. In Industrial manufacturing, it’s used for predictive and prescriptive technologies to determine when a system is going to break for downtime, etc.”
This is a tricky question. Artificial intelligence can be used in robotics and vision systems most basically, and to control certain parts of a machine. The data, however, is most important for any AI system to work as it is supposed to and to its highest potential.
“AI is used to augment controls in robotics. In Industrial manufacturing, it’s used for predictive and prescriptive technologies to determine when a system is going to break for downtime, etc.,” explains Huschka. Vision is deployed across a large swath of industrial environments—quality control, safety, inspection and also in pick and place applications—usually as a piece of vision equipment, he adds.
Cory Knight, automation engineer, food and beverage segment at Festo, believes that AI is about data and digitalization. “You can predict, write algorithms and execute code all day long, but if you don't have any data to analyze then it's simply machine logic, not AI.” Cory says that companies first need to get the data to the algorithms in order to start deterministic or programming to determine what you need to do or how to improve the process.
An M-2iA delta robot using 2D iRVision® and iRPickTool line tracking software to pick randomly oriented waffles from an infeed conveyor and place them on an outfeed conveyor. Next, an LR Mate 200iD/7LC clean room robot picks the group of product from the conveyor and places them into a cardboard box to simulate packing. Photo courtesy of FANUC
“What’s new is the way AI anticipates what’s coming, and the way AI can do it more quickly and even with a feedback into quality of future products coming down the line.”
According to Dr. Irene Petrick, senior director of Industrial Innovation, Internet of Things Group, at Intel Corporation, “AI is the use of data that is processed thru algorithms that are generally trying to find some kind of pattern or assess data based on an already existing pattern. It’s simply using data to drive insights into the real world.” Intel works on artificial intelligence algorithms for the software itself and how to collect data at the source, how to pick it and move it to wherever it’s necessary to process, and how to store it.
Lawson believes that intelligence at a system level is more important than at the machine level. There is a lot of focus on building smarter machines, but what we really need are smarter systems at a higher level. The machine level problems don’t need AI or ML; they just need good algorithms and good engineering and software, and you can solve a lot of those problems.
Part of implementing AI successfully is the learning process. If you're not going to write thousands of lines of code using closed loop feedback, you have to "educate" the system some way. “The learning is only as good as the information you give it,” Knight advises.
Use of AI in food and beverage manufacturing is limited at this stage. There are some companies behind it 100%, but most are still dipping their toes in the water.
One area that is taking off is crop harvesting, with Knight saying he has seen agricultural harvesting robots. One Festo customer has a solution that uses machine-guided vision for not only telling the robot where strawberries are on the plant, for example, but also grading strawberry ripeness and picking based on ripeness and where that strawberry is being transported.
“So they’ve connected it to the supply chain,” says Knight. “They can pick it based upon transport time to each destination.” This allows the company to pick strawberries days before they’re ripe to make sure it is ripe when the shipment gets to its destination.
Huschka agrees that AI use in crops is working well.
“AI is already leading in the global farming industry and weather prediction. When is the right time to plant crops? It’s making these macro decisions. What crops work best in the right time? And it’s providing insight at the beginning of the process. Farmers are attaching wearables to their cows that feed them data on how their cows are doing… whether or not they’re healthy.” Huschka thinks that this could lead to controlling greenhouse gas emission that cattle can cause.
AI is also frequently used in inspection equipment, which is not new in food and beverage manufacturing.
“What’s new is the way AI anticipates what’s coming, and the way AI can do it more quickly and even with a feedback into quality of future products coming down the line,” says Petrick. “AI comes in and makes visual inspection different than in the past. It plugs into difference pieces of the production process to reduce quality defects in future runs.”
“You can predict, write algorithms and execute code all day long, but if you don't have any data to analyze then it's simply machine logic, not AI.”
Left: The VTEM motion terminal by Festo is a smart valve terminal, which measures and controls pressure and flow and air consumption for every valve and offers predictive failure analysis. Right: The MSE6-E2M gathers air usage data for an entire machine, for each system. The intelligent module automatically monitors and optimizes the compressed air consumption of line equipment and can alert plant personnel to system leaks. Photo courtesy of Festo
“A couple years ago it was a lot of hype; everyone was excited and thought it was going to take care of a lot of things. And now, it’s not what everyone thought it would be. It’s got limitations.”
“Although we haven’t seen a massive impact in the industry just yet, we have seen AI being used to acquire raw materials more intelligently...”
Last year, A3 surveyed its members about their companies’ adoption of AI. More than 76% of respondents felt that AI will be important to their companies in the next three years. However, just 3.3% of those surveyed said that AI was being widely applied in their organizations. This has been called the Ambition-Execution Gap.
“Even if you get started, we have seen companies who struggle to move from the labs, so to speak, into real-world production. Scalability has been a challenge in some cases. And making it work outside of a more controlled environment,” Huschka cautions.
“I would say AI, machine learning, digital twinning… it hasn’t been fully adopted yet. It’s rare you see it,” says Knight. “You need to know who’s in the game and forward thinking, because it’s an investment. You have to figure out how to apply it to your process.” He adds that once applied, it offers the flexibility and modularity to take control of your process.
So, is AI a useful tool or is it all hype?
Most agree that it’s somewhere in between. “It’s not a magic wand; it’s not something that will solve all of your problems. But at the same time there are companies that are making use of—whatever you want to call it AI, data analytics, smart automation—to make a real difference in their businesses,” says Huschka. He adds that it takes the vision, know-how and finding the right strategic partners to make that dream a reality for your individual business.
There are a lot of buzzwords being thrown around regarding AI. “We have algorithms that make decisions, but they’re simplistic decisions. So is that AI? Yes, but it’s very basic. Because of the way models have evolved and the amount of data that’s now available, the AI software models and algorithms are so much more complex than ‘is this a defect or not a defect, yes or no?’” explains Petrick. Today, AI has gone far beyond that as it’s using a lot more data, visual, audio, text, numerics and combining them in ways—much like the human mind will do—and putting it together to answer much more complex questions.
Knight sees AI as a valuable tool, especially in food. “The big reason it’s highly important in food is the because of the variability and irregularity of food products. They are the most challenging product to process in the world.” That’s where Knight sees AI really making a huge impact.
Garrett thinks it’s both. “A couple years ago it was a lot of hype; everyone was excited and thought it was going to take care of a lot of things. And now, it’s not what everyone thought it would be. It’s got limitations,” he says. He believes it may come in 10 years, but not in the short foreseeable future.
Lawson agrees with Garrett. “While there are practical success stories out there, there’s also an extraordinary level of hype around AI and how it will revolutionize the manufacturing industry. This hype poses an issue in that it gets in the way of those providing real innovative technology to their customers.” FE
Intro video courtesy of Getty Images