Artificial intelligence (AI) has powered quality assurance systems forward over the past few years and is helping food manufacturers detect more and smaller foreign objects while maintaining throughput.

“Artificial intelligence has become a key component of quality control in the food production industry,” said Daniel Greb, head of machine vision, Schubert. “Artificial intelligence fills a gap where algorithmic approaches often fail or are limited. It can perform tasks where algorithmic approaches cannot. An algorithmic approach usually requires measurable parameters to make a qualitative assessment. However, there is often a lack of correlation between different error classes. With the help of artificial neural networks, these correlations can be trained to a particular network. This makes quality control much easier.”

Metal detection software and technology have helped speed up processes, said Craig Lorei, global market manager, light industry, Eriez. 

“Advancements in metal detection software and technology have enabled us to detect metals on faster conveyor belts and identify smaller metal pieces, even when buried deeper or when using larger detector openings than possible in the past,” he said.

AI is continuing to revolutionize the way the food industry inspects food, said Todd Grube, product manager inspection systems, Heat and Control..

“AI can be used to improve order accuracy, to reduce overall waste, to enhance safety and to comply with regulatory guidelines,” he said. “Data from detectors, sensors and X-rays is analyzed to make real-time adjustments on the processing line.”

The processing power for vision systems has gone through the roof, said Andrew McGhie, business development director, KPM vision inspection, KPM Analytics.

“Where we might have been measuring seven attributes on a bun when we first started in vision serving bakeries, now we can measure 40, 50 things on a bun,” he said. “The lines are getting more throughput as well so we can analyze 1,600 buns or 10,000 biscuits a minute, no problem, and we can make a decision within 2 feet if an item is good or bad. And if it’s bad, the inspection system can reject it off the line.”

For vision systems, AI has allowed companies to look at items a bit more holistically, McGhie said. For instance, traditional systems often look at bake color, width and height of products, but AI can also look at shape and overall appearance. In other applications such as pizza toppings, inspecting similarly colored toppings, such as pepperoni slices, tomatoes and tomato paste, can be distinguished and measured separately to ensure the right quantity and distribution. It’s also helping to identify foreign objects that systems could not identify earlier. 

“For example, clear plastics and foreign materials similar in color to the product have always been difficult to detect on a traditional vision system,” McGhie said. 

But AI-enhanced inspection systems can be taught to identify it as a foreign material.

Technology is making vision systems easier to set up and simpler to use. Additionally, the vision system can communicate with equipment along the production line. For instance, if hamburger buns are too brown as they come out of the oven, the system can turn down the temperature or make other adjustments in the oven to get them back within their set parameters.

“AI is also being used to help simplify product setup using images of good products,” McGhie explained. “This learning from these images can provide a much faster, easier setup. The 30 or so attributes that are being measured can be set up with minimal operator intervention.”

And AI can identify smaller items like low-density plastics, which have traditionally been hard to see, said Kyle Hermes, vice president, TDI Packsys. 

“X-ray with AI is now able to identify small items consistently but at a miniscule scale that our other programs and software wouldn’t be able to identify,” he said. “There’s a baseline 50% improvement across the board on every material type. It has completely revolutionized what can be detected or not detected.” 

Hermes said X-ray systems will always have some false rejects, but AI can improve this problem, especially when it comes to certain challenging products, like those with a small metallic piece of packaging or a product with a desiccant package or another component that could throw off an X-ray system.

“With AI, we don’t really encounter any of those issues because we can teach it, for lack of a better term, what the good product looks like, and it can become accustomed to that,” he said.

Tying various quality assurance systems together becomes easier with AI, which leads to strong food safety programs, said Eric Garr, regional sales manager, Fortress Technology.

“Intuitive data management and the accessibility of AI now makes integration of any of these four inspection technologies — X-ray, metal detection, checkweighing and vision — possible,” he pointed out. “When integrated into a single system, this synergistically enhances the performance of each technology.”

He added that Fortress’ system reviews, collects data and oversees the performance of multiple metal detectors, checkweighers or combination inspection machines connected to the same network.

Mark Friesen, global director of marketing, Bunting Magnetics, said that AI has helped improve automatic setup and product learning of systems, simplified reporting processes, improved monitoring capabilities and facilitated simplified final product inspection.

What’s coming next

As the technology that runs various quality assurance systems advances quickly, bakers and snack makers should stay tuned in the next few years for more progress in this field.

“The future of our global food system relies on transparency, traceability and data-informed decision-making,” Garr said. “This will inevitably accelerate the adoption of digital recordkeeping technologies. AI, especially when a data center is completely integrated into a single system, takes machine-learning, rule-based algorithm technologies and makes sense of the data the machine has collected. It enhances human intelligence and adds greater scientific input, assisting the actions taken to respond to issues rather than operating independently or replacing human decision-making.”

Greb said that new sensor technologies will become established in the food production industry.

“The areas of SWIR (short wavelength infrared) and radar sensor technology are currently developing rapidly and are becoming more attractive, especially from a financial perspective,” he said. “Furthermore, neural networks will continue to establish themselves and increasingly replace traditional algorithms.”

Advances in vision are going to make detection capabilities better, McGhie said.

“Those advances are not just in AI but also improvements in the resolution and speed of cameras, improvements in processing power to analyze the images,” he added. “Other advances in vision include hyperspectral imaging, which provides increased capability to detect foreign materials and other defects we previously weren’t able to reliably detect. The price of this technology is also falling, making it more accessible.”

In the next five years, bakers and snack producers can anticipate advancements in hardware that will enable the processing of more data without requiring additional equipment, Goasmat said.

“Another evolution would be achieving full integration with all equipment on the line, enabling the system to detect defaults and automatically regulate, ensuring self-correction,” he said. “Furthermore, it remains imperative to prioritize secure data storage. Given that bakers and snack makers entrust experts with their data to ensure seamless operations, it is essential for suppliers to furnish a robust privacy framework to safeguard this information effectively.”

The next big push is going to be dual- and multi-
 energy, Hermes pointed out. It’s a new approach to using X-rays in determining different types of foreign materials.

“Dual-energy utilizes two different distinct wavelengths of X-rays, and each one will pass through or get stopped by different materials in a certain unique way,” he explained. “By blasting those through at the same time, and comparing how the product stops certain wavelengths in certain ways, we’re actually able to determine the composition of it. If it has a different molecular composition, we’re able to detect it now.”

Friesen said to look for improved detection accuracy through advanced algorithms and enhanced data integration and analysis capabilities with AI. He also sees more compact and energy-efficient systems, better detection of smaller contaminants and greater automation and remote monitoring systems coming in the near future.

Bakers and snack manufacturers have many quality assurance systems to choose from. One system or many combinations exist to help them keep their products safe from foreign contaminants. They must assess their risks and products then decide which system or systems work best for their operation.

This article is an excerpt from the August 2024 issue of Baking & Snack. To read the entire feature on Quality Assurance, click here.