Visit our Content Hub!
Access free downloadable content curated by our editors.

New Technologies Feed the Hunger for Information

Trying to keep up with productivity, quality, safety, sustainability, and other industry demands, food and beverage plants are turning increasingly to the innovations found in today’s sensor instrumentation.

Image-based scanners are displacing standard laser scanners, providing the capability to read 2D barcodes, which can be much smaller than traditional barcodes. Photo courtesy of Cognex.
Image-based scanners are displacing standard laser scanners, providing the capability to read 2D barcodes, which can be much smaller than traditional barcodes. Photo courtesy of Cognex.

The reference technology of Leuze’s DRT 25C enables the detection of packaging with a wide range of colors and transparency as a deviation from the conveyor belt surface. Photo courtesy of Leuze.The reference technology of Leuze’s DRT 25C enables the detection of packaging with a wide range of colors and transparency as a deviation from the conveyor belt surface. Photo courtesy of Leuze.The food and beverage industry faces plenty of challenges. Under constant pressure to create new and innovative products, manufacturers are dealing with increased variability in the raw ingredients coming into the plant. Operations need to see higher productivity while also remaining profitable. They need to stay competitive in local and global markets and protect their brands by ensuring food safety and product quality.

“This is all at the same time where it’s getting increasingly difficult to retain and find people,” says Ola Wesstrom, senior industry manager for food and beverage at Endress+Hauser. “Companies need to find innovative ways to leverage existing or new instrumentation to increase productivity while protecting their brand.”

Hungry for information, food and beverage producers and packagers are adding more sensors to the factory floor. They’re looking for additional process information, and they’re looking to get better control over their processes.

“The food and beverage industry is a little late in adopting sensor technology, but now clearly it does,” says Sabine Busse, managing director of measurement and analytics for ABB. A key driver is the fact that budgets are being stretched everywhere. “Sensors to some extent provide one of the most cost-effective ways to understand what is happening in my plant and how I can optimize.”

As sensors gain capabilities along the way, the bottom line remains the same: Control processes and control time, money, and waste.

Measure it in real time

As sensors proliferate, a key trend is the ability to bring more inline quality monitoring to operations. “We’re looking at where and how traditional lab measurements can be replaced with inline and at-line instrumentation,” Wesstrom says.

Inline measurements can speed up operations and reduce costs as well. In the food industry, there’s a 14-minute average hold time waiting for lab results to release product for further processing, according to Wesstrom. “That can have an impact on tank space utilization, processing time loss, and possible product quality,” he says.

Product variability between sampling could have an impact on the quality itself, Wesstrom also points out, adding that inline measurements could capture anomalies quicker.

In an Endress+Hauser webinar presented in April, Wesstrom ran through several examples of products and processes that would benefit from inline measurements—measuring oxygen in yeast fermentation, viscosity in batter and chocolate, fermentation rate in apple cider, salinity in brine, and many more. But he also made it clear that it should not be considered a replacement for lab measurements. “In most cases, lab measurements will remain the final quality or safety check,” he says. “We’re looking at how to supplement that to capture things in between or speed things up to increase productivity.” It might also be possible to reduce the number of samples being taken and make better use of those resources.

One type of analysis that is moving increasingly from the lab to the factory floor is pH measurement. Providing examples in measuring the pH of mayonnaise and ketchup, Wesstrom noted its importance for taste and consistency, as well as food safety. In this case, getting continuous measurement throughout the process could not only eliminate hourly sampling, but also provide continuous quality assurance.

One reason pH measurements are typically made in a cooled sample line or via grab samples is because of the hostile production environment. ABB’s new pH/ORP sensors have advanced built-in diagnostics to ensure accuracy and reliability even in harsh processes. Sugar refining is an example of a high-temperature, caustic process that requires accurate pH control at almost every stage. Used within the production process, the sensors help to reduce the lag time of lab measurements.

>>Learn how digital protocols are making it easier to get the information you need from sensors in "IO-Link Connects the Food and Beverage Industry."

Image-based sensors are also seeing a trend that brings more data acquisition from the lab to the production line, according to Matt Moschner, senior director of ID products at Cognex. Though not new in concept, 3D sensors, for example, are more able to work online largely because of an increase in horsepower of embedded computers. The smart sensors are capable of making decisions at the point of image acquisition, Moschner notes. “They don’t have to pass information to the cloud or to a PC,” he says. “They make decisions as the line is running and making parts.”

What this means in practical terms for the food and beverage industry is the ability to do quality inspection on every single item being produced or packaged, rather than typical piece parts (10 out of 100 samples, say). “We’re now able to do that on the line for every single part,” Moschner says.

The growth of edge computing helps to interpret all the images on the fly, enabling decision-making on the plant floor. Cognex Edge Intelligence stores information about the raw results along with 2D or 3D images to be analyzed later. But that data can also be analyzed in real time, Moschner explains, giving operators the information needed to make real-time decisions.

The Rosemount 708 Wireless Acoustic Transmitter can continuously determine steam trap status, including identifying steam trap failures and inactivity. Photo courtesy of Emerson.The Rosemount 708 Wireless Acoustic Transmitter can continuously determine steam trap status, including identifying steam trap failures and inactivity. Photo courtesy of Emerson.

Artificial intelligence deals with complexity

Those kinds of computing capabilities—along with more thermally efficient packages and lower price points—have also enabled a growth in inspection algorithms based on machine learning for more advanced tasks. Food processing has always been a quality control challenge because of the high variability. “That has really challenged our classic machine vision tools,” Moschner says. “The advent of machine learning can really be a breakthrough for applications that have either been complexity-prohibitive or cost-prohibitive.”

Cognex’s new In-Sight D900 embedded vision system, which features the company’s ViDi deep learning software inside an industrial-grade smart camera, is well-suited to optical character recognition (OCR), assembly verification, and defect detection.

Traditional vision systems struggle to check food packaging portion control, as well as presence of the correct food, notes John Petry, director of marketing for vision software at Cognex. “For example, training a rule-based system to spot the pepperoni on a pizza is a huge challenge. Simultaneously teaching it to distinguish pepperoni from ham is completely impractical in the real world,” he explains. “But with deep learning machine vision, this is becoming a reality, allowing manufacturers to provide fully automated 100% quality inspection.”

The trend seems to be moving toward artificial intelligence, agrees Jeremy Bergh, president, North America, for IDS Imaging Development Systems. “There are a lot of applications where a human eye can notice whether something is produce, for example, or apples. A human can look and say that looks good or it doesn’t look good,” he says. “In order to program that information in a computer, it’s a very difficult task. There’s more interest in using artificial intelligence and deep learning to figure out solutions.”