Four decades of development and refinement are making the electro-magnetic spectrum a tool of both line operators and chemical engineers

A lab technician conducts oil and fat analysis adjacent to production lines. Juxtaposing quality assurance with production is the most common application of NIR spectroscopy; the next step is applying it to process control. Photo courtesy of ABB Bomem Inc.
Simple ideas often require complex calculations to execute, and turning them into easy-to-use devices requires another quantum leap forward. That, in a nutshell, describes the application of near infrared spectroscopy in the food processing industry.

Testing instruments that use near infrared reflectance (NIR) to measure moisture, protein content and many other variables in food trace back to the 1960s (see related story), and the technology's application has spread to many other industries. Farmers use NIR spectroscopes to verify everything from the consistency of planting depth to the moisture content of harvested grain. Process industries continue to find new applications; Eastman Kodak Co. alone has more than 14,000 NIR spectroscopes deployed throughout its Rochester, N.Y., plant. Researchers recently adapted the technology for a commercial unit that assays the pH, glucose and a dozen other dimensions in premature babies' blood, avoiding the need for multiple blood samples that could compromise the most fragile infants.

Food processors, on the other hand, are just beginning to warm to NIR analysis. Once bitten, twice shy, the saying goes, and in years past NIR instrumentation didn't always deliver as advertised. ESE Inc., a Marshfield, Wis., engineering firm that began manufacturing food analysis equipment in recent years, bases its devices on near infrared transmittance (NIT), a variation of NIR. The distinction is usually downplayed, according to company president Tom Walther, "unless we're talking to one of the old guys who was burned by NIR in the past." Even though the current crop of NIR analyzers deliver as promised, he explains, earlier failures left a bad taste in users' mouths.

Current generation units not only are reliable, QA professionals report, they analyze multiple constituents in a fraction of the time of wet chemistry analysis. Standard testing of fat content in cheese using the Babcock method takes 15 minutes and the use of sulfuric acid and a centrifuge, QA manager Diane Gatzke of Pace Dairy Foods Co. in Rochester, Minn., points out. Since Pace began using ESE's Food Quantifier analyzer two years ago, benchtop testing "takes about 15 seconds," she reports, and simultaneous measures of pH, salt and moisture content also are accomplished. Lactose and protein analysis also could be done with the same unit.

Results from moisture tests at Foremost Farms' Decorah, Ia., cheese plant weren't known for 18 hours before near infrared testing was added, reports lab supervisor Jill Lensch, another ESE client; now she has results within 35 seconds, and the accuracy is close to that of wet chemistry. "The use of the machine is simple," she adds, "but not just anybody can do the calibration maintenance and the number crunching that goes along with it."

Simplifying the operation and maintenance of these systems is uppermost in vendors' minds as they try to push the technology out of the lab and onto the production line. But having a champion in the lab is also important, ESE's Walther points out. "We wanted to prove this technology to the lab guys first, because if you can show them that it is feasible, they'll tell the processing guys, 'Yeah, I have one of these units and it's as reliable as hell.'"

NIR in the trenches

While NIR is simply the ratio of incident and transmitted light, complex computations against a known base are required to deduce what that ratio is saying about an organic substance's composition. "One of the limiting factors of NIR is that it is very calibration intensive," notes Ole Rasmussen, market support manager for Foss North America, the Eden Prairie, Minn., division of Denmark's Foss Group that focuses on the food industry. To simplify calibration, Foss developed Artificial Neural Network (ANN), a massive database that provides reference values for similar products. Sampling procedures and calibration maintenance are thereby simplified.

ANN contains calibrations for about 45,000 grain samples, Rasmussen says, allowing users to switch from barley to wheat analysis without re-calibrating the instrument. "Coupled with our new FoodScan device, ANN lets end-users without any analytical chemistry or ANN knowledge use near infrared technology," he says. "It's plug and play, if you will."

Food manufacturers are clamoring for those kinds of robust systems, adds Rick Dempster, director of product and technological development at the American Institute of Baking in Manhattan, Kan. Veteran mixer operators who can eyeball a 2,000-lb. dough to determine moisture content are riding off into the retirement sunset, he notes, so "we're trying to move the art into the science."

An electrical engineer with expertise in machine vision and the electro-magnetic spectrum, Dempster is working with manufacturers to incorporate near infrared devices in their mixer designs. The solution for a bowl-style mixer was simple -- a hole was drilled in the bottom of the bowl and a spectroscope was mounted on the outside -- but a solution is more elusive with horizontal mixers, the most common style used at U.S. commercial bakeries.

"NIR is quite useful because molecular vibration of the salt, sugar and shortening that occurs in the dough is within NIR's range," Dempster reports. "We hope to see a mixer out with an NIR device in about a year."

Improved accuracy in NIR measurement makes the technology applicable throughout the dough makeup and baking processes, he says. Unfortunately, accuracy is not always a given. The first attempt to add online fat analyzers to meat grinders was a flop, according to Doug Brown, food business manager for Quebec, Canada-based ABB Bomem. The problem had more to do with inadequate sample sizes than the technology, he explains. "How representative is a 200-gram sample of a 500 or 2,000 lb. batch of nonhomogeneous product?" he asks. ABB's solution is a sidestream that diverts up to 17 percent of the meat being blended in a batch for analysis before putting it back in the process. A fiber-optic network connects the sample cell to a spectrometer that can monitor up to four blenders, holding down equipment costs for large processors. "We have tied this all together with a statistical process control package" for real-time monitoring and adjustment, says Brown.

Accuracy of NIR measurements is improving thanks to higher speed gauges and more rugged instruments that can better withstand harsh operating environments. The MM710 gauge from NDC Infrared Engineering, for example, processes data five to seven times faster than the unit it replaced, according to Jim Psotka, applications specialist at the Irwindale, Calif., firm. A faster motor spins a filter wheel that dissects light waves every 7.5 milliseconds, allowing the unit to collect more data than before. As a result, four constituents can be monitored, twice the number of earlier generation gauges.

"Historically, online process gauges weren't quick enough" in processing signals, Psotka observes, and changes in humidity levels and lighting could create maintenance headaches. More rugged design and technical refinements have put those limitations to rest, he assures. As a result, food industry use of NIR measurement tools is expected to increase 26 percent per year.

Moisture checking in cookies as they emerge from the oven band is a new application that would not have been possible two years ago. "The system is gated, so you're not getting a false reading from the belt," Psotka says. NIR only measures surface conditions, which is why lab technicians grind samples into a fine powder before testing, but once a plant has sufficient data to relate surface moisture to internal moisture in a specific cookie, operators can be confident in making oven adjustments based on those surface readings.

Snack foods are another category where moisture is an important measure. An emerging measure is fat content. "If you're losing a lot of oil in the product going out the door, you're replenishing oil at a higher rate than you usually have to, and that can be costly," notes Psotka. With NIR, a single online gauge performs both moisture and fat analysis.

Near infrared isn't the only emerging quality-assurance technology available to food processors. Nuclear magnetic resonance, for example, offers some advantages over NIR, though it is limited to fluid products. Oil refineries have successfully applied NMR for years, and recent field tests with Land O' Lakes for butter fat control validated NMR's precision. Invensys has received 3A approval for an NMR device for online analysis in dairies. Regulatory approval is pending.

As with NIR, proponents will have to demonstrate NMR's reliability as an analytical instrument. If it wins converts among lab professionals, it could join near infrared in the process control revolution taking shape.

Sidebar: Three cheers for father of NIR

The race to outer space was shifting into high gear when researchers at USDA's Agricultural Research Service labs in Beltsville, Md., were presented with a more earthly challenge: why were 5 lb. bags of flour in supermarkets chronically underweight, despite having been filled to the proper weight? The problem turned out to be in-package moisture loss, and the solution involved a technique used to explore celestial bodies.

The year was 1963, and Karl H. Norris was a young ARS engineer when the National Bureau of Standards asked the department to investigate the situation. Norris was aware of the application of near infrared reflectance (NIR) to ascertain the presence of water on celestial bodies. He began applying NIR to measure moisture in wheat flour, soy meal and ground corn, but signal interference occurred because of protein content. After building in a correction for protein, interference resulted because of the grains' oil. NIR clearly had applications beyond helping millers adjust their scaling systems, and ARS set out to develop that potential.

"In consulting with the manufacturers of spectrometers, we determined that none was accustomed to making measurements on powders, so the department put out a contract to two companies to make instruments that could," Norris recalls. Other techniques have been developed to assay foods' content, but none can match the simplicity, accuracy, and versatility of using the absorption range of light.

Still active as a consultant to industry, Norris was feted in November by the Society for Applied Spectroscopy in New York, which presented him with the SAS Gold Medal Award for 2001. "Karl Norris is a grand gentleman and a fantastic person who blazed a trail for the rest of us," says society member Bill Fateley, distinguished chemistry professor emeritus at Kansas State University and a leader in applied spectroscopy.