Glanbia Nutritionals make sports nutrition powders containing whey and dairy powders and many other ingredients according to carefully controlled formulas. The production facility in Middlesborough UK makes hundreds of different powders for own brand products and a number of leading brands in the UK. To help the quality controllers, an NIR instrument was recently installed for rapid analysis of protein, fat and moisture. In addition, the company could also see the potential of NIR for discriminating against unusual samples as well as purely quantitative measurement of certain parameters.
Programming the machine
The NIR analyser was measuring amounts of protein, fat and moisture in products according to set tolerances. The new discriminant analysis supplements this by giving a total profile of the sample reflecting the many other ingredients of the whey and milk powder samples such as amino acids and other ingredients that are not currently quantified. Even though the discriminant analysis cannot identify or quantify these, it can give an immediate alarm if something looks different from normal. This is because any sample will give a unique infrared spectra when measured on an
NIR analyser. It is like a fingerprint. If something is different in the product, the spectrum pattern will look different and the analyser can give an alarm and the operators can check further into the recipe.
“We wanted to look at the application of discriminant analysis to identify our own products in house and to check if there were any inconsistencies in the formulation,” says technical director Paul Mousley. “We are making over 500 different products at the site, some similar and some quite different, but we want to make sure that all the recipes are always being followed by people in the dispensary, that there is no variability.”
First though, the quality control team had quite a lot of work to do to collect sample data which was then used to program the NIR instrument. The data tells the machine exactly what a sample should look like, so that it can check if a sample does not match-up to spec.
The system was installed in July 2012. Several product samples were required for each product formulation to show what it should look like. With over 500 products in production this has involved considerable work and the data collection is still ongoing. Discriminant analysis is now in use for the majority of products.
The discriminant analysis check is performed simultaneously with the routine tests and does not involve any extra time or work for operators. If either the quantitative tests or the discriminant test is out, the quality control staff can do further physical analysis of the sample.
Discriminant analysis is sometimes called spectral pattern recognition. The purpose is to classify samples into well defined groups based on a “training set” of similar products.
Peace of mind and reduced costs
Mousley outlines how the discriminant aspect of testing helps to rationalize the use of physical tests, saving money in the process.
“Historically, we would have to send out samples for wet chemistry analysis and taste every batch, but now the NIR is in the system it reduces time and cost involved, particularly in wet chemistry,” he says. “It gives peace of mind to customers too to know that we are doing these checks.”
New dimension to food checks
The quality controllers at Glanbia Nutritionals have the impression that they are unique in the food industry in applying discriminant analysis to so many products. Whether it is chocolate, banana, whey protein or amino acids, all their customers obviously want something different in their nutritional powder and each one has been treated individually.
On a note of caution Mousley explains that discriminant analysis has some limitations in that it is not an actual test of authenticity, but the fact that they get a warning that something is out of profile is of great value for physical testing.
That said, there is obvious scope to expand the system to raw material entering the process. Not all the whey powder used in products comes from within the Glanbia organization. “We want to know that if the certificate on a delivery states 80% protein that it is right and when we blend for 75% protein for a particular product that the customers are getting exactly what they want. What we would like to do now is not necessarily an authenticity test, but a check to see if it is in any way different from the standard before we use it,” he says.
Another interesting aspect of NIR testing is some newer instruments have the broader wavelength and measurement stability required to measure amino acid powders, both at receival and in the production.
Discriminant analysis with NIR also has exciting potential for protecting the supply chain.
“You can make an analogy to the meat industry with the horse meat scandal,” says Mousley. “If we could use it as a tool for authenticity testing, it would be fantastic.”