Raise a glass to FT-NIR spectroscopy

Good news for analytical scientists who like their ale: food scientists from Italy and Denmark have shown that Fourier transform NIR (FT-NIR) spectroscopy can be used to monitor beer fermentation.

Such monitoring is essential for ensuring that the resultant beer has the right taste and correct alcohol content, which depends on factors such as the sugar content of the mashed barley, known as wort, and the pH of the fermenting solution. Up to now, however, such monitoring has involved techniques that are either highly subjective, such as tasting, or require samples to be taken away for subsequent analysis. What is really needed is a monitoring technique that is both objective and able to be conducted in-line during the brewing process.

FT-NIR offers one option, but is hampered by the fact that wort is a highly complex mixture consisting of various different sugars and proteins, all of which absorb similar NIR wavelengths. To make matters worse, wort also contains lots of water, which strongly absorbs at NIR wavelengths and so tends to mask weaker absorption by other compounds. All of which makes monitoring pH levels and sugar content using FT-NIR spectroscopy fairly tricky.

Nevertheless, the team led by José Manuel Amigo at the University of Copenhagen decided to give it a go. First off, they used a pH-meter, a refractometer and a UV-visible spectrometer to monitor changes in pH levels, sugar content and wort biomass respectively during the brewing process. They monitored this process for two different strains of yeast, and at three different temperatures, and found that the three factors changed in different ways for each set of conditions, influencing the taste and alcohol content of the resultant beer. This was especially the case for the two yeast strains, with one producing a British ale and one producing a Belgian ale.

As reported in Food Chemistry, the team next analysed the brewing processes for each set of conditions with FT-NIR and related the spectral data to the measured pH levels, sugar contact and wort biomass using partial least squares (PLS) regression. This produced models that weren’t particularly accurate at determining these three factors from NIR data, most likely because of the chemical complexity of the fermenting solution.

To deal with this complexity, Amigo and his team first treated the NIR data with a technique known as standard normal variate and then produced a model using a non-linear variant of PLS known as locally weighted regression-PLS. The resultant models turned out to be much more accurate at determining changes in the pH level, sugar content and wort biomass during the brewing process, proving that FT-NIR spectroscopy can certainly handle its beer.

Blog tags: