Using near infrared spectroscopy in the food industry

A Technical Insight by Dr Holger Keller, Büchi Labortechnik AG

Dr Holger Keller, Product Manager NIR at Büchi Labortechnik AG, outlines the use of the Cluster ID method for distinct identification of quality aspects in the food industry.

Testing raw materials or finished products using near infrared (NIR) spectroscopy is a common procedure in the chemical and pharmaceutical industry. The main focus is the identification and qualification of, in most cases, pure chemical substances or formulations with well-defined matrices. Pharmaceutical and chemical companies rely on NIR technology because of its speed and the proven specificity to, not only identify and qualify chemicals, but also because of important physical parameters.

Besides the rapidness, the possibility to perform non-invasive analytics is a major advantage of the NIR technique founded on the properties of the NIR radiation, or to be more precise, on the physical nature of the excited vibrational transitions. The wavelength region between 10,000 and 4000 cm -1 is characterised by a weak degree of absorption. Thus, a high ratio of the incident light is reflected from the solid surface after interaction with the sample and can be collected and analysed in the so-called reflectance mode. To analyse the resulting spectral information, various chemometric methods are used with respect to identity control. The Cluster-ID method, a classification method based on principal component analysis (PCA), combines different advantages of other algorithms such as spectral comparison or SIMCA. As a result, the operator of such an analytical tool will receive clear and easy-to-interpret information: ‘Identity OK’ or ‘Identity NOT OK’. This allows its use outside of an analytical laboratory where qualified analytical personnel are not present e.g. directly in the warehouse to perform quality control checks of raw materials or on the production floor to verify the quality of finished products before shipping.

Testing the quality of raw materials or finished products is also an important issue for the food industry. In contrast to the chemical and pharmaceutical industry, the majority of raw materials and finished products in the food industry are highly complex materials from plant material or animals that show seasonal, geographical and species variation. Next to sensory and visual inspections, the quality of a raw material or finished product in the food industry is traditionally defined by a set of chemical or physical parameters, which are quantified using suitable analytical technology.

Companies specialising in industrial bread production have a high demand for ready-to-use bread baking mixtures. The market for bread mixtures at retail stores for use in consumer households is growing continuously as well. The following case study will demonstrate that quality verification of this kind of mixture can be done not only with a single measurement using NIR spectroscopy to acquire quantitative results but also to establish an easy-to-use, distinct identification/qualification method using Cluster ID.

Bread mixtures are much more than just flour, water and yeast. They are very complex formulas containing various components in different ratios such as wheat flour, rye flour, cereal flakes, seeds, whole grain, malt, salt, spices, yeast and additives to enhance the baking result.

Figure 1: Partial Least Squares (PLS) calibration curve for Protein.

The development of these complex formulas requires intensive know-how and time consuming tests in which the relevant quality parameters are evaluated and optimised. The identified critical quality parameters need to be monitored closely before shipment to ensure a satisfying result at an industrial bakery or in the family home. In the following study, the parameters protein, starch, fat and salt were analyzed and calibrated with the use of NIR spectroscopy for a total of eight different bread mixtures. Figure 1 above shows an example of such an NIR calibration.

Table 1 below summarises important parameters derived from the Büchi NIRCal calibration software. They show the calibration range and consistency of calibration data (SEC) and validation results (SEP).

Table 1: SEC: Standard error of calibration, SEP: Standard error of prediction

Different bread mixtures show large differences within the critical parameters but within different batches of the same bread mixture the parameters remain in a narrow band. Therefore it is possible, in a quantitative way, to identify certain composition patterns, which can be analysed with chemometric methods as well.

Figure 2: Score plot of 8 different bread mixtures based on the significant differences of analytical results.

Figure 2 above shows that classification of the eight different bread mixtures based on the data acquired from the reference analysis methods is possible. Because the relevant quantitative information can be calibrated based on the NIR spectral data and correlated with the classification of the different bread mixtures, a classification model based on the spectral data can be derived directly (Figure 3 below).

Figure 3: 3D-Score plot that shows the classification of different bread mixtures into individual clusters.

The results show that there is more than one way to determine the quality of different bread mixtures. The more traditional quantitative approach delivers a number of quantitative results with a single NIR measurement. These results need to be interpreted in order to link them to a quality related pattern. Cluster analysis is an elegant way to merge a traditional quantitative approach with direct examination of spectral information. This delivers a classification of different bread mixtures with a clear message, which leaves no room for interpretation. The results are distinctive: identity conforms or identity does not conform.

Case study: Identification by quantitative PLS analysis

Near Infrared (NIR) Spectroscopy is a well known and established method for the identification and qualification of raw materials in the chemical and pharmaceutical industry.
Testing the quality of raw materials is an important issue for food production, especially in the bakery industry for example. Wheat flour can be produced in several kinds of different qualities. This is obtained by milling together different types of wheat and by using the same mixture in different ratios. Each different quality is produced for a specific purpose. Depending on the quality, the flour will be used for bread production, cakes, biscuits or other bakery products. To ensure the quality of the finished products it is necessary to monitor the characteristics of the flour very closely. This is usually been done by determining a set of chemical and rheological parameters such as moisture, protein, water absorption, stability, dough development time, baking strength of dough or the ratio between dough strength and extensibility, in a quantitative way.

Interpretation of the results is necessary in order to determine whether all quantitative results are in the allowed tolerance range for the expected flour quality or not. The overall quality aspect is therefore characterised as a combination of the different quantitative chemical, alveographic and farinographic results.

This study was developed to evaluate the capability of NIR technology to identify and qualify a certain flour quality with the use of PLS Analysis. For each quantitative parameter a number of samples, ranging from about 150 to 500 samples, were collected. The samples represent six different wheat flour qualities and were collected over a period of approximately 10 months.

The study was carried out by Barilla G. & R. Fratelli Spa company R&D. The results show that with the use of a BUCHI NIRFlex N-500 FT-NIR spectrometer it is possible to develop quantitative PLS calibration models that allow the clear identification of different flour qualities by a non-destructive analysis which avoids labour intensive, traditional, alveographic and farinographic reference measurements.

Now the correct identity of every batch of flour can be immediately verified against an established pattern of several different quantitative parameters immediately after its arrival to the bakery plant.