Chemometrics in Food and Beverage

Scientists in the food and beverage industry are faced with many different quality control tasks, such as making sure that flavors meet certain standards, identifying changes in process parameters that may lead to a change in quality, detecting adulteration in any ingredient and identifying the geographical origin of raw materials. Food scientists who work for regulatory agencies, such as the Food and Drug Administration, are interested in detecting economic fraud due to product substitution and adulteration, as well as health risks from possible contamination. Many of these quality control issues have traditionally been assessed by experts, who are able to determine a product's quality by observing its color, texture, taste, aroma, etc. However, it takes years of experience for one to acquire these skills. It would therefore be advantageous if there were a way for food scientists to measure the quality of a product by instrumented means. 15-1214 Food and Beverage Applications Overview © 2014 Page 2 Unfortunately, quality is a difficult parameter to quantify. It is difficult to find direct sensors for quality parameters such as freshness or expected shelf life; therefore we are forced to measure an indirect set of parameters which, individually, may be only weakly correlated to the properties of interest. In analyzing this multivariate data, patterns emerge which are related to product quality and can be recognized by either a human interpreter or a computer. Figure 1. Whisky samples can be classified based on the relative composition of trace constituents (21) For example, a chromatogram or spectral profile can be thought of as a fingerprint, where a pattern emerges from the relative intensities of the chromatographic sequence or spectrum. If these fingerprints are repeatable for every batch packaged for sale, it is possible for an automated quality control system to interpret those patterns in the data. Chemometrics is a statistical approach to the interpretation of patterns in multivariate data. When used to analyze instrument data, chemometrics often results in a faster and more precise assessment of composition of a food product or even physical or sensory properties. For example, composition (fat, fiber, moisture, carbohydrate) of dairy products or grain can be quickly measured using near infrared spectroscopy and chemometrics. Food properties (e.g., taste, smell, astringency) can also be monitored on a continuous basis. In all cases, the data patterns are used to develop a model with the goal of predicting quality parameters for future data. The two general applications of chemometrics technology are:  to predict a property of interest (typically adherence to a performance standard); and  to classify the sample into one of several categories (e.g., good versus bad, Type A versus Type B versus Type C ...) Food and Beverage Applications This overview describes several applications in which chemometrics software has simplified methods development and automated the routine use of robust pattern matching in the food and beverage industry. The examples cited can be duplicated using Pirouette® multivariate modeling software and automated in a routine quality assurance setting using InStepTM. Process Monitoring and Control  Grading of raw materials (1)  Routine on-line quality checks (2, 3)  Minimizing sample preparation (4)  Determining process by which product was made (5) Much of the research and the quality control effort are aimed at assessing a product's consistency or identifying changes in process parameters that may lead to a degradation of quality standards. In most cases, no single measurement is sufficient to categorize samples for QC purposes. By examining a series of parameters simultaneously, an instrumental technique can be utilized that is considerably more precise than the manual spot quality checks that are the tradition. The speed and efficiency of the instrument allows chemometrics technology to be used for batch-tobatch product control (6). Chemometric profiling is useful for detecting changes in a process or in the ingredients; it can also be used to monitor plant-to-plant product variations. For example, near-infrared Chivas Regal