The ongoing digitization of industry and agriculture can benefit significantly from optical spectroscopy. In many cases, optical spectroscopy enables the estimation of properties such as substance concentrations and compositions. Spectral data can be acquired and evaluated in real time, and the results can be integrated directly into process and automation units, saving resources and costs. Multivariate data analysis is needed to integrate optical spectrometers as sensors. Therefore, a spectrometer with integrated artificial intelligence (AI) called SmartSpectrometer and its interface is presented. The advantages of the SmartSpectrometer are exemplified by its integration into a harvesting vehicle, where quality is determined by predicting sugar and acid in grapes in the field.

An automatic determination of grape must ingredients during the harvesting process would support cellar logistics and enables an early termination of the harvest if quality parameters are not met. One of the most important quality-determining characteristics of grape must is its sugar and acid content. Among others, the sugars in particular determine the quality of the must and wine. Chiefly in wine cooperatives, in which a third of all German winegrowers are organized, these quality characteristics serve as the basis for payment. They are acquired upon delivery at the cellar of the cooperative or the winery and result in the acceptance or rejection of grapes and must. The whole process is very time-consuming and expensive, and sometimes grapes that do not meet the quality requirements for sweetness, acidity, or healthiness are destroyed or not used at all, which leads to economic loss. Near-infrared spectroscopy is now a widely used technique to detect a wide variety of ingredients in biological samples. In this study, a miniaturized semi-automated prototype apparatus with a near-infrared sensor and a flow cell was used to acquire spectra (1100 nm to 1350 nm) of grape must at defined temperatures. Data of must samples from four different red and white Vitis vinifera (L.) varieties were recorded throughout the whole growing season of 2021 in Rhineland Palatinate, Germany. Each sample consisted of 100 randomly sampled berries from the entire vineyard. The contents of the main sugars (glucose and fructose) and acids (malic acid and tartaric acid) were determined with high-performance liquid chromatography. Chemometric methods, using partial least-square regression and leave-one-out cross-validation, provided good estimates of both sugars (RMSEP = 6.06 g/L, R² = 89.26%), as well as malic acid (RMSEP = 1.22 g/L, R² = 91.10%). The coefficient of determination (R²) was comparable for glucose and fructose with 89.45% compared to 89.08%, respectively. Although tartaric acid was predictable for only two of the four varieties using near-infrared spectroscopy, calibration and validation for malic acid were accurate for all varieties in an equal extent like the sugars. These high prediction accuracies for the main quality determining grape must ingredients using this miniaturized prototype apparatus might enable an installation on a grape harvester in the future.

The selection of training data determines the quality of a chemometric calibration model. In order to cover the entire parameter space of known influencing parameters, an experimental design is usually created. Nevertheless, even with a carefully prepared Design of Experiment (DoE), redundant reference analyses are often performed during the analysis of agricultural products. Because the number of possible reference analyses is usually very limited, the presented active learning approaches are intended to provide a tool for better selection of training samples.

With the rising trend of consumers being offered by start-up companies portable devices and applications for checking quality of purchased products, it appears of paramount importance to assess the reliability of miniaturized sensors embedded in such devices. Here, eight sensors were assessed for food fraud applications in skimmed milk powder. The performance was evaluated with dry- and wet-blended powders mimicking adulterated materials by addition of either ammonium sulfate, semicarbazide, or cornstarch in the range 0.5–10% of profit. The quality of the spectra was assessed for an adequate identification of the outliers prior to a deep assessment of performance for both non-targeted (soft independent modelling of class analogy, SIMCA) and targeted analyses (partial least square regression with orthogonal signal correction, OPLS). Here, we show that the sensors have generally difficulties in detecting adulterants at ca. 5% supplementation, and often fail in achieving adequate specificity and detection capability. This is a concern as they may mislead future users, particularly consumers, if they are intended to be developed for handheld devices available publicly in smartphone-based applications.