Application of AI based measurement systems for characterization of raw materials in viticulture

Like all agricultural products, grapes vary considerably in terms of their quality. Numerous external factors including climate, soil conditions and time of harvest have a significant impact on the composition of the grapes and, hence, on the wine that is made from them. This diversity is a desirable factor in viticulture as it results in a wide range of wines with all kinds of different characters. In order to ensure the required raw material quality for the desired wine, the grapes need to be characterized on the basis of selected quality parameters. This characterization process should be easy to carry out, it should not damage the grapes and it needs to provide large amounts of information. Infrared spectroscopy fulfills these requirements particularly well. The aim of the SmartGrape joint project is to develop a compact measuring system for the rapid, non-destructive determination of grape quality on the basis of infrared spectroscopy in the mid-infrared (MIR) range.

Using infrared spectroscopy to characterize raw materials

Infrared spectroscopy is a non-destructive optical technique which uses infrared radiation to provide information about the chemical composition of a sample. It is most commonly used in the near-infrared (NIR) range at wavelengths between around 780 and 2500 nm. In this project, however, infrared spectroscopy is to be used in the mid-infrared (MIR) range at wavelengths between 2500 and 50,000 nm to characterize the quality of grapes. The information content in the mid-infrared range is significantly higher than near-infrared, which means that more precise information can be provided.

Using AI in equipment development and in the final MIR measuring system

The large quantities of information generated by infrared spectroscopy and the accompanying chemical analytics result in high-dimensional data sets which require complex evaluation. In the SmartGrape project, artificial intelligence is used to record and evaluate these high-dimensional data sets. The artificial intelligence takes into account non-linear correlations and interaction effects in the data set which can pose problems or take an extremely long time to process when using conventional mathematical/statistical methods. Using artificial intelligence should make it possible to develop a synergistic measuring system where the hardware adapts to the software and vice versa.

Digitalizing viticulture using an MIR measuring system

The AI-based MIR measuring system will provide producers in the field of viticulture or in the agricultural sector in general with a tool for digitalization. The system will allow raw materials to be characterized and digitalized simultaneously in a single, direct process — a possibility not offered by conventional methods. Digitalizing the data will in turn make it possible to apply new methods and measures which can be used within a broad context. For example, the data can be used within the GAIA-X digital ecosystem developed by the German Federal Ministry for Economic Affairs and Energy (BMWi). Exchanging data within a digital ecosystem promotes the sharing of information between the various stakeholders throughout the entire value chain (e.g., agriculturalists, machinery rings, research institutions, etc.). This in turn creates opportunities for process optimization in order to conserve resources and ensure efficiency within the agricultural sector, not least in view of the new challenges presented by climate change. Examples include the possibility of recording changes in quality over many harvest years or the correlation of external influencing factors (e.g., climate, soil quality) and their impact on the composition of the grapes and, ultimately, on the quality of the wine.

Description of the joint project

Fraunhofer is coordinating the SmartGrape joint project in close collaboration with the project partners: IRPC Infrared-Process Control GmbH (Hamburg), LiquoSystems GmbH (Kirchheim am Neckar), QuoData GmbH (Dresden) and Weincampus Neustadt (Neustadt an der Weinstraße). Within the project team, Fraunhofer IVV is involved in carrying out the reference analyses, conducting tests on model substances and identifying marker substances for characterizing the grapes.