Project Goal: Determining Shelf Life
More Realistically Throughout the food value chain—from production through logistics and retail to the end consumer—environmental conditions such as temperature and light change. These factors influence food quality but have so far been taken into account only to a limited extent in shelf-life assessments.
The research team from ILT.NRW, inIT, and industry partners therefore investigated chemical, physical, and sensory changes in beverages under various storage conditions. In parallel, machine learning methods were developed to analyze this data and make it usable for quality predictions.
New Approach: Multi-sensor Technology and AI
Many existing methods for determining shelf life are based on external factors such as temperature or humidity. However, these often provide only indirect indications of a food product’s actual condition.
The project approach therefore combines:
intrinsic parameters such as pH or gas composition
optical parameters such as changes in color and turbidity
sensory evaluations
AI-based information fusion and pattern recognition
This combination allows for more accurate detection of quality changes and the creation of more realistic shelf-life predictions.
Project
Approach The project examined various lemonades, their raw materials, and orange juice. In aging and storage experiments, the team analyzed changes in the products using chemical, physical, and spectroscopic methods. The results were evaluated kinetically and multivariately.
At the same time, a Monitoring Device (MD) was developed—a sensor system for continuous data collection. It includes, among other things:
A camera for analyzing color and turbidity data
Temperature sensors
Integrated lighting
Control unit for data acquisition and processing
The system was tested in laboratory environments and prepared for use in real-world logistics processes.
Key findings
The investigations show that some standard parameters, such as pH, density, or turbidity, exhibit only minor changes over time and temperature. Other factors react much more sensitively to storage conditions, including:
Sugar content and degradation products
Color changes (e.g., anthocyanins, browning index)
Viscosity
L-ascorbic acid content
Sensory properties
Aging kinetics were established for orange and raspberry raw materials as well as orange juice. Based on sensory tests, an acceptance limit for orange juice was also determined. From this, a dynamic cut-off point (COP) was derived—a potential approach for a dynamic best-by date.
Demonstrator for practical application
Another project success is the construction of a monitoring device in the form of a beverage crate. The system can potentially be integrated into the logistics chain of beverage distribution to continuously collect data on product changes.
The project has already succeeded in modeling changes in color and turbidity in various model solutions and lemonades.
Outlook
The projects’ results demonstrate that the combination of multi-sensor technology, data fusion, and AI opens up new possibilities for a realistic assessment of food quality. In the long term, dynamic shelf-life data could help make supply chains more efficient and significantly reduce food waste.
