Fraunhofer Institute for Integrated Circuits IIS, Division Engineering of Adaptive Systems EASThe coronavirus crisis has made it crystal clear to us: extraordinary medical emergencies brought on by the COVID-19 pandemic have led to shortages of much-needed medical supplies – a situation made even worse by worldwide trade restrictions and long delivery times for medical equipment.
To avoid such scenarios in the future, we are supporting a project team led by KEX Knowledge Exchange AG that is developing an early-detection and prediction system for the medical sector. The aim is to identify supply bottlenecks early on so suppliers, distribution points and medical facilities can take countermeasures in good time. This should ensure the sustainable, cost-effective supply of necessary items and, overall, make supply chains resilient to market fluctuations.
Solution approach
In this AI lighthouse project sponsored by the German state of North Rhine-Westphalia, we primarily work on one of the core functionalities: the AI-based early-detection system for consolidating volumes and calculating safety stock levels. The basis for this is a model that forecasts current and future demand for bottleneck products and derives a safe level of inventory from this. The system is designed to detect deviations from optimum inventory levels and provide an early warning whenever the current or future risk of shortages is high. This allows medical facilities to respond to the situation in good time.
We are refining AI algorithms for this functionality and training them with various data such as infection rates, utilization and relevant production metrics. These algorithms will support flexible management of safety stocks in distribution centers and warehouses. In addition, we calculate various forecasts required for the products under consideration, such as demand, delivery quantity, delivery time or price. And it is essential to develop the right strategy for procuring and maintaining safety stocks. Finally, regular stress tests will be conducted to optimize the early-detection system’s resilience.
