Building a Digital Twin and an AI Optimization Algorithm for APotheek Voorzorg

Apotheek Voorzorg specializes in automated customized packaging of medicine for patients. Based on received prescriptions, they package medicines per intake moment in a Medication Roll that is personalized for each customer. Deliveries are made to care centers, correctional facilities, nursing homes, and home care patients across the Netherlands within several days after an order is made. Their goal is to help people that use different prescriptions to independently, easily, and safely take the right amount of medicine at the right time.

When creating the Medication Roll, Apotheek Voorzorg selects tablets, dragées, and capsules for each individual patient according to the time of ingestion and then packages the medicine in a special plastic bag. One sachet contains all the pills that should be taken at the same time. For the most part, sachets are filled automatically using their industrial machines working around the clock. The machines can store a maximum of 400 pills and a prescription can be processed only if all the required pills for that sachet are in the same machine. This allows them to create Medication Rolls in automatic mode when all the pills are available in the machine. When not all the products are available in the same machine the productivity in the machine is significantly reduced.

With an assortment of more than 1500 available pills and thousands of possible prescriptions, what is the optimal combination of pill assortment that should go into each machine to ensure maximum efficiency while keeping the highest quality without any errors? This is what Anchormen set out to do.

In collaboration with the client, we looked at their entire production process to see whether smart data analysis could help them. We quickly pinpointed that the machines were the biggest bottleneck and the way to improve efficiency is to make sure all the necessary pills were assorted in such a way that prescriptions could be filled automatically, without human intervention. We also couldn’t stop their whole process to test different pill-to-machine combinations because that would prevent people from getting their medicine on time.

So, instead, we built a Digital Twin of their whole production chain which could serve as a real-time digital counterpart where we could simulate different combinations of pill arrangement. We also created an optimization algorithm that analyzed all of their historical data and based on that gave recommendations of which pills should go in which machine to handle as many possible prescriptions automatically.

Building the Digital Twin allowed us to run hundreds of different simulations and assess the impact of each of them without having to disturb Apotheek Voorzorg’s production line. We were even able to estimate exactly how much more efficient they would become with each different combination. Within the first year, we were able to achieve 14% increase in automatically handled prescriptions. In practice, this all added up to a cost saving of €0.5 million per year for the client.

Furthermore, we also built a Planner Module that would automatically allocate the right prescription to the right machine at the right time in order to increase efficiency of the production process even more. Thus, helping Apotheek Voorzorg handle non-urgent orders only after pill assortments were optimized instead of accepting lower efficiency. In the end, we managed to help the client overcome several bottlenecks, allowing them to increase efficiency and be able to help more people receive their medicine at the right time.