Hybrid combinatorial remanufacturing strategy for medical equipment in the pandemic

The COVID-19 pandemic hit the medical supply chain, creating a serious shortage of medical equipment. To meet the urgent demand, one realistic way is to collect abandoned medical equipment and then remanufacture, where the disassembled modules are shared with all stock-keeping units (SKUs) to improve utilization. However, in an emergency, the equipment should be processed sequentially and immediately, which means the decision is short-sighted with limited information. We propose a hybrid combinatorial remanufacturing (HCR) strategy and develop two reinforcement learning frameworks based on Q-learning and double deep Q network to find the optimal recovery option. In the frameworks, we transform HCR problem into a maze exploration game and propose a rule of descending epsilon-greedy selection on reweighted valid actions (DeSoRVA) and Espertate knowledge dictionary to combine the cost-minimizing objective with human judgment and the global state of the problem. A real-time environment is further implemented where the quality status of the in-transit equipment is unknown. Numerical studies show that our algorithms can learn to save cost, and the larger scale of the problem is, the more cost-down can be achieved. Moreover, the sophisticated knowledge refined by Espertate is effective and robust, which can handle remanufacturing problems at different scales corresponding to the volatility of the pandemic.