Identification of Key Node Sets in Tunneling Boring Machine Cutterhead Supply Chain Network Based on Deep Reinforcement Learning

The supply chain is a network structure prone to disruptions due to its complexity. Specifically, tunnel boring machines (TBMs) are extensive and intricate equipment that undergo design, production, and construction simultaneously, further exacerbating the risks in the TBM cutterhead supply chain (TBMCSC). When a problem arises in an enterprise within the TBMCSC, the risk propagates along the supply chain, impacting other enterprises in the network. Although predicting risks in advance is deemed impossible, identifying the most vulnerable enterprises, which are referred to as key node sets, enables improved risk management. In light of this, this paper proposes a deep reinforcement learning (DRL)-based method for identifying key node sets in a TBMCSC. The approach involves the following steps: first, the entire TBMCSC is modeled using complex network theory (Step 1). Next, risk propagation processes on the network are revealed using the coupled map lattice (CML) method (Step 2). Finally, the DRL algorithm is used to identify key node sets in the TBMCSC, with the aim of maximizing the impact of risk propagation (Step 3). By comparing the extent of risk propagation of the critical node sets identified by the DRL method with the traditional methods when facing the same risks, the superiority of this approach is demonstrated.