A Knowledge Graph-Based Case Similarity Calculation Method for Ship Sub-Assembly Welding Process

As the shipbuilding industry advances toward intelligent manufacturing, the design of ship sub-assembly welding processes faces significant challenges due to the fragmented and heterogeneously formatted nature of welding process knowledge, alongside a substantial dependence on engineers’ empirical expertise in process planning. To address these issues, a method for computing the similarity between ship welding cases was proposed by leveraging knowledge graphs (KGs) and a graph matching network (GMN). First, a KG of ship welding processes is constructed to enable standardized representation and unified management of process knowledge. Then, a multi-level GMN incorporating local normalization and segmented encoding is employed. The principal function of this network is to assess the similarity between welding cases, thereby facilitating the intelligent retrieval and reuse of process cases. Experimental results demonstrate that the proposed method can effectively identify analogous welding process cases, providing valuable technical support for the intelligent transformation of shipbuilding processes.