The intelligent upgrading of manufacturing has placed higher demands on programmable logic controller (PLC) technical personnel. This paper proposes an industrial case-driven teaching model for PLC technology courses. Based on knowledge graphs and Petri nets, an aggregation model for process knowledge discovery is constructed to achieve analogy transfer for multi-domain engineering problems. A hierarchical embedding collaborative recommendation model (HECR) is designed, utilizing multi-scale graph convolutions, residual connections, and LightGCN to optimize the matching of engineering problems and teaching resources. The model’s loss function stabilizes around 0.0512 as the number of recommendation method learning iterations increases. When the resource hierarchy is set to 10 levels, the model’s two metric values reach their highest values of 0.629 and 0.686. Students’ satisfaction with the reformed teaching methods exceeded 0.830 across all four emotional evaluation themes.
