Production logistics resource recommendation based on 'look-around' reasoning mechanism in discrete manufacturing

Abstract

Production logistics (PL) involves high levels of complexity and unpredictability, driven by volatile resource demands and a lack of synchronicity in operational workflows. Resilient and efficient resource allocation in PL is essential for optimizing production processes and forms a foundational element in managing resources to achieve zero inventory targets in an uncertain environment. This paper introduces a recommendation-driven approach for real-time resource allocation in PL. Firstly, a resource spatial-temporal knowledge graph (RSTKG) is constructed to capture and analyze the relationships among entities and historical allocation data. Then, we propose a ‘look-around’ reasoning mechanism, which leverages the temporal and spatial attributes of material buffers on shop floors to assess the cost-effectiveness of requested nodes compared to available resources, ultimately generating a resource allocation plan. Finally, to validate our approach, a case study is conducted in an air conditioning manufacturing company, where our method demonstrates a punctuality rate exceeding 90%, outperforming previous allocation methods.