Knowledge Graph-Driven Reasoning for Automated Crane Lift Monitoring

Crane lifting operations strongly influence construction productivity, yet current monitoring approaches offer limited ability to interpret lifting activities beyond motion tracking.This study presents a knowledge graph approach that uses a Neo4j property graph with in-database reasoning to interpret crane lifts.A conceptual graph schema is proposed to organize fused sensor and vision data, while Cypher rules operate on the graph to infer eight transient behaviors and segment continuous monitoring into discrete lifts.The method is validated on a field deployment with 21 lifts, and 2,974 labeled samples captured every 3 seconds.The system reaches accuracy 0.911 and F1 0.907 for transient behavior recognition and segments all lifts with median timing errors below 10s per lift.The results show that the proposed rule-based reasoning inside the graph database enables efficient automated monitoring of crane lifts and provides a basis for productivity evaluation and progress tracking.