Character-based hazard warning mechanics: A network of networks approach

Abstract

Hazard warnings derived from hazard associations help in guiding safety inspectors, while detailed descriptions may limit risk perceptions. However, outlining characteristics of hazards may help inspectors to conduct subjective searches for hazards with fewer preset conditions. This study aimed to facilitate safety inspections by determining critical characters of hazards using a character-based network of networks (NoN) with actual construction site data. First, characters were extracted using text analysis and categorized by hierarchical clustering. Then, a character-based NoN was established using network analysis. Critical characters and hazards were generated by considering association strengths and node measures. Finally, the practicability and reliability of associated characters were validated through a case study. Results indicated that (1) “facility/equipment/device,” “setting,” “site/construction site,” and “power distribution/distribution box” were critical characters by evaluation of outdegree, betweenness, closeness, and eigenvector centrality; (2) the hazard warning route from “railing” to “facility/equipment/device” was obtained through associations within and between different layers of the NoN. The case study indicates that the proposed approach based on character associations not only simplifies hazard association routes but also discovers hazards omitted from the hazard network. In practice, the proposed method may assist safety inspectors to focus on critical characters and thereby improve the efficiency of risk identification.