Using network theory to explore the complexity of subway construction accident network (SCAN) for promoting safety management

Abstract

Accident case analysis has been widely adopted to promote construction safety. Learning from past accidents is effective to avoid similar dangerous situations or accidents. An accident is often the result of a sequence of previous accidents, or the cause of the next accidents. There is an accident chain or network in practice. Instead of analyzing a single accident, this study uses network theory to explore the complexity of the subway construction accident network (SCAN). Pajek was employed to identify SCAN and analyze corresponding topological characteristics. As a result, an unweighted directed network with 26 vertices and 49 edges was obtained. Five parameters were calculated for better capturing the structure of SCAN. The cumulative degree distribution obeys power-law distribution. This indicates that SCAN is resilience to random attacks. If some high-degree vertices are attacked at the same instant, SCAN is turned to be vulnerable and isolated. The characteristics of big clustering coefficient and short average path length denote that SCAN is a small-world network. This type of network demonstrates faster accident propagation than regular networks. Almost 60% of shortest paths contain collapse of soil, struck-by, explosion and collapse of machine. Effectively controlling these four types of accidents can increase average path length and diameter. As a result, accident propagation efficiency can lower, and chain reaction is dampened in this accident network. Topological parameters analysis is beneficial to understanding the mechanism and capturing the complexity of SCAN. It is helpful to restraint original accidents, and prevent secondary and derivative accidents, which can assist in improving safety management on subway construction sites.