Exploring the interrelationships of variables in Australian road tunnel incidents using Bayesian networks

Exploring the interrelationships of variables in Australian road tunnel incidents using Bayesian networks

Authors

  • Edwin Hidayat
  • David Lange
  • Jurij Karlovsek

DOI:

https://doi.org/10.58674/phpji.v16i1.411

Keywords:

Incidents, Road tunnels, Bayesian network, Vehicle crash, Fire incidents

Abstract

Incidents in road tunnels can have serious consequences, so it is critical to respond quickly and effectively. To
achieve this, road tunnels are generally equipped with incident detection and fire suppression systems. However,
the system can occasionally generate false alarms, which brings drawbacks for both the user and the tunnel
operator. This paper investigates incident response in Australian road tunnels by analyzing the cause-effect
relationships between incident variables. Bayesian Networks (BN), a machine-learning technique, is applied to
diagnose and predict the interrelationship between variables. Structure learning refers to the process of building
BN structure using two score-based algorithms with three different scoring functions, which are then validated
using four indicators to determine the appropriate network structure that fits the data. Then, parameter learning
is executed to estimate the probability of consequences. The diagnostic and predictive reasoning are then applied
for what-if scenarios to identify variables that have a high influence on the number of injury victims as the worst
consequence variable. The diagnostic analysis reveals some variables that have a significant impact on the number
of injuries. The predictive analysis indicates that the transportation area and incident type (vehicle crash or fire
incident) have a higher probability of injured victims than other location and incident types. The study's findings
can help tunnel operators develop mitigation strategies to reduce the number of people injured in road tunnel
incidents.

Downloads

Published

2023-10-16

Issue

Section

TOPIC B : Road Safety
Loading...