Prediction of Bridge Fires Characteristics Using Machine Learning

Abstract

Bridges are an essential component of transportation systems for traffic passage across natural or man-made obstacles. Modern urban trends and growing travel patterns have resulted in a rise in the number of bridges on motorways to avoid road conflict. A major fire might cause irreversible structural deterioration or perhaps the bridge's failure. This research studies the critical factors of bridge fire incidents using a comprehensive database containing 171 bridges. Using an Artificial Neural Network (ANN) model, the vulnerability of bridges in fire is estimated, and the extent of damage is determined based on several key factors including bridge proximity to urban, suburban, or rural areas, structural system, construction materials, annual average daily traffic (AADT), ignition source, combustible types, position of the fires, and the fire-caused damage level. The influence of each parameter on the bridge fires damage levels is investigated. Suggesting measures to reduce the risk of damage, damage levels of fire incidents in different bridges for different scenarios is predicted and a comprehensive and accurate model for definition of vulnerability of bridges under fires is proposed. The outcome of this research will help predict the risk of fire in bridges with various characteristics and the level of damage.