Fire protection strategies for cultural heritage sites: Evacuation optimization and water mist system design via Machine Learning, MCDM and Performance-based Methods
Year: 2025
Type: article
Abstract: Protection of cultural heritage from fire threats requires a delicate balance between historical integrity and incorporating modern provisions on safety. This paper provides an integrated solution to the protection of fire in cultural heritage sites through incorporating evacuation optimization and latest suppression system design. The hybrid solution employs machine learning techniques, MCDM and performance-based fire safety engineering (PB-FSE) principles to enhance occupant safety and asset preservation. Evacuation scenarios are replicated with agent-based models, for which optimization by reinforcement learning identifies the most optimal egress schemes under various fire conditions. A data-driven design methodology for water mist fire extinguishing systems is introduced simultaneously, utilizing supervised learning algorithms to extrapolate optimum nozzle placement, droplet size, and system pressure tailored to building constraints and fire loads. The Hybrid Approach is the highest with an artifact protection rate of 90.03%, which is significantly higher than the baseline approaches (80% for Approach 1 and 85% for Approach 2). This indicates clearly that the system provides a good compromise between fire extinguishing and prevention of damage to priceless cultural artifacts by water. With dynamic flow and nozzle position control, the hybrid system can efficiently protect susceptible regions without over-exposing regions to water. The maximization of artifact protection is paramount in preserving the integrity of cultural heritage buildings. The Hybrid Method is a very efficient way for protecting artifacts during the outbreak of fire, with only minimal damage to replaceable objects.
Authors: Safiye Turgay, Elif Çataklı
Institution: Sakarya University
Language: English
Subfield: Safety, Risk, Reliability and Quality
Field: Engineering
Domain: Physical Sciences
Open Access status: hybrid
OpenAlex ID: W7081930690