A Three-Part Bayesian Network for Modeling Dwelling Fires and Their Impact upon People and Property

Area of Science:

• Fire Safety Engineering
• Risk Assessment
• Probabilistic Modeling

Background:

• Dwelling fires in the UK disproportionately cause fire-related fatalities.
• Assessing dwelling fire risk is complex due to numerous interacting variables, leading to ambiguity in safety planning.
• Existing methods often struggle to comprehensively address the multifaceted nature of fire development.

Purpose of the Study:

• To propose a novel three-part Bayesian network model for analyzing dwelling fires.
• To enhance confidence and reduce ambiguity in dwelling fire safety assessment and decision-making.
• To provide a robust framework for studying fire dynamics from ignition to extinguishment.

Main Methods:

• Development of a three-part Bayesian network model.
• Integration of both hard (e.g., physical data) and soft (e.g., expert opinion) data inputs.
• Utilizing posterior probabilities to quantify outcomes and uncertainty.

Main Results:

• The model successfully simulates dwelling fire progression from ignition to extinguishment.
• Posterior probabilities for key fire outcomes were delivered, enhancing risk assessment.
• Case studies validated the model's functionality and applicability.

Conclusions:

• The proposed Bayesian network model offers a significant advancement in dwelling fire risk assessment.
• The model improves confidence in fire safety planning and decision-making processes.
• It serves as a valuable tool for both proactive safety planning and post-incident investigation.