Investigating the Impacts of Architectural Finishes on Fire Patterns in Support of Developing Data-driven Tools for Pattern Analysis

Interpretation of fire patterns on walls and other surfaces is integral to the investigation of any fire incident. These patterns are, in part, a function of architectural finishes. Nevertheless, prior studies of these phenomena did not account for the potential impacts of such finishes on their conclusions, despite fires usually breaking out in rooms with finished surfaces. As such, efforts to establish the influence of commonly used architectural finishes on fire patterns are long overdue. To address this important knowledge gap, this project will conduct extensive fire tests on drywall with a variety of popular architectural finishes, and use the test results along with data collected from other relevant sources to develop data-driven tools for automatic quantitative fire-pattern analysis. Such analysis, in turn, will address another issue in fire forensics: that fire-pattern analysis is essentially qualitative, relying on the investigators’ experience and skills to correctly deduce the relationship between specific patterns and the overall scene. Several prior studies have warned that using such an approach carries a high risk that evidence will be misinterpreted, casting doubt on the validity of the conclusions reached and on the ability of the process to meet the Daubert Standards. This project’s data-driven tools will utilize advanced techniques in computer vision and machine learning, and will: 1) quantify the degree of fire damage sustained by a dryall surface, either by rating the level of fire-damage, or by predicting the calcintion or heat energy the surface has experienced; 2) classify a  fire pattern by shape, as triangular, columnar, conical, etc.; and 3) identify fire patterns’ causal relationship to fire dynamics. The milestones for this two-year study are, first, collection of data on post-fire scenes, including but not limited to fire-pattern data; second, holistic analysis of the impact of architectural finishes on fire patterns; and third, the development of data-driven tools for quantative and automatic fire-pattern analysis. It is hoped that the outcomes of this work will revolutionize our understanding of how architectural finishes influence fire patterns, while also providing fire investigators and researchers with vital tools for analyzing their data objectively. Both those aspects of this research will improve the accuracy of data interpretation in the field of fire forensics, and ultimately, help make communities safer. CA/NCF