Data and Model Requirements for Statistically Weighted Determination of Fire Origin for Fire Forensics

火灾法医重建在许多方面都是法医学中最复杂、最容易出错的领域之一。美国民事和刑事司法系统依靠专家证人的证词来评估法律案件中的过错和责任。越来越多的专家证人依靠科学模型来解释或促进火灾如何演变的特定假设/理论。火灾调查人员面临的挑战是，火灾本身就会破坏车厢内的物品，并模糊其早期演变过程中产生的特征。在火灾现场提取具有统计意义的特征需要表征火灾环境中典型材料的热化学损伤。这些损伤特征必须以自洽的方式与大规模火灾演变联系起来。该项目的总体目标是开发一种严格的统计方法，将这些测量数据和火灾特征与火灾演变联系起来，用于法医重建。这为今后制定火灾法医标准和最佳做法提供了科学基础。这些标准将成为培训更有科学经验的火灾调查人员的基础。该项目将评估模型偏差和数据不确定性对发布火灾起源预测的影响。在典型火灾现场出现的测量和观察将被提炼成数学术语，可以建模并进行可量化的评估。完成项目目标需要完成几个任务。首先，将在小尺度和大尺度上对凝聚相材料的降解进行实验研究，以表征热损伤。接下来，将开发随机损伤模型来描述材料退化。同时，将开发一个使用计算模型和测量数据的贝叶斯框架，以统计评估火灾起源和进化假设。最后，验证研究将在一个密集的仪器和执行器控制的大型火灾试验室进行。该项目将通过表征火灾热暴露过程中凝聚态材料的特性变化，将特性变化特征封装到损伤模型中，并确定如何将这些损伤模型与气相火灾模型耦合应用于火灾法医重建的限制，来推进火灾科学知识。

Fire forensic reconstruction, in many respects, is one of the most complex and error-plagued areas of forensic science. The United States civil and criminal justice systems rely on expert witness testimony in the evaluation of fault and blame in legal cases. Increasingly, these expert witnesses rely on scientific models to either explain or promote a particular hypothesis/theory of how the fire evolved. The challenge for fire investigators is that by its very nature, a fire damages the contents of the compartment and obscures signatures produced in its early evolution. Extracting statistically meaningful signatures in a fire scene requires characterizing the thermochemical damage to typical materials present in the fire environment. These damage signatures must then be connected in a self-consistent manner to large-scale fire evolution. The overarching goal of this project is to develop a rigorous statistical methodology to connect these measured data and fire signatures to fire evolution for forensic reconstruction. This provides a scientific foundation for future development of fire forensic standards and best practices. Such standards will be the basis for training of more scientifically sophisticated fire investigators.This project will evaluate the effects of model bias and data uncertainty in issuing predictions about the origin of a fire. Measurements and observations present in typical fire scenes will be distilled into mathematical terms that can be modeled and subjected to quantifiable assessments. Several tasks are required to accomplish the project goals. First, degradation of condensed phase materials will be experimentally investigated at small and large scales to characterize thermal damage. Next, stochastic damage models will be developed to describe material degradation. Concurrently, a Bayesian framework using computational models and measured data will be developed to statistically evaluate fire origin and evolution hypotheses. Finally, a validation study will be conducted in a densely-instrumented and actuator-controlled large-scale fire test room. The project will advance knowledge in fire science by characterizing property changes for condensed phase materials during fire thermal exposure, encapsulating property change features into damage models, and identifying the limits on how these damage models coupled to gas-phase fire models should be applied to fire forensic reconstruction.