Fire models are routinely used to evaluate life safety aspects of building design projects and are being used more often in fire and arson investigations as well as reconstructions of firefighter line-of-duty deaths and injuries. A fire within a compartment effectively leaves behind a record of fire activity and history (i.e., fire signatures). Fire and arson investigators can utilize these fire signatures in the determination of cause and origin during fire reconstruction exercises. Researchers conducting fire experiments can utilize this record of fire activity to better understand the underlying physics. In all of these applications, the fire heat release rate (HRR), location of a fire, and smoke production are important parameters that govern the evolution of thermal conditions within a fire compartment. These input parameters can be a large source of uncertainty in fire models, especially in scenarios in which experimental data or detailed information on fire behavior are not available. To better understand fire behavior indicators related to soot, the deposition of soot onto surfaces was considered. Improvements to a soot deposition submodel were implemented in a computational fluid dynamics (CFD) fire model. To better understand fire behavior indicators related to fire size, an inverse HRR methodology was developed that calculates a transient HRR in a compartment based on measured temperatures resulting from a fire source. To address issues related to the uncertainty of input parameters, an inversion framework was developed that has applications towards fire scene reconstructions. Rather than using point estimates of input parameters, a statistical inversion framework based on the Bayesian inference approach was used to determine probability distributions of input parameters. These probability distributions contain uncertainty information about the input parameters and can be propagated through fire models to obtain uncertainty information about predicted quantities of interest. The Bayesian inference approach was applied to various fire problems and coupled with zone and CFD fire models to extend the physical capability and accuracy of the inversion framework. Example applications include the estimation of both steady-state and transient fire sizes in a compartment, material properties related to pyrolysis, and the location of a fire in a compartment. / text

No description available.

The interaction between the heat flux and fluid flow of an engine compartment fire and the windshield of a post-collision passenger vehicle has been studied using analytical methods. A computational fluid dynamics model of this scenario has been developed with TASCflow using a turbulent, reacting, multi-component fluid flow in a multi-grid domain with conjugate heat transfer objects. A group of computer programs have also been created to automate the grid generation and model construction processes. Calculation tools have also been developed using aspects of fire dynamics for the purpose of making comparisons to the results of CFD modeling as well as experimental measurements. A framework has been established for the modeling and validation of the windshield problem using the tools developed in this study.

La prévision des incendies de forêt reste un défi puisque vitesse et direction de propagation dépendent des interactions multi-échelles entre la végétation, la topographie du terrain et les conditions météorologiques. Un modèle à l’échelle régionale peut donc difficilement prendre en compte le détail des processus physiques mis en jeu. Toute modélisation est entachée de nombreuses incertitudes (modélisation incomplète, méconnaissance du terrain, de la végétation et des interactions flamme/atmosphère, etc.) qu’il est nécessaire de quantifier et de corriger. Ces travaux de thèse proposent ainsi une modélisation régionale des incendies qui a des meilleures capacités de simulation et prévision, basée sur une évaluation des modèles et l’assimilation de données. L’évaluation de modèles a consisté à développer des simulations multi-physiques à l’échelle de la flamme, incluant la résolution des équations de Navier-Stokes réactives, l’évaluation du transfert radiatif vers la végétation, la construction d’un modèle de pyrolyse de la végétation ainsi que la modélisation de l’interface flamme/végétation afin de mieux comprendre la dynamique des incendies. La seconde approche proposée a consisté à mettre en place un prototype d’assimilation de données pour le suivi de la propagation du front de feu. L’idée est de rectifier la trajectoire simulée du front au fur et à mesure que de nouvelles observations sont mises à disposition, la différence entre les positions observées et simulées du front étant traduite en une correction des paramètres de vitesse de propagation ou directement de la position du front via l’algorithme du filtre de Kalman d’ensemble. Ces approches, tenant compte des incertitudes à la fois sur la modélisation des incendies et sur les observations disponibles, permettent ainsi d’améliorer la prévision de la dynamique des feux ainsi que des émissions atmosphériques, ce qui constitue un enjeu de taille pour la protection civile et environnementale. / Because wildfires feature complex multi-physics occurring at multiple scales, our ability to accurately simulate their behavior at large regional scales remains limited. The mathematical models proposed to simulate wildfire spread are currently limited because of their inability to cover the entire range of relevant scales, because also of knowledge gaps and/or inaccuracies in the description of the physics as well as inaccuracies in the description of the controlling input parameters (i.e., the vegetation, topographical and meteorological properties). For this purpose, the uncertainty in regional-scale wildfire spread modeling must be quantified and reduced. In this context, the goals of this thesis are two-fold. First, multi-physics detailed simulations of fire propagation, solving for the flame structure using Navier-Stokes equations for multi-species reacting flow and including radiation heat transfer, biomass pyrolysis as well as a flame/vegetation interface, were performed at the flame scale. These simulations were compared to measurements to provide a comprehensive understanding of the mechanisms underlying fire propagation. Second, the use of a data-driven simulator that sequentially integrates remote sensing measurements and relies on an empirical spread model was explored for regional-scale fire front tracking. The idea underlying this data assimilation strategy was to translate the differences in the observed and simulated fire front locations into a correction of the input parameters of the empirical model or directly of the fire front location via the ensemble Kalman filter algorithm. Since these two approaches account for uncertainties in fire spread modeling and measurements, they improve our ability to forecast wildfire dynamics and plume emissions. These challenges have been identified as a valuable research objective with direct applications in fire emergency response for civil defense and environmental protection.

The ignition and subsequent burning of polyurethane foam based mattresses poses a significant danger to life and safety in North American homes. The development of fire models which can predict the full scale fire behavior of these mattresses using bench scale data would assist manufacturers and regulators to manage this danger in a cost effective manner. This thesis builds on previous work by the University of Saskatchewan and University of Waterloo fire research groups and focuses on the evaluation of one such scaling model, which was originally developed during the Combustion Behavior of Upholstered Furniture (CBUF) project. The evaluation of the CBUF model conducted in this thesis isolates the heat release rate (HRR) density sub-model and explores the effects of 1) cone calorimeter incident heat flux setting, 2) specimen thickness and 3) ignition location on the predictive capability of the CBUF model. To provide input for the CBUF model cone and furniture calorimeter tests were conducted. Cone calorimeter tests were conducted on foam specimen thicknesses of 2.5, 5.0, 7.5 and 10.0 cm at incident heat flux settings of 25, 35, 50 and 75 kW/m2. Furniture calorimeter tests were conducted on foam specimen thicknesses of 2.5, 5.0, 7.5 and 10.0 cm in both edge and center ignition configuration. Flame area spread rates were measured from infrared video of the furniture calorimeter tests using an automated algorithm. It was found that HRR curves predicted by the CBUF model showed good agreement with experimental results. Experimental results from tests of thinner foams were predicted with greater success than results from thicker foams, and results from edge ignition tests were predicted with greater success than results of center ignition tests. The results of this study indicated that specimen thickness and ignition location need to be considered when selecting an appropriate incident heat flux setting for producing input data for the CBUF model.

Since the mid-1990s, formal verification has become increasingly important because it can provide guarantees that a software system is free of bugs and working correctly based on a provided model. Verification of biological and medical systems is a promising application of formal verification. Human neural networks have recently been emulated and studied as a biological system. Some recent research has been done on modelling some crucial neuronal circuits and using model checking techniques to verify their temporal properties. In large case studies, model checkers often cannot prove the given property at the desired level of generality. In this thesis, we provide a model using the Coq proof assistant and prove some properties concerning the dynamic behavior of some basic neuronal structures. Understanding the behavior of these modules is crucial because they constitute the elementary building blocks of bigger neuronal circuits. By using a proof assistant, we guarantee that the properties are true in the general case, that is, true for any input values, any length of input, and any amount of time. In this thesis, we define a model of human neural networks. We verify some properties of this model starting with properties of neurons. Neurons are the smallest unit in a human neuronal network. In the next step, we prove properties about functional structures of human neural networks which are called archetypes. Archetypes consist of two or more neurons connected in a suitable way. They are known for displaying some particular classes of behaviours, and their compositions govern several important functions such as walking, breathing, etc. The next step is verifying properties about structures that couple different archetypes to perform more complicated actions. We prove a property about one of these kinds of compositions. With such a model, there is the potential to detect inactive regions of the human brain and to treat mental disorders. Furthermore, our approach can be generalized to the verification of other kinds of networks, such as regulatory, metabolic, or environmental networks.

"Alternate models for extended ceiling convection heat transfer and ceiling vent mass flow for use in the Harvard Computer fire Code are developed. These models differ from current subroutines in that they explicitly consider the ceiling jet resulting from the fire plume of a burning object. The Harvard Computer fire Code (CFC) was used to compare the alternate models against the models currently used in CFC at Worcester Polytechnic Institute and with other available data. The results indicate that convection heat transfer to the ceiling of the enclosure containing the fire may have been previously underestimated at times early in the fire. Also, the results of the ceiling vent model provide new insight into ceiling vent phenomena and how ceiling vents can be modeled given sufficient experimental data. this effort serves as a qualitative verification of the models as implemented; complete quantitative verification requires further experimentation. Recommendations are also included so that these alternate models may be enhanced further. "

<p dir="ltr">Fires cause significant casualties and property damage. As critical component of indoor and building fires, fires along a surface (vertical or horizontal) contribute significantly to fire spreading and resulted damage. Accurately predicting the interactions between a wall surface and fire is crucial to minimizing losses. Computational methods, such as large-eddy simulations (LES), can result in errors in fire modeling along a surface due to various model and numerical errors among which the error in the convective wall heat transfer models is an important source. The convective heat transfer model error grows when the grid resolution near a thermal boundary layer along a wall surface decreases. Traditional wall-function based heat transfer models, mostly developed for forced convection heat transfer problems, tend to fail in the buoyancy-driven fire wall heat transfer. It is imperative to develop accurate and efficient convective wall heat transfer models for fire modeling.</p><p dir="ltr">In this study, machine learning is employed as an alternative to traditional physics-based modeling approach for wall heat transfer in fire modeling. A significant advantage of machine learning over physics-based modeling is that machine learning does not require thorough knowledge of fire wall heat transfer which is generally hard to acquire due to the complexity of the problem. A machine-learning assisted convective wall heat transfer model, aiming to enhance wall fire predictions, is developed in this work. The objective is to improve predictions of convective heat flux to a wall in under-resolved LES of wall fires. An amplification factor ($\beta$) is introduced to compensate the under-prediction of temperature gradients normal to a wall surface in coarse grid simulations. Machine learning is then employed to assist the construction of models for $\beta$ with the training data obtained directly from fine-resolution LES. Extensive studies are conducted to identify suitable machine learning architecture, input features, training data generation strategies, training procedure, and testing and validation approaches.</p><p dir="ltr">A vertical wall fire test case is considered first to develop a baseline machine learning model. The focus is on identifying suitable input features and training strategies for machine learning of convective wall fire heat transfer. A four-parameter (input) machine learning model for $\beta$ is constructed. Both \textit{a priori} and \textit{a posteriori} testing are developed in the vertical wall fire case to provide preliminary model performance assessment. The fully tested model is also examined in an intermediate-scale parallel-wall fire spreading case that was not seen in the model training to assess the applicability of the developed machine learning model. In general, excellent model performance is observed in the vertical wall fire case.</p><p dir="ltr">The established machine learning approach for the vertical wall case is then extended to horizontal surfaces like floor fires and ceiling fires to expand the training scope of the machine learning model. The unique challenges in these new fire scenarios are investigated separately to identify the need of additional input features and training strategies. It is found that a fifth input parameter, in addition to the four parameters identified in the vertical wall, is generally needed in order to correctly identify different fire scenarios. Data augmentation techniques are also found to be a useful technique to handle data sparsity during model training. Different machine learning architectures like random forest and deep neural network are also compared.</p><p dir="ltr">The above studies are finally integrated into a unified machine learning model suitable for both vertical and horizontal surfaces. Extensive testing shows that the unified model reproduces the model performance of the separately trained models. The work is significant in demonstrating the feasibility of using machine learning approaches to enhance fire simulations. The developed machine learning modeling techniques improve predictions in various fire scenarios by using relatively coarse grid to maintain low computational cost, a critical consideration when simulation approaches are employed in real fire simulations.</p>

Ces travaux de thèse ont porté sur le développement d’un modèle de propagation d’un incendie de forêt et son intégration dans une chaîne opérationnelle d'aide à la lutte tactique. C'est un modèle dont la propagation s'effectue sur un réseau de sites combustibles qui prend en compte les mécanismes principaux de transfert de chaleur radiatifs et convectifs des sites en feu vers les sites sains et l'environnement. Ce modèle tient également compte du relief et des conditions locales de vent et végétation. La simulation « temps réel » a nécessité deux développements distincts. Le premier a consisté à combiner la méthode de Monte Carlo à un algorithme génétique pour créer une base de données des facteurs de vue radiatifs de la flamme sur la végétation environnante, pour une large gamme de propriétés de flammes et de conditions environnementales. Le second repose sur une méthode de suivi du front de feu afin de limiter les données manipulées aux seules données utiles au calcul de sa propagation. La phase de validation a porté sur l’analyse comparative des contours de feux calculés par le modèle avec ceux mesurés lors de deux brûlages dirigés, dont un réalisé en Thaïlande dans le cadre de cette thèse, et ceux mesurés lors du feu de Favone de 2009 en Corse et d'un feu de grande ampleur aux États-Unis. Les temps de calcul sont très inférieurs au temps réel. Le modèle a été ensuite étendu pour permettre une évaluation du risque incendie à l’interface forêt-habitat. Dans le cadre du projet TechForFire, porté par la société NOVELTIS, il a été enfin couplé aux différents modules de la chaîne opérationnelle. La chaîne complète a été validée sur le feu historique de Velaux de 2015. / This thesis work is focused on the development of a wildfire spread model and its integration into a decision-making tool for planning firefighting operations. The fire spread model is based on a network model to represent vegetation distribution on land and considers the main heat transfer mechanisms from burning to unburnt vegetation items (i.e. radiation from the flaming zone and embers, surface convection and wind-driven convection through the fuel bed, and radiative cooling from the heated fuel element to the environment). The effects of local conditions of wind, topography, and vegetation are included. To address the challenge of real-time fire spread simulations, the model is also extended in two ways. First, the Monte Carlo method is used in conjunction with a genetic algorithm to create a database of radiation view factors from the flame to the fuel surface for a wide variety of flame properties and environment conditions. Second, the front-tracking method is introduced in order to reduce the amount of data to store and handle during propagation. The fire spread model is validated against data from different fire scenarios, showing it is capable of capturing the trends observed in experiments in terms of rate of spread, and area and shape of the burn, with reduced computational resources. It is then extended to evaluate fire risk at the wildland fire interface. In the frame of the TechForFire project coordinated by the NOVELTIS company, the new version of the fire spread model is coupled with the other modules of the operational chain. Finally, data from the fire of Velaux in 2015 are used to evaluate the TechForFire solution.

"Review of the literature shows that the reported correlation between predictions and experimental data of flame spread vary greatly. The discrepancies displayed by the models are generally attributed to inaccurate input parameters, user effects, and inadequacy of the model. In most experiments, the metric to which the model is deemed accurate is based on the prediction of the heat release rate, but flame spread is a highly complex phenomenon that should not be simplified as such. Moreover, fire growth models are usually made up of distinctive groups of calculation on separate physical phenomena to predict processes that drive fire growth. Inaccuracies of any of these “sub-models” will impact the overall flame spread prediction, hence identifying the sources of error and sensitivity of the subroutines may aid in the development of more accurate models. Combating this issue required that the phenomenon of flame spread be decomposed into four components to be studied separately: turbulent fluid dynamics, flame temperature, flame heat transfer, and condensed phase pyrolysis. Under this framework, aspects of a CFD model may be validated individually and cohesively. However, a lack of comprehensive datasets in the literature hampered this process. Hence, three progressively more complex sets of experiments, from free plume fires to fires against an inert wall to combustible wall fires, were conducted in order to obtain a variety of measurements related to the four inter-related components of flame spread. Multiple permutations of the tests using different source fuels, burner size, and source fire heat release rate allowed a large amount of comparable data to be collected for validation of different fire configurations. FDS simulations using mostly default parameters were executed and compared against the experimental data, but found to be inaccurate. Parametric study of the FDS software shows that there are little definitive trends in the correlation between changes in the predicted quantities and the modeling parameters. This highlights the intricate relationships shared between the subroutines utilized by FDS for calculations related to the four components of flame spread. This reveals a need to examine the underlying calculation methods and source code utilized in FDS."