Flame and smoke image processing and analysis could improve performance to detect smoke or fire and identify many complicated fire hazards, eventually to help firefighters to fight fires safely. Deep Learning applied to image processing has been prevailing in recent years among image-related research fields. Fire safety researchers also brought it into their studies due to its leading performance in image-related tasks and statistical analysis. From the perspective of input data type, traditional fire research is based on simple mathematical regressions or empirical correlations relying on sensor data, such as temperature. However, data from advanced vision devices or sensors can be analyzed by applying deep learning beyond auxiliary methods in data processing and analysis. Deep Learning has a bigger capacity in non-linear problems, especially in high-dimensional spaces, such as flame and smoke image processing. We propose a video-based real-time smoke and flame analysis system with deep learning networks and fire safety knowledge. It takes videos of fire as input and produces analysis and prediction for flashover of fire. Our system consists of four modules. The Color2IR Conversion module is made by deep neural networks to convert RGB video frames into InfraRed (IR) frames, which could provide important thermal information of fire. Thermal information is critically important for fire hazard detection. For example, 600 °C marks the start of a flashover. As RGB cameras cannot capture thermal information, we propose an image conversion module from RGB to IR images. The core of this conversion is a new network that we innovatively proposed: Dual-Attention Generative Adversarial Network (DAGAN), and it is trained using a pair of RGB and IR images. Next, Video Semantic Segmentation Module helps extract flame and smoke areas from the scene in the RGB video frames. We innovated to use synthetic RGB video data generated and captured from 3D modeling software for data augmentation. After that, a Video Prediction Module takes the RGB video frames and IR frames as input and produces predictions of the subsequent frames of their scenes. Finally, a Fire Knowledge Analysis Module predicts if flashover is coming or not, based on fire knowledge criteria such as thermal information extracted from IR images, temperature increase rate, the flashover occurrence temperature, and increase rate of lowest temperature. For our contributions and innovations, we introduce a novel network, DAGAN, by applying foreground and background attention mechanisms in the image conversion module to help reduce the hardware device requirement for flashover prediction. Besides, we also make use of combination of thermal information from IR images and segmentation information from RGB images in our system for flame and smoke analysis. We also apply a hybrid design of deep neural networks and a knowledge-based system to achieve high accuracy. Moreover, data augmentation is also applied on the Video Semantic Segmentation Module by introducing synthetic video data for training. The test results of flashover prediction show that our system has leading places quantitative and qualitative in terms of various metrics compared with other existing approaches. It can give a flashover prediction as early as 51 seconds with 94.5% accuracy before it happens.

[ES] El objetivo de esta tesis es la investigación y análisis de la estructura interna de los chorros diésel reactivos y el efecto de las condiciones de contorno en los parámetros asociados a la combustión. Este objetivo se consigue por medio de la simulación numérica del chorro con modelos de turbulencia RANS y LES usando un modelo de combustión avanzado basado en el concepto flamelet. Para este estudio, se aplica una aproximación simplificada de las flamelets de difusión, conocidas en la literatura como Flamelets de Difusión Aproximadas (ADF en inglés), como fundamento del modelo de combustión. En una primera etapa, el modelo se valida con combustibles de diferente complejidad química en regímenes estacionarios y transitorios para el conjunto de posibles velocidades de deformación. Una vez se confirma su idoneidad para condiciones encontradas en chorros diésel, se aplica a la simulación del chorro A del Engine Combustion Network (ECN), representativo de chorros diésel. Para proporcionar un cuadro completo de los fenómenos subyacentes, la combustión se analiza inicialmente para condiciones homogéneas y llamas laminares para las distintas condiciones de contorno de este experimento. Después este análisis se complementa con la simulación de diferentes mecanismos químicos para determinar cómo las características del encendido predichas por el esquema de oxidación afectan a la propagación de llama. Los resultados obtenidos en esta etapa se enlazan con el análisis del chorro turbulento en el contexto de simulaciones RANS y LES para describir cómo el fenómeno de la combustión se modifica con los diferentes niveles de complejidad física. La estructura del chorro turbulento se describe profundamente para las distintas condiciones de contorno y mecanismos químicos en términos de mezcla y escalares reactivos para las fases temporales y las regiones espaciales de la llama. La satisfactoria concordancia con los resultados experimentales muestran que el concepto flamelet, y más particularmente el modelo ADF, es adecuado para las simulaciones de chorros diésel. / [CAT] L'objectiu d'esta tesi és la investigació i anàlisi de l'estructura interna dels dolls dièsel reactius i l'efecte de les condicions de contorn en els paràmetres associats a la combustió. Este objectiu s'aconsegueix per mitjà de la simulació numèrica del doll amb models de turbulència RANS i LES usant un model de combustió avançat basat en el concepte flamelet. Per a este estudi, s'aplica una aproximació simplificada de les flamelets de difusió, conegudes a la literatura com Flamelets de Difusió Aproximades (ADF en anglés), com a fonament del model de combustió. En una primera etapa, el model es valida amb combustibles de diferent complexitat química en règims estacionaris i transitoris per al conjunt de possibles velocitats de deformació. Una vegada es confirma la seua idoneïtat per a condicions trobades en dolls dièsel, s'aplica a la simulació del doll A del Engine Combustion Network (ECN), representatiu de dolls dièsel. Per a proporcionar un cuadre complet dels fenòmens subjacents, la combustió s'analitza inicialment per a condicions homogènies i flames laminars per a les distintes condicions de contorn d'aquest experiment. Després esta anàlisi es complementa amb la simulació de diferents mecanismes químics per a determinar com les característiques de l'encesa predites per l'esquema d'oxidació afecten la propagació de flama. Els resultats obtinguts en esta etapa s'enllacen amb l'anàlisi del doll turbulent en el context de simulacions RANS i LES per a descriure com el fenomen de la combustió es modifica amb els diferents nivells de complexitat física. L'estructura del doll turbulent es descriu profundament per a les distintes condicions de contorn i mecanismes químics en termes de mescla i escalars reactius per a les fases temporals i les regions espacials de la flama. La satisfactòria concordança amb els resultats experimentals mostren que el concepte flamelet, i més particularment el model ADF, és adequat per a les simulacions de dolls dièsel. / [EN] The objective of this thesis is the investigation and analysis of the internal structure of diesel-like reacting sprays and the effect of boundary conditions on combustion related parameters. This objective is achieved by means of the numerical simulation of the spray with RANS and LES turbulence models using an advanced combustion model based on the flamelet concept. For this study, a simplified approach for diffusion flamelets, known in the literature as Approximated Diffusion Flamelet (ADF), is applied as the basis of the combustion model. In a first step, this model is validated for fuels with different chemical complexity in steady and transient regimes for the whole set of possible strain rates. Once its suitability is confirmed for conditions found in diesel sprays, it is applied to the simulation of spray A from the Engine Combustion Network (ECN), representative of diesel-like sprays. In order to provide a complete picture of the underlying phenomena, combustion is initially analysed in homogeneous conditions and laminar flames for the different boundary conditions of this experiment. Later, this analysis is complemented with the simulation of different chemical mechanisms in order to determine how the ignition characteristics predicted by the oxidation scheme affect to the flame propagation. The results obtained at this stage are connected with the analysis of the turbulent spray in the context of RANS and LES simulations as a way to track how combustion phenomenon is modified at the different levels of physical complexity. The turbulent spray structure is thoroughly described for the different boundary conditions and chemical schemes in terms of mixing and reactive variables for both temporal phases and spatial flame regions. The satisfactory agreement with experimental results shows that the flamelet concept, and more particularly the ADF model, is suitable for diesel-like sprays simulations. / Pérez Sánchez, EJ. (2019). Application of a flamelet-based combustion model to diesel-like reacting sprays [Tesis doctoral no publicada]. Universitat Politècnica de València. https://doi.org/10.4995/Thesis/10251/117316 / TESIS