Study, design and validation of a framework model for smoke and particle-filled atmospheres

Resumen

Vision is by far the most important source of information regarding situation awareness in humans. This is relevant in the case of temporary loss of vision for certain professionals due to safety implications -specially when conducting exposed tasks-. This thesis makes a number of contributions to ease the future development of devices able to overcome the risks and problems related with that kind of temporary blindness. First, a concept overview of the problem from a remote sensing perspective is given. An analysis of the state of the art of several technologies used today in applications involving loss of vision in harsh environments is conducted. Their possibilities are studied from a theoretical point of view as well as from the user's perspective in order to determine advantages, disadvantages and possible areas of improvement. The theoretical and practical aspects of different signal domains and their application to artificial vision systems are studied next, surveying their suitability to imaging in media such as particle-filled atmospheres. A comparison between electromagnetic and pressure waves as a means of sensing the environment is made, extracting conclusions pointing to the need to explore new technologies to enhance the performance of current systems. Two cases of study are defined, one related to fire suppression operations and another one to low visibility underwater operations. The characterization of the disturbing medium is identified as the cornerstone in the modeling of signal propagation through particle-filled atmospheres. In that sense, an in-depth analysis of smoke as a propagation medium is carried out. The characterization of smoke as a suspension of particles in an atmosphere with a variable degree of combustion products is conducted. Main characteristics of smoke are also studied: particle sizes, optical properties, motion, etc. A reference set of parameters to feed a computer model is defined. Next, the generalization of those characteristics to other environments such as turbid water is done, although the present work is more focused on smoke-filled atmospheres. With the medium completely characterized, it is possible to create an event-based, time resolved Monte Carlo model to describe interactions between an incident electromagnetic radiation and a particle-filled atmosphere. The model is conceived as a framework model: a wide scope of particle-filled environments with arbitrary particle types and optical properties can be simulated without major changes. The definition of spatial environments with objects is also included. Although primarily intended for use in optical and infrared wavelengths, the model can be generalized to other parts of the spectrum as well. This model, along with the characterization of the medium, are the main original contribution of this PhD thesis. The implementation of the model in a highly parallel computer architecture for higher performance is presented and discussed subsequently. The identification of performance drainers, and the analysis of the underlying hardware for performance optimization have been worked out. Two implementation strategies were used and analyzed, extracting practical hints to maximize the effectiveness of the optimizations implemented. The huge amount of data generated potentially by the model makes it convenient to have some sort of tool for rapid analysis and exploitation of the results obtained. A visual tool has been developed as a proof of concept, and has proven to be very useful in practice. Finally, an experimental setup was designed and built to validate the model against real world tests. The workflow necessary to provide with a rapid design of a cost-effective instrument to run the experiments was planned, making use of advanced techniques such as parametric design. The instrument was fabricated using additive manufacturing techniques (3D printing), and included a way to inject and confine the medium inside an ad-hoc built chamber. A procedure to create stable and reproducible atmospheres at a microscopic level using calibrated microspheres was also designed and tested, accounting for an original contribution of the present thesis. Several experiments were run and results were obtained. In this first phase of the experiments, qualitative data showing good agreement between simulations and experiments was obtained. Due to the capabilities of the hardware used, the obtention of accurate numerical data is left for future work, along with the exploration of –among others–, compressed sensing techniques for signal processing.