An experimental and numerical study on the smoke layer growth and movement within large-volume spacesatrium fires

Resumen

This thesis aims to assess smoke behaviour in atrium fires. To this end, 27 new full-scale experimental fire tests in an atrium and a series of numerical Computational Fluid Dynamics (CFD) models are presented. Moreover, with the aim of reducing the excessive computational cost that the simulation of these large volume spaces require, a new methodology based on the Fractional Factorial Design (FFD) is proposed to predict fire scenarios. Finally, with the objective to verify the existing codes and standards, an analysis of the influence of the make-up air on smoke production is performed. The thesis starts with an overview of smoke management systems. This overview identifies smoke hazard parameters, and describes the main fire stages as well as the most important fire phenomena when this system is activated in atrium fires. Subsequently, a literature review of atrium fires involving smoke production and exhaust is provided, being complemented with a review of the few full-scale fire tests carried out in atria reported in the literature. These revisions provide the framework of the research hereinafter carried out. Next, the novel full-scale fire tests carried out in the ``Fire Atrium'', located in Murcia (Spain), are exhaustively described, including the different devices and sensors employed. The experimental results serve as a benchmark data for numerical modelling as well as for profound understanding of the smoke behaviour under both different make-up air conditions and steady and transient exhaust flow rates. The numerical models are elaborated and adjusted using Fire Dynamics Simulator (FDS, version 6), which is the main fire modelling software, in order to be validated as a tool to simulate the smoke behaviour in atrium fires. Particular emphasis is placed on the smoke response under different transient exhaust flow rates. The results show that, when the experimentally measured heat release rate (HRR) is introduced as an input to the models, fire scenarios are accurately reproduced. In order to mitigate the excessive computational cost of the atrium fire simulations, a novel methodology based on a set of short-listed simulations defined by the FFD method to predict atrium new fire scenarios is presented. Once a variable of interest in an atrium fire scenario is identified, such as the smoke layer drop, this methodology shows the influence on this variable of each of the main factors that define the fire scenario, such as the HRR or the exhaust flow rate, among others. Furthermore, in order to validate the proposed approach, a comparison with the aforementioned full-scale fire tests is performed, also considering the specific corresponding FDS models. The use of this methodology allows performing representative sensitivity analyses and also the precise prediction of smoke temperatures as well as the smoke layer position in case of atrium fires. Finally, the thesis addresses the make-up air influence on the smoke production by the flow patterns generated around the flame. On the one hand, a numerical-experimental comparison shows the effects of the domain extensions at the openings in the simulations and the cases in which its use is strongly recommended. On the other hand, the fire plume and flame deflections caused by the flow patterns induced by the make-up air are evaluated numerically in detail focusing on the inlet velocity, opening distribution and the pan location. From this study, symmetric opening distributions are recommended, with inlet velocities below the prescribed limit by the codes and standards. However, in asymmetric inlet distributions, important flame perturbations are encountered implying that a further analysis of this prescribed limit is advisable.