Developing guidelines for modeling air pollutants dispersion

This Ph.D. thesis is devoted to the analysis and development of numerical models to predict the dispersion of pollutants and hazardous substances, with a focus on Computational Fluid Dynamics (CFD) models. Indeed, numerical models are needed to aid the Decision Support System for environmental impact assessment and industrial safety studies. Some models have been massively employed through the years, such as gaussian, lagrangian or integral models. However, they can suffer from inaccuracies especially when dealing with local scale cases and complex layouts. For this reason, CFD models can be considered a reliable and promising tool in the context of dispersion studies, to overcome these drawbacks. The present thesis aims at improving the CFD modeling of gas dispersion in complex industrial layouts. The first improvement in the use of CFD in gas dispersion analysis is the simulation of an accidental LNG release. The LNG pool formation is studied with an integral model while the subsequent methane dispersion is investigated with CFD. In this case, the pool boundary condition is treated in the CFD model as variable with time, enabling for a more precise analysis of the dynamic of gas dispersion. Moreover, the influence of an obstacle, with dimensions similar to those of a perimeter wall is investigated involving all the possible atmospheric stability conditions. The CFD models improvement is also performed through the application of an Uncertainty Quantification (UQ) technique to calibrate parameters and also through a novel CFD model for the analysis of the Atmospheric Boundary Layer (ABL). In particular, the first strategy is based on the use of the generalized Polynomial Chaos (gPC) expansion for the calibration of two uncertain boundary conditions in the simulation of an accidental release of LNG, which are the radius of the LNG pool and the vaporization rate. The UQ technique is applied to a well-known experimental series of tests, which is employed as a reference to perform the calibration. The methodology is able to provide values of these parameters that ensure an optimal setup of the CFD model and, moreover, it allows to have a deep understanding of the influence of important physical phenomena on the gas dispersion, such as convection and diffusion. Another improvement is focused on the description of the characteristics of the ABL within a CFD model. In particular, a novel sub-model which employs the SST k-ω turbulence model is implemented ensuring the consistency with the ABL features in open field simulations. In addition, a strategy to consider the presence of obstacles in the domain is developed within this model performing a blending from a turbulence model to another one in the region influenced by the presence of obstacles. Comparison with experimental data of a single building obstacle and an array of buildings shows that the proposed model is perfectly able to reproduce the ABL characteristics both in the undisturbed region and in that influenced by the buildings. Beside the improvement of CFD models for gas dispersion analysis, some applications to real case studies are carried out demonstrating the reliability of this kind of models in aiding the Decision Making for the environmental and safety management of industrial sites. In particular, the above mentioned gPC expansion is applied to a real industrial case study, in which the accidental release of LNG is considered in a pipeline of a storage plant. The technique is employed in this case to investigate the influence of variable meteorological conditions, in terms of wind direction and speed. The UQ methodology allows to build concentration maps at the Lower Flammable Limit (LFL) level depending on the variation of input parameters. In this manner, the mean and maximum maps, computed employing the standard deviation of the dispersion results, are provided. Therefore, this particular study extends the common practice of considering only the most frequent wind conditions in dispersion and consequence assessment studies. One further application of CFD to aid the industrial safety in a real case is presented considering the implementation of a novel algorithm for the optimal gas detectors placement in a storage plant. CFD simulations are performed to analyse the dispersion of a vapour cloud generated by the accidental release of gasoline from a storage tank. The CFD output is then employed to develop a strategy for the computation of the optimal economic number of sensors based on a multi-criteria analysis, coupled with another algorithm which provides the optimal locations of sensors in the plant