Experimental and Theoretical Studies of Contaminant Transport Due to Human Movement in a Hospital Corridor

This thesis considers numerical, physical and theoretical modelling approaches to investigate the influence of a person's wake on the dispersion of an airborne pathogen in a hospital corridor and the implications this has in terms of infection risk. The various physical and computational modelling approaches were conducted using geometries corresponding to a 1:15 reduction in length-scale, when compared to the full-scale, and then interpreted in the context of a full-scale scenario in a hospital corridor. The movement of people in a corridor was approximated using a translating circular cylinder. A physical water-bath model was used to investigate contaminant transport using food-dye in a channel with different sized cylinders and translation frequencies. Dye concentrations were quantified through a calibration method dependent upon changes in light-intensity, leading to accurate tracking of the dye and allowing the amount of dye in different regions of the water-bath to be calculated over time. The centre of mass of the dye cloud was found to be dependent upon the square root of the translation frequency, amplitude, cylinder diameter and elapsed time. Based on the hypothesis that the dispersal of the dye could be described by a turbulent diffusion process, a theoretical model was constructed to predict the evolution of the dye concentration using a Gaussian function, which agrees well with experimental data for a broad range of cylinder diameters and translation properties. Two and three-dimensional computational fluid dynamics (CFD) models were developed to investigate the transport of a passive scalar due to a translating cylinder in a channel, their geometries and boundary conditions bearing close resemblance to the water-bath. Seven turbulence models were tested to determine the most suitable, using the water-bath data for validation. The shear-stress transport (SST) model was found to offer solutions in closest agreement with experimental results and theoretical predictions, as well as offering up to a 70% reduction in computation time compared to SAS, DES and LES turbulence models. The commonly used k-epsilon model was found to be inappropriate for modelling the flows encountered here. The numerical and theoretical models were used to investigate a number of scenarios in a corridor at the full-scale where an infectious contaminant is released. This includes a unidirectional flow applied along the corridor, where it was shown that the wake of the cylinder was still able to transport contaminant `upstream' against the direction of the flow. This implies that a walking person may be able to transport an airborne contaminant in their wake even in the presence of ventilation. Infection risks were calculated for a person making a single pass and multiple passes of the corridor based on the amount of contaminant inhaled and published data on the infectiousness of different pathogens. Results showed that the theoretical model developed here led to each individual breath having its own infection risk based on temporal and spatial differences, whereas a model assuming a well-mixed contaminant distribution did not. Results demonstrate that a person's wake is likely to influence the spread of an airborne contaminant in a hospital corridor, even if ventilated within current recommended guidelines. This highlights that a person's risk of infection, in the presence of airborne pathogens, is partly determined by any human traffic passing through the space before them and not solely on any ventilation within the space, as is often assumed in airborne infection models. Furthermore, the experimental work has provided strong validation data for the CFD models and allowed for the construction of uncomplicated yet powerful theoretical models. It has been shown that, when appropriate modelling assumptions are taken, confidence can be had in CFD predictions of contaminant transport involving complex flow behaviour, such as eddy shedding, within a built environment. The study also confirms that poor selection of `default' modelling assumptions, for example use of the k-epsilon turbulence model, will provide very poor predictions, highlighting need for careful selection of each aspect of a model.