No abstract

Steady-state axisymmetric simulations using the Reynolds-Averaged Navier-Stokes equations have been carried out in order to optimise the performance of a Chemical, Biological, Radiological and Nuclear (CBRN) canister filter for its use in a powered air-purifying respirator (PAPR). Alterations have been made to the shape of the canister, the spacing of the rear wall of the canister with regard to the carbon filter, and the bracketing between (i) the particulate filter and the carbon bed and (ii) the carbon bed and the canister wall. The pressure drop across the canister and the residence time distribution at the rear of the carbon bed have been analysed in detail based on an extensive parametric analysis involving the aforementioned variations. It has been demonstrated that the non-uniform porosity profile of the carbon bed resulted in alternating regions of high and low velocity close to the canister wall, providing a possible route for breakthrough. Designs, which included a bracket at the rear of the carbon bed, blocked this route and consequently had a longer minimum mean residence time than those, which did not. It has also been shown that the spacing between the carbon bed and the canister rear wall had a large impact on both residence time and pressure drop. In cases where the carbon backed directly onto the canister rear wall flow in the axial direction from the outside wall towards the canister axis resulted in far greater pressure drop and a reduction in minimum mean residence time within the carbon bed.

A numerical model has been developed with the intention of better understanding the relationship between flow characteristics and adsorption behavior in small activated carbon beds. The model uses the Dubinin–Radushkevich adsorption isotherm with Linear Driving Force kinetics and has been validated against experimental data for the breakthrough of cyclohexane. Simulations have been carried out varying the inlet flow rate and the adsorption kinetic constant. The adsorption behavior has been analyzed by assessing the furthest point into the bed that a threshold concentration of a contaminant could be seen in the vapor phase. It was seen in all cases that the adsorption was a two-stage process with a rapid initial penetration into a certain bed depth followed by a relatively slow gradual progression of the front through the bed over time. The initial distance penetrated into the bed has been seen to be inversely proportional to the kinetic constant, the same trend as predicted from a plug flow mass balance analysis. The rate of steady front advancement after this showed a slight tendency to increase with decreasing adsorption kinetics, compared to the expected plug flow speed.

A model has been developed to represent the stochastic nature of the activated carbon bed within a generic chemical biological radiological and nuclear (CBRN) respirator canister. The porous region is subdivided into discrete sections which are assigned a porosity based on their radial position according to a longitudinally‐averaged porosity model, and then perturbed by some amount according to a Gaussian distribution. The porosity model was used in Reynolds‐averaged Navier‐Stokes (RANS) simulations in order to assess the impacts that the choice of section size and the porosity variation would have on the pressure drop and residence time distribution. It was shown for small section sizes that increasing porosity variation would increase pressure drop and minimum residence time in the carbon bed, while decreasing the average residence time. As the section sizes became larger the reverse trend was seen as a greater extent of flow channelling throughout the bed became apparent. For the given domain size, there was an upper limit to the section size, beyond which statistical convergence could not be guaranteed. It was also shown that for section sizes close to the particle diameter, the results would depend only on the ratio of section size to porosity variation, reducing the porosity model to a single parameter. A few selected cases were simulated at higher flow rates, where the previously mentioned trends were seen to persist.

No abstract