Numerical modelling of fluid flow through a powered air-purifying respirator filter

Abstract

Powered Air-Purifying Respirators (PAPRs), in which air is forced through a filter by means of an integrated motor and blower can provide a significantly higher level of protection than other types of respirator. There is considerable scope for improvement over existing designs, particularly with regards to breath-responsive PAPRs, which provide air only in response to the wearer’s inhalation. Understanding the flow of air through such respirators presents a challenging fluid dynamics problem. In this thesis, the state-of-the-art of respiratory protection is reviewed in order to identify shortcomings and potential areas of improvement. A system-level model has also been developed, which considers each of the key components of a PAPR and presents a simplified depiction of various PAPR parameters under a range of operating conditions. This tool has been designed to integrate more in-depth models of the various PAPR components as they are developed. The area with the greatest scope for development was found to be the filter. A series of computational fluid dynamics (CFD) simulations using the finite volume method to solve the discretised Reynolds-averaged Navier-Stokes (RANS) equations were conducted based on a typical Chemical Biological Radiological and Nuclear (CNRN) filter canister. An initial analysis has been conducted to measure the impact of canister geometry on performance metrics such as the pressure drop and the residence time distribution throughout the adsorbent bed by performing a parametric analysis of steady-state axisymmetric simulations. A selection of these geometries were then investigated under transient flow conditions. A full 3D model of an existing canister geometry has then been developed and simulations are conducted to investigate the extent to which canister performance under steady flow conditions can be used to predict flow behaviour when under a range of real human breathing patterns. A model to predict adsorption performance of volatile organic compounds on activated carbon has been developed in conjunction with experimental adsorption data, with the intention of relating adsorption behaviour to flow. The distribution of porosity throughout the adsorbent bed has been shown to have a considerable impact on the distribution of flow. A new method has been developed to better represent the porosity profile by subdividing the bed into discrete sections and assigning a local porosity based on a longitudinally-averaged porosity profile which has been pseudo-randomly perturbed. This porosity model has then been implemented within a 3D CBRN canister model to investigate how different parameters of the porosity model will affect flow. Axisymmetric simulations have been carried out on a simplified carbon bed geometry using both of these novel adsorption and porosity models in order to gain a better understanding of how the residence time distribution can be related to adsorption.