S&C Thermofluids Ltd CFD modelling of adsorption in carbon filters E Neininger*, MW Smith** & K Taylor* * S&C Thermofluids Ltd ** Dstl, Porton Down

Overview Background to filter model development Physics of adsorption modelling Validation Implementation in PHOENICS Future developments

Typical filter application Air Flow Impregnated granular activated carbon Glass Fibre Filter Canister filter for respirator

Modelling Requirements Pressure drop Contaminant breakthrough time

Other filter geometries Small scale filter test bed - 2 cm diameter carbon bed Carbon monolith filter - Courtesy of MAST

Flow through filter bed

Flow through packed bed Pressure drop - local voidage distribution coupled to Ergun equation for pressure loss through bed: p/L = 5 S o 2 (1- ) 2 U/ S o (1- ) U 2 / 3 viscous loss turbulent loss - earlier work using this equation given good agreement with experimental data for pressure drop. Voidage distribution - Mueller model good for uniform spherical particles - uniform voidage gives better comparison with measured breakthrough times for granular carbon

Adsorption rate Two scalar equations solved - one for transport of contaminant vapour - one for rate of uptake of adsorbed phase A linear driving force approach is used for the adsorption rate, whereby this is proportional to the amount of remaining capacity - C/ t = 1/ S o k m (C - C i ) Equilibrium uptake determined by adsorption isotherm = f(C,T)

Adsorption isotherm Pentane adsorption isotherm on BPL carbon at 295K X – experimental data __ - Dual Dubinin-Astakhov equation

Validation Breakthrough of pentane (3lpm flow, 295K, various bed depths)

Validation Breakthrough of pentane (3lpm flow, 1cm bed depth, 295K)

Saturation of filter bed

Variable inlet concentration Outflow concentration from pulsed inflow With no filter 0.5cm filter – experimental 1cm filter – experimental 0.5 filter – CFD 1cm filter - CFD

Adsorption in wet air - C/ t = 1/ S o k m (C - C i ) but C i for pentane limited so that uptake </= total pore volume - water uptake

Implementation in PHOENICS Pre-processor User interface allows rapid input of geometry and property data Writes Q1 file and runs FEMGEN to create mesh Run steady-state to establish flowfield then transient to model adsorption Run full transient if inlet flowrate varies with time

Implementation in PHOENICS Pre-processor User interface allows rapid input of geometry and property data Writes Q1 file and runs FEMGEN to create mesh Run steady-state to establish flowfield then transient to model adsorption Run full transient if inlet flowrate varies with time

Implementation in PHOENICS Pre-processor User interface allows rapid input of geometry and property data Writes Q1 file and runs FEMGEN to create mesh Run steady-state to establish flowfield then transient to model adsorption Run full transient if inlet flowrate varies with time

Implementation in PHOENICS Customised GROUND Coding Pressure drop and adsorption source terms Outlet contaminant concentration can be monitored as run progresses Modelling issues Cell blockages

Monolith filter model Activated carbon monolith Low pressure drop Single channel model detailed model of one flow path contaminant diffuses into porous monolith can model several monoliths in series

Monolith – vapour concentration vapour concentration after 6 mins outlet vapour concentration vs time

Future development of model Multiple adsorbents Non-linear driving force for adsorption Property database/GUI Heat of adsorption source terms Improved solver speed - optimisation of GROUND coding - parallel processing

Conclusions Requirement for CFD modelling of filters CFD model of adsorption process developed Validation of packed bed model promising Monolith model requires validation Customised user interface Ongoing developments