 |
Center for Nonlinear Studies |
As the UK nuclear sector prepares for largescale decommissioning, the amount of effluent containing contaminated colloidal particles is expected to increase. Traditionally, the approach to cleaning of this effluent is tiered. Starting with the coarser grained particles and finishing with the much smaller scale radionuclides. One area which is expected to be exposed to a large number of heterogeneous particulate materials is the sand bed filtration system. This system relies on a graded bed structure in which the larger particulates are captured earlier within the system and finer particulates are captured later. This bed system, is well modelled using traditional continuum dynamics modelling in which the amount of mass within the system is predicted. However, it is believed that this model can significantly under- or over-estimate the amount of captured materials due to its treatment of the deposition behaviour. With this in mind, the behaviour of the bed can vary significantly as particulates which were once captured can become de-attached and penetrate deeper into the bed. Overtime, this can produce a significant pressure drop across the bed accumulating in the breakthrough of undesirable products into more sensitive regions of the clean-up procedure. By understanding the localised interactions around the collector (sand grain) it is possible to refine the deposition behaviour and hence, increase the accuracy of the continuum model. To achieve this, two mesoscale models were coupled. Firstly, a lattice Boltzmann model was constructed and introduced into the agent based modelling suite NetLogo and secondly, an agent based model was produced. Lattice Boltzmann modelling is a well-established way in predicting fluid flow through porous media. The lattice Boltzmann model is based on the lattice gas model, however, this approach tracks the distribution of the lattice particles. By solving the lattice Boltzmann equation, the implementation of this technique is significantly simpler. Agent based modelling, is a newer technique which can trace its origins back to complex adaptive systems and the Game of Life models. It assumes that localised interactions between agent’s results in emergent behaviour on the macro-scale. These agents exist as entities which can interact with each other along with the environment in which they are suited. The coupled model proposed here, uses the lattice Boltzmann model to produce the fluid environment around the collector at Stokes flows (Re < 1). Agents are then introduce into the model, these agents have physical parameters (i.e. mass, volume). By analysing the interactions between the agents and the collector surface the deposition behaviour can be predicted and hence, the large scale model can be refined.
Host: Blas Pedro Uberuaga