Multi-scale Model Coupling for CFD Simulations of Discharge Dispersion in the Sea
Abstract:The processes that influence the dispersion of effluent discharges in the sea occur over a wide range of length and time scales. The distance that effluent can travel before it is considered mixed can be several kilometres, whereas the turbulent eddies that affect the near-field mixing of a discharge can be as small as a few centimetres.
The range of scales that are involved mean that it is not generally practical to include all influencing physical phenomena within one model. Typically, the modelling of effluent dispersion is performed using two separate numerical models: a local model of the outlet(s), including the near-field effects of momentum, buoyancy and turbulence; and a larger scale model that can include the far-field effects of tidal-, wind- and wave-driven-currents, water depth variations, atmospheric fluxes, and Coriolis forces.
The boundary between the two models is often not strictly defined, but is usually placed at the transition from where the behaviour of the effluent is dominated by the ambient environment, rather than the discharge characteristics and outfall configuration. In most real applications, this transition line varies considerably in time and space.
This paper presents the findings of collaborative research between the Applied Modelling and Computation Group (AMCG) at Imperial College London, UK, and HR Wallingford Ltd. Results are presented using a range of coupling methods to link the near- and far-field mixing regions. An idealised domain and tidal conditions are used, with the outfall and ambient conditions typical of those found at small coastal desalination plants. Open-source CFD code Fluidity is used for both the near-field and far-field modelling. Fluidity scales well when run in parallel on large numbers of cores. It also has an anisotropic adaptive mesh capability which allows local control over solution accuracy throughout the domain. This combination means that accuracy can be achieved without excessive time costs, with nodes adjusting to efficiently concentrate computational effort only on those areas of the flow that are influential to model accuracy.
The simulation results are used to give a pragmatic cost / benefit analysis of the various coupling methods, comparing their accuracy and speed where used in outfall design studies.