5. Future developments
Although uncertainties still exist concerning modelling
of the complete hydrodynamics within hydrocyclones,
significant progress has been made. The
following identify present research objectives in order to
progress hydrocyclone modelling towards a complete
description:
For the turbulence modelling, the applied highorder
differential-stress model as applied for the research
of Cullivan et al. (2003) was not identified as the
appropriate level of turbulence modelling, but rather as
a lower bound. It remains to identify the importance of
aspects such as non-equilibrium turbulence (short residence-
time), wall effects such as intermittency (affecting
turbulence generation) as well as appropriate turbulence
modelling adjacent to the air-core.
Although these issues are challenging, an initial address
of these may include application of large-eddy
modelling coupled with a knowledge base developed
through small scale detailed modelling and measurements
of the turbulence wall boundary and air-core
interface phenomena.
The issue of accurate high-solids multiphase modelling
is of paramount importance to the CFD community.
The DNS technique is being considered as a
component of a multi-resolution approach to guide
construction of models and to generate correlations for
development of constitutive relations. Certainly the
DNS approach is not straight forward and requires
careful determination of the appropriate boundary
conditions for the fluid particle interactions.
To develop automatic design methods, the numerical
simulations need to be combined with automatic
search and optimisation procedures. The principle idea
is to define a set of geometry design parameters in the
form of weights ai applied to a set of shape functions
LiðxÞ for example, such that the shape is represented as
Pai LiðxÞ. Then the cost function I measuring the
performance of the system is selected, which may be
the cut size parameter or the sharpness of separation.
The sensitivities dI=dai may then be estimated by making
a small variation dai in each design parameter in turn
and recalculating the flow to obtain the change in I.
Such an embedded procedure would allow for optimisation
of hydrocyclone design to reflect complex multiobjective
requirements.
Parallel computing is essential for such multiobjective
hydrocyclone design. However, load balancing
(assigning equal load to processors) could be challenging.
Utilization of advanced solving strategies for complicated physical phenomena in conjunction with
highly complex hydrocyclone geometries may show a
dynamic behaviour with regard to computing time per
grid point. Therefore, the same workload for each processor
during the computation can only be achieved by
ensuring dynamic load balancing. In addition, the
numerical algorithm may cause a nonlinear increase in
communication demand.
Validation of proposed predictions of the complex,
highly turbulent, high-solids multiphase flow within the
hydrocyclone presents a considerable challenge. Research
is underway to address this issue. In particular
laser techniques such as dual-planar laser induced fluorescence
are being investigated for their potential resolution
of the hydrocyclone flow-field detail.
Tomographic methods are being assessed and their
development investigated in order to provide solids
concentration profiles, as well as the velocity field
through cross-correlation techniques. In particular research
is focussed upon the application of X-ray and
electrical impedance tomography. More information
regarding these issues may be found in Cullivan et al.
(2001).