APEX - Awareness and Promotion EXercise
by Mike Ashworth and David Emerson
APEX is a two-year project funded by the European Union ESPRIT
IV programme. Its aim has been to promote the benefits of using
High Performance Computing (HPC) in combination with Computational
Fluid Dynamics (CFD) to support the next generation approach to
industrial research and development in the process industries,
with emphasis on the use of Information Technology for simulation
and prediction. A multi-media CD-ROM describing the benefits of
CFD and HPC has been widely and freely distributed throughout
the European Community.
The very fast pace of change within HPC hardware and continuous
developments of CFD software require a way of conveying information
very rapidly. A specific goal of the APEX project was to develop
a multimedia CD-ROM that could be distributed to relevant industrial
companies. The CD contains detailed information about HPC and
CFD and results from two industrial applications from the ESPRIT
III HP-PIPES project (#8114). An additional important feature
is a Cost Benefit Analysis (CBA) section that will allow senior
management to consider the investment decisions and risks concerning
such technology. The partners involved in APEX are:
- CLRC Daresbury Laboratory, Warrington, UK, responsible for the
HPC section and industrial application of Tioxide reactor
- Instituto Superior Tecnico, Lisbon, Portugal, responsible for
the CFD section and industrial application of EDP power station
burner
- Paras Ltd., Isle of Wight, UK, responsible for the CBA section
and project management.
The APEX CD highlights two industrial applications from the process
industries. In collaboration with industrial partners, the HP-PIPES
project developed a CFD code to solve the 3D incompressible Navier-Stokes
equations and to augment the core code with modules describing
the physics and chemistry of particular applications. It was demonstrated
that advanced numerical simulation can be used as a realistic
predictive tool for the design and operation of complex chemically
reacting flows and combustion systems. The core code, developed
at Daresbury Laboratory, describes fluid flow and heat transfer
in process engineering systems and has been designed to exploit
the power of massively parallel high performance computing systems.
It has been written in a portable and modular form to facilitate
the future development of modules for the specific features of
particular applications.
An Industrial Application Example: Production of Titanium Dioxide
Tioxide plc, formerly owned by ICI, is one of the world’s largest
producers of the white pigment titanium dioxide that is widely
used in materials such as paints because of its good light scattering
characteristics. The critical parameters of the product are the
mean size of the particles, which dominates its scattering power,
and the distribution of the different particle sizes about the
mean diameter, which affects the range of wavelengths that are
scattered and the resulting effectiveness of the pigment in materials
such as paints.
The HP-PIPES code was designed to overcome limitations in the
existing model of the titanium dioxide process, in particular
in the areas of the physical and chemical modelling and in the
resolution of the flow field. At the start of the project axi-symmetric
simulations were taking approximately one month to converge. The
parallel implementation reduced simulation times to a few hours.
This opened up the possibilities of exploring in more detail the
phase space of physical and chemical parameters embedded in the
model, enabling improvements in the theoretical bases of the various
physical and chemical sub-models and moving the models to three
spatial dimensions better to simulate real reactor geometries.
The ultimate aim of the HP-PIPES project was to develop a predictive
model capable of determining the optimum reactor configuration
and operating conditions so as to produce particles of a specified
mean diameter and standard deviation.
The Titanium Dioxide chloride process occurs in a basic pipe reactor.
The oxidation process involves several complicated stages which
may include the nucleation, growth and coagulation of the titania
particles and the dissociation and recombination of chlorine produced
in the reaction. In general the reactors have multiple inlets
through which the feed materials, titanium tetrachloride and molecular
oxygen, are fed at various temperatures, velocities and mixture
fractions. The products, and any unreacted input materials, flow
through the reactor at high speed. Initially, titania particles
are assumed to form as nuclei and these become larger through
surface deposition and coagulate as they flow through the reactor.
These various processes produce a continuous size distribution
of titania particles with a broad distribution of sizes.
The particle growth process is modelled by discretising the size
distribution in terms of a fixed number of size intervals characterised
by a mean diameter. Each of the size classes is then treated as
a distinct transported variable in its own right The models can
employ anything up to 100 class sizes for high accuracy computations.
The calculations of the various source terms describing the nucleation,
growth and coagulation is extremely time consuming as the coagulation
terms fully couple all the size class variables in each finite
volume. It is obvious that three-dimensional computations for
systems with this number of class size variables would be completely
impractical and uneconomic in the absence of parallel systems.
The APEX CD-ROM
The APEX CD-ROM is an interactive multi-media presentation, which
introduces CFD and HPC, shows results from two typical applications
(a chemical reactor and a power station boiler) and guides the
viewer through the investment decision process.
The CD-ROM is available cost-free from the contact address below
or by filling in the electronic form on our Web Page http://www.dci.clrc.ac.uk/Activity/APEX
Please contact:
Mike Ashworth - CLRC Daresbury Laboratory
Tel: +44 1925 60 3663
E-mail: M.Ashworth@dl.ac.uk
