Abstract:
Numerical
optimisation methods have successfully been used for a variety of
aerodynamic design problems over quite a few years. However the application of these
methods to the
aerodynamic blade shape optimisation of industrial axial fans has
received much less attention in the literature probably given the fact that the majority of
resources available to
develop these automated design approaches is to be found in the
aerospace field.
This work
presents the development of an automated design process which was
developed to aerodynamically optimise the fan blade geometry. It involves the
application of Genetic Algorithm (GA) methods to the aerodynamic shape optimisation
of a two-dimensional axial fan cascade as well as the development of a three-
dimensional
shape optimisation routine. Navier-Stokes CFD codes were used for the 2
and 3-D
analyses using steady simulations.
The effects of the Variation of the control parameters on the performance of the GA as
a
optimisation tool is presented. The tournament selection, uniform crossover, creep
mutation scheme with elitism appears to work the best of this application. The
parallelisation of genetic algorithm was also developed using the Message Passing
Interface
(MPI) scheme. This essentially reduces the running time for each generation to
the amount of time
required for performing the genetic operations on just one
individual.
The
aerodynamic optimisation of a low speed fan cascade based on genetic algorithm is
presented. A commercial turbo machinery CFD code, CFX-TASCow, was used in the
evaluation of the
objective function. The optimisation process reduces the total pressure
Aerodynamic Optimisation of Industrial Axial Fan Blade O.Lotï
¬
iii
Executive
Summary
Cfae UNIVERSITY fold
loss while
maintaining the same loading and mass flow rate. This development is
related to a
significant change in profile curvature in the vicinity of the trailing edge.
The flow field inside low
speed axial fans characterized by low hub-tip ratio can be
highly three-dimensional and particularly complex. As a consequence three-dimensional
automated
design process was developed to aerodynamically optimise the fan blade
geometry taking account of the predicted three dimensional flow. The optimiser
employs a genetic algorithm for global optimization purposes and is coupled to the
academic Navier-Stokes solver MULTIP. The numerical investigation of the overall
performance, efficiency and work-input characteristics of the fan were found to agree
well with the
previously reported experimental results.
The
optimization task is accomplished by modifying the blade camber line, lean and
sweep while keeping the blade thickness distribution and mass flow rate constant. The
optimisation process demonstrated that the fan efficiency can be improved by changing
the
profile curvature and giving the blade a proper forward sweep. Nevertheless the
effect of introducing lean and backward sweep did not improve the fan performance for
this
particular application.
This
study demonstrated that the present method offers a promising approach to
industrial axial fan
designers to help design better machines while contributing to the
softening of the design cycle. The results obtained show that the genetic algorithm
when
coupled to a CFD tool is not only capable of achieving a improvement in the
designs of existing axial fan blades effectively but also that they achieve these results
with a minimum amount of user
expertise.