Use this URL to cite or link to this record in EThOS:
Title: Design optimisation of centrifugal pump impellers using parallel genetic algorithm
Author: Wahba, Waleed A.
Awarding Body: Cranfield University
Current Institution: Cranfield University
Date of Award: 2001
Availability of Full Text:
Access from EThOS:
Access from Institution:
Computational Fluid Dynamics (CFD) techniques have settled to a stage, where it is possible to gain significant insight into fluid flow processes of turbomachinery. However, the purpose of fluid dynamics naturally goes beyond improved understanding to the aim of improving the performance of the engineering systems. Consequently, the present thesis investigates the use of a automated design optimisation method using CFD. This presents a new design method for a important turbomachinery part, blade profiles of centrifugal pump impellers, based on a shape optimisation algorithm in combination with CFD. The use of genetic algorithms in shape optimisation dose not allow the design engineer to use any derivative information on the evolution of the shape, but only simple evaluation techniques. A optimisation library (GAlib), based on a genetic algorithm (GA), was used. GA controls the evolution of a population of profiles towards an optimum design. The optimisation process can handle simple objectives as well as conicting ones. The fitness value of each population element is evaluated using a CFD flow solver (Mac_LNS) based on nite-difference discretisation of the incompressible, Navier-Stokes (N-S) equations on structured polar-coordinate meshes. A number of design examples have been developed and the behaviour of the genetic algorithm has been tested using different kinds of objective functions. I addition, the algorithm was tested with a multi-objective mction. Bézier curves were selected to represent the impeller profile. A symmetric profile, identical profile for the pressure side (PS) and suction side (SS), was used as a basic shape to generate the population elements. GAlib was modified to run as a parallel algorithm using Message Passing Interface (MPI). It is indicated that parallelisation using MPI is good technique to overcome the time taken by GA and CFD, and quite good optimisation convergence criteria was obtained by using parallelisation. The obtained results show that the genetic algorithm is capable of achieving satisfactory designs of centrifugal impeller blade profiles effectively and with a minimum amount of user expertise.
Supervisor: Elder, R. L. ; Tourlidakis, A. Sponsor: Not available
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available