Shape optimization of turbine blade firtrees

Abstract

The effective application of various optimisation techniques including classic gradient-based and modern evolutionary computation methods in engineering design practice can not only deliver better quality, but also shorten design cycle time. However, the success of using these techniques relies on a number of factors, such as efficient design parameterisation, complete automation, and expertise in deploying various search tools and managing the high computational cost associated with the use of high-fidelity simulation code. A CAD-based shape optimisation method is investigated in this work using knowledge-based ICAD* system with focus on the optimum shape design of turbine blade firtrees. The design of such a structure component involves a large number of constraints derived from industrial experience. The overall aim of this work is to employ some effective and efficient search techniques to explore various new shapes based on an automated design-to-analysis integration, which is achieved by incorporating a knowledge-based intelligent computer-aided design system (ICAD) into the process using sequential rule-based modelling methods. Analysis-related information as well as geometric data is integrated together to produce a general template for the firtree. A high-fidelity finite element analysis code is used as the assessment tool of structural strength, and different types of stress criteria are used in the formation of the optimisation problem. Both the existing shape features inherent to CAD systems and new features offered by the use of free-form shape modelling using Non-Uniform Rational B-Splines (NURBS) are investigated. This leads to a combined feature-based and free-form shape parameterisation method. A two-stage (Genetic Algorithms + Local Search) procedure is used in order to make use of the advantages offered by these two methods while overcoming some of their weaknesses. The problem of high computational cost problem is also tackled by the efficient use of a Gaussian Process based surrogate model coupled with Genetic Algorithms. Both the combined shape parameterisation methods and framework for incorporating surrogates with GA can be applied to general engineering design problems.

More information

Identifiers

Catalogue record

Export record

ASCII CitationAtomBibTeXData Cite XMLDublin CoreDublin CoreEP3 XMLEndNoteHTML CitationHTML CitationHTML ListJSONMETSMODSMPEG-21 DIDLOpenURL ContextObjectOpenURL ContextObject in SpanRDF+N-TriplesRDF+N3RDF+XMLRIOXX2 XMLReferReference ManagerSimple Metadata