Abstract
Most response surface methods typically work on isotropically sampled data to predict a single variable and fitted with the aim of minimizing overall error. This study develops a metamodel for application in preliminary design of aircraft engine nacelles which is fitted to full-factorial data on two of the eight independent variables, and a Latin hypercube sampling on the other six. The specific set of accuracy requirements for the key nacelle aerodynamic performance metrics demand faithful reproduction of parts of the data to allow accurate prediction of gradients of the dependent variable, but permit less accuracy on other parts. The model is used to predict not just the independent variable but also its derivatives, and the Mach number, an independent variable, at which a certain condition is met. A simple Gaussian process model is shown to be unsuitable for this task. The new response surface method meets the requirements by normalizing the input data to exploit self-similarities in the data. It then decomposes the input data to interpolate orthogonal aerodynamic properties of nacelles independently of each other, and uses a set of filters and transformations to focus accuracy on predictions at relevant operating conditions. The new method meets all the requirements and presents a marked improvement over published preliminary nacelle design methods.
Highlights
Response surface models (RSM) are interesting in the design of engineering systems, as they promise not just near-instant evaluation of complicated designs and to maximize the use that can be made of existing data
The distribution of nacelle drag prediction errors for different operating conditions was characterized by comparing predictions from the RSM and CFD for several example nacelle designs
The problem of nacelle drag prediction from RSM presents a unique combination of challenges which prevent the application of standardized methods
Summary
Response surface models (RSM) are interesting in the design of engineering systems, as they promise not just near-instant evaluation of complicated designs and to maximize the use that can be made of existing data For this reason, they have become frequently used in various engineering studies, both to augment and accelerate optimisation processes [1] and as substitutes for the higher-fidelity methods used to generate the training data, in preliminary design [2]. Preliminary design applications permit comparably low accuracy on absolute performance figures as long as the effects of design changes are modelled correctly This allows the use of methods with very selective accuracy, where data of low interest can be simplified to a large extent
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a callDisclaimer: All third-party content on this website/platform is and will remain the property of their respective owners and is provided on "as is" basis without any warranties, express or implied. Use of third-party content does not indicate any affiliation, sponsorship with or endorsement by them. Any references to third-party content is to identify the corresponding services and shall be considered fair use under The CopyrightLaw.
