Doctoral Dissertations
Keywords and Phrases
airplane design; design space exploration; inverse design; machine learning; neural networks; surrogate models
Abstract
"A key challenge in airplane design has been the long-desired ability to quickly create an airplane design with specific performance properties. To tackle this problem, researchers have traditionally leveraged design space exploration methodologies to iteratively explore a design space via surrogate models. On the other hand, inverse design is a process wherein performance objectives are fixed while a suitable design for satisfying these requirements is calculated. In recent years, artificial neural networks (ANNs) have been explored as viable surrogate models because traditional surrogate models struggle with the large number of output to input mapping needed for inverse design problems. In this study, the feasibility of developing ANNs for calculating airplane configuration design parameters when provided with time-series, mission-informed performance data is investigated. Shallow ANNs were developed, trained, and tested using data pertaining to two design cases: tube-and-wing (TAW) and blended wing body (BWB) configurations. Performance for each configuration was obtained via a physics-based performance assessment in order to construct the training data sets. The effects of varying ANN architecture, along with the application of different data filtering schemes, on the predictive accuracy of the models have been examined. The results for the TAW scenario demonstrated that cascade-forward (CF) shallow ANNs exhibited better generalization across design sites, when trained with a limited amount of training data. Meanwhile, results for the BWB scenario showed that feed-forward (FF), shallow ANNs yielded significantly better predictive accuracy when a larger training dataset is used. The inference based on the results of the two design cases is that the optimal ANN architecture identified for each case was not due to characteristic differences in the airplane configurations; instead, it was a function of the size of the training dataset employed for each scenario. The results presented in this work are expected to assist future efforts aimed at developing surrogate models capable of handling time-series, mission-informed performance data for airplane inverse-design methodologies" -- Abstract, p. iii
Advisor(s)
Hosder, Serhat
Committee Member(s)
Han, Daoru Frank
Riggins, David W.
Dagli, Cihan H., 1949-
Reddy, Sudhakar
Department(s)
Mechanical and Aerospace Engineering
Degree Name
Ph. D. in Aerospace Engineering
Publisher
Missouri University of Science and Technology
Publication Date
Summer 2024
Pagination
xi, 104 pages
Note about bibliography
Includes_bibliographical_references_(pages 97-103)
Rights
©2024 Rohan Srinivas Sharma , All Rights Reserved
Document Type
Dissertation - Open Access
File Type
text
Language
English
Thesis Number
T 12410
Electronic OCLC #
1477970559
Recommended Citation
Sharma, Rohan Srinivas, "Mission-Driven Inverse-Design of Conventional and Unconventional Airplane Configurations using Machine Learning" (2024). Doctoral Dissertations. 3330.
https://scholarsmine.mst.edu/doctoral_dissertations/3330