Study of a global design space exploration method for aerospace vehicles

Abstract

In the early stages of the design process of aerospace vehicles, the search for optimal configurations encompasses a broad range of possibilities, and the use of local optimization tools may risk missing the best designs. Therefore, global optimization methods are attractive for the early design stage. Unfortunately, global design optimization usually requires the evaluation of a very large number of designs, a formidable computational challenge. The present work uses a highly effective Lipschitzian global optimization algorithm in the multidisciplinary optimization of a High Speed Civil Transport (HSCT). The use of massively parallel computers to handle this computational challenge is also demonstrated. A variety of load balancing methods are evaluated to assess the utilization of the computer nodes.

Review and Computational Study on Practicality of Derivative-Free DIRECT-Type Methods

“…The authors adopted a straightforward approach to parallel function evaluations for expensive objective functions in the original DIRECT. In contrast, other practical applications employed enhanced DIRECT algorithms, which proved to be more computationally demanding (Baker et al, 2001;Dapšys et al, 2023;He et al, 2004). Two studies (Baker et al, 2001;Dapšys et al, 2023) utilized the "Aggressive" version of the DIRECT algorithm, relaxing the selection of POH criteria and subdividing hyperrectangles in each iteration of every diameter, leading to improved parallel efficiency.…”

“…In contrast, other practical applications employed enhanced DIRECT algorithms, which proved to be more computationally demanding (Baker et al, 2001;Dapšys et al, 2023;He et al, 2004). Two studies (Baker et al, 2001;Dapšys et al, 2023) utilized the "Aggressive" version of the DIRECT algorithm, relaxing the selection of POH criteria and subdividing hyperrectangles in each iteration of every diameter, leading to improved parallel efficiency. Furthermore, He et al (2004) proposed a hierarchical parallel scheme for DIRECT, where the initial domain was decomposed into smaller parts and each sub-domain was optimized independently using a master-slave scheme.…”

Experimental Study of Excessive Local Refinement Reduction Techniques for Global Optimization DIRECT-Type Algorithms

“…In [33], the authors introduced the aggressive selection of POH. This strategy selects at least one hyper-rectangle from each group of different measures (δ i k ) with the lowest function value.…”

Untitled

“…Preliminary experience of applying DI-RECT to the route-finding problem in [14] shows that it minimizes (2.2) more efficiently than than the methods in [1] and [5] which employ random searching. The use of DIRECT in aerospace design problems (with quite large numbers of variables) is reported in [2], [3], [4], [6], [7], [12] and [13]. In [7], in particular, it is noted that it does better than a trajectory-following approach on functions with deep and widely spaced minima.…”