A method of sequential log-convex programming for engineering design

作     者：Karcher, Cody Haimes, Robert 

作者机构：MIT Dept Aeronaut & Astronaut 77 Massachusetts Ave Cambridge MA 02139 USA 

出 版 物：《OPTIMIZATION AND ENGINEERING》 (最优化与工程学)

年 卷 期：2023年第24卷第3期

页      面：1719-1745页

核心收录：

学科分类：12[管理学] 1201[管理学-管理科学与工程(可授管理学、工学学位)] 08[工学] 0701[理学-数学] 

基　　金：U.S. Air Force [FA8650-20-2-2002] 

主　　题：Geometric programming Log-convexity Non-linear programming Sequential quadratic programming Sequential convex programming 

摘      要：A method of Sequential Log-Convex Programming (SLCP) is constructed that exploits the log-convex structure present in many engineering design problems. The mathematical structure of Geometric Programming (GP) is combined with the ability of Sequential Quadratic Program (SQP) to accommodate a wide range of objective and constraint functions, resulting in a practical algorithm that can be adopted with little to no modification of existing design practices. Three test problems are considered to demonstrate the SLCP algorithm, comparing it with SQP and the modified Logspace Sequential Quadratic Programming (LSQP). In these cases, SLCP shows up to a 77% reduction in number of iterations compared to SQP, and an 11% reduction compared to LSQP. The airfoil analysis code XFOIL is integrated into one of the case studies to show how SLCP can be used to evolve the fidelity of design problems that have initially been modeled as GP compatible. Finally, a methodology for design based on GP and SLCP is briefly discussed.