Development Of Optimization Software SIPOPT And Its Applications In Engineering

The amazing progress of computer technology has propelled the steady expansion of engi-neering optimization implementations. The challenges from the optimization applications, such as high computing time consumption, hard multi-objective optimizations, and so on, have raised concerns over the complexities of real-world engineering optimization processes. Optimization software, which bridges the advantages of computer technology, edges of optimization method academic research, and practical optimization applications, plays a central role in approaches tackling these challenges. However, lots of the optimization software is hard to exploit the ad-vances of optimization research or harness the computer's potential. This dissertation analyses and designs multi-level extensible optimization software, named as SIPOPT (Software Integra-tion Platform for Optimization), as well as investigates approaches that enhancing effectiveness of engineering optimization proceeding in several disciplines collaborating with CAE software, parallel computing technology, high throughput computing environment, and other methods.This dissertation is composed of three parts:the introduction, the design and implementa-tion of SIPOPT, and engineering optimization applications with SIPOPT.The introduction, chapter 1, presents background and target of this dissertation. Firstly, developments of optimization algorithms and optimization software are reviewed. Then, a mea-sure of engineering optimization application computational complexity is examined; methods used to enhance optimization application computational efficiency are summarized. Finally, the research route of this dissertation is proposed.In the second part of the dissertation, SIPOPT is developed based on a software architec-ture with multi-level extensible scheme. SIPOPT integrates optimization algorithms modules, DOE (Design of Experiment) module, surrogate model module, GUI (Graphic User Interface), and other modules on the basis of JavaScript language environment. With its generalizing, inte-grating, and extensible features, SIPOPT can be installed cross software operation systems, and supports optimization approaches both in stand-alone and Internet environment.The third part of this dissertation aims to demonstrate SIPOPT's capabilities of integrating CAE software, solving complex optimization problems, and improving engineering optimiza-tion application computational efficient. Thus, engineering optimization problems of structure design, dynamic and control, build energy analysis are selected, and approaches for tackling these problems were proposed respectively.Chapter 4 consists of two structural mechanics optimization applications. Garteur airplane structural dynamical model updating optimization application is tackled by SIPOPT integrat-ing ANSYS, and wind turbine blade structure optimization application by SIPOPT integrating ABAQUS. The two approaches demonstrate SIPOPT's capability of solving complex simulation based practical optimization problems.In chapter 5, a set of spacecraft orbit control optimization problems are tackled from dif-ferent directions with SIPOPT:a multi-peak single objective optimization model is solved by hybrid-optimization algorithms, a three-objective optimization model by divide and conquer s-trategy, a computational time sensitive optimization application by surrogate model method. These research works demonstrate SIPOPT's capability of solving complex practical optimiza-tion problems and extensibility.Regarding building energy optimization problems, a web-based distributed high-through computation framework is presented in chapter 6. Based on distributed high throughout environ-ment of 16 computing nodes and 48 CPU cores, parallel NSGA-II algorithm, data management module, and task management modules are constructed. A case study of a KUBIK building en-ergy consumption optimization shows the speedup of the parallel NSGA-II reached up to 39.3. The research demonstrates SIPOPT's capability of increasing complex practical optimization problems computational efficient.SIPOPT is one trial of independent research and development of general purpose optimiza-tion software in China. The research presents software architecture, development scheme, and object-oriented design for optimization software; constructs a multi-level extensible optimiza-tion platform; lays the basis for a optimization software ecosystem; and provides a reference for the design of related simulation software. The case studies demonstrate SIPOPT is capable of providing similar functionalities of international optimization software, additionally, SIPOPT shows advantages in extensibility, openness, and customization.