多学科优化方法及其在水轮发电机方案设计中的应用
发布时间：2018-07-28 21:20
【摘要】：随着科技的发展，大型机械产品的设计逐渐从零件设计转向系统设计，涉及的学科领域由单一学科领域变成多个学科领域，对设计系统的优化从单目标优化发展到多目标优化。 大型水轮发电机作为一种大型机械，它的整体方案设计包含电气、机械、水力等多个学科领域，需要不同学科领域的专家互相协作，这其中包含了多个关键技术：为提高效率，需要对水轮发电机方案设计中各领域的复杂模型进行近似建模；各个方案设计领域模型的解耦问题；为实现整体方案优化，需要研究多目标优化问题。 研究从水轮发电机设计案例中通过近似建模来获取隐含的方案设计模型的方法。本文研究了两种近似建模方法，针对变量个数少，设计空间维数低，要求精度不高的系统采取多项式响应面近似建模方法；针对变量个数较多，设计空间维数高，要求精度高的系统采取支持向量机近似建模方法。并对水轮发电机的材料成本和电磁领域分别进行了近似建模，验证了两种方法的适用性。 对于近似建模后获得的设计模型中存在的耦合问题，本文研究了两种解耦优化方法，基于学科耦合变量最小方差的解耦方法和基于遗传算法的多学科解耦优化方法。对水轮发电机的定子重量耦合模型采用解耦方法进行了解耦优化。 针对设计方案中的多目标优化问题，提出了基于特征合成的多目标优化方法，采用多个模型对提出的算法进行了算法验证，并将算法运用到了水轮发电机的材料重量优化的设计中，获得较好的优化结果。 基于上述研究结果，开发了水轮发电机多学科优化系统，特别针对设计过程中的近似建模、解耦优化和多目标优化问题，开发了对应的功能模块，可用于水电设备方案设计的多学科优化。
[Abstract]:With the development of science and technology, the design of large-scale mechanical products is gradually changing from part design to system design, and the subject field is changed from a single discipline to a multi-disciplinary field. The optimization of design system develops from single-objective optimization to multi-objective optimization. As a kind of large machinery, the overall design of large hydrogenerator includes electrical, mechanical, hydraulic and other fields of study, which require experts from different disciplines to cooperate with each other, which includes several key technologies: to improve efficiency, It is necessary to approximate modeling the complex models in each domain of hydrogenerator scheme design; decouple the models in each scheme design domain; and study the multi-objective optimization problem in order to realize the overall scheme optimization. The method of obtaining implicit scheme design model by approximate modeling from hydrogenerator design case is studied. In this paper, two approximate modeling methods are studied. The polynomial response surface approximation modeling method is adopted for the system with less variables, low dimension of design space and low precision, and the number of variables is large, and the dimension of design space is high. Support vector machine (SVM) approximate modeling method is adopted in the system with high accuracy. The material cost and electromagnetic field of hydrogenerator are modeled approximately, and the applicability of the two methods is verified. For the coupling problem in the design model obtained by approximate modeling, this paper studies two decoupling optimization methods, one is based on the minimum variance of the subject coupling variable, the other is the multi-disciplinary decoupling optimization method based on genetic algorithm. Decoupling method is used to decouple the stator weight coupling model of hydrogenerator. Aiming at the multi-objective optimization problem in the design scheme, a multi-objective optimization method based on feature composition is proposed, and the algorithm is verified by using multiple models. The algorithm is applied to the design of material weight optimization of hydrogenerator, and better results are obtained. Based on the above research results, a multi-disciplinary optimization system for hydrogenerator is developed, especially for the approximate modeling, decoupling optimization and multi-objective optimization in the design process, and the corresponding functional modules are developed. It can be used for multi-disciplinary optimization of hydropower equipment scheme design.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TM312
本文编号：2151611
[Abstract]:With the development of science and technology, the design of large-scale mechanical products is gradually changing from part design to system design, and the subject field is changed from a single discipline to a multi-disciplinary field. The optimization of design system develops from single-objective optimization to multi-objective optimization. As a kind of large machinery, the overall design of large hydrogenerator includes electrical, mechanical, hydraulic and other fields of study, which require experts from different disciplines to cooperate with each other, which includes several key technologies: to improve efficiency, It is necessary to approximate modeling the complex models in each domain of hydrogenerator scheme design; decouple the models in each scheme design domain; and study the multi-objective optimization problem in order to realize the overall scheme optimization. The method of obtaining implicit scheme design model by approximate modeling from hydrogenerator design case is studied. In this paper, two approximate modeling methods are studied. The polynomial response surface approximation modeling method is adopted for the system with less variables, low dimension of design space and low precision, and the number of variables is large, and the dimension of design space is high. Support vector machine (SVM) approximate modeling method is adopted in the system with high accuracy. The material cost and electromagnetic field of hydrogenerator are modeled approximately, and the applicability of the two methods is verified. For the coupling problem in the design model obtained by approximate modeling, this paper studies two decoupling optimization methods, one is based on the minimum variance of the subject coupling variable, the other is the multi-disciplinary decoupling optimization method based on genetic algorithm. Decoupling method is used to decouple the stator weight coupling model of hydrogenerator. Aiming at the multi-objective optimization problem in the design scheme, a multi-objective optimization method based on feature composition is proposed, and the algorithm is verified by using multiple models. The algorithm is applied to the design of material weight optimization of hydrogenerator, and better results are obtained. Based on the above research results, a multi-disciplinary optimization system for hydrogenerator is developed, especially for the approximate modeling, decoupling optimization and multi-objective optimization in the design process, and the corresponding functional modules are developed. It can be used for multi-disciplinary optimization of hydropower equipment scheme design.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TM312
【参考文献】
相关期刊论文 前1条
1 谢延敏;于沪平;陈军;阮雪榆;;基于Kriging模型的可靠度计算[J];上海交通大学学报;2007年02期
，本文编号：2151611
本文链接：https://www.wllwen.com/kejilunwen/shuiwenshuili/2151611.html
