【作者】 张于轩；

【导师】 何桢；

【作者基本信息】 天津大学 ， 管理科学与工程， 2004， 硕士

【摘要】 本文研究的是多响应问题的稳健性设计优化方法。稳健性设计优化作为持续质量改进的一项重要支撑技术，在产品设计和开发过程中加以实施，能够有效的提高产品质量。传统的稳健性研究集中于单响应的情况，已经不能满足当今竞争环境的需要。如何对具有多个质量特性值的产品或过程进行稳健性设计优化，也就是多响应问题的稳健性设计优化，成为了亟待解决的问题。本文将该问题的解决划分为三个阶段——稳健性设计方法的选择、多响应优化技术的确定以及实现从一般多响应优化到多响应稳健性设计优化的转变。在此基础上，提出了分阶段解决问题的思路。首先，对两种稳健性设计方法——田口稳健性参数设计和基于双响应曲面模型的稳健性设计，进行了研究，分析了各自的优缺点，提出了选择的依据。接下来，本文研究了三种传统的多响应优化技术，并提出了一种崭新的优化方法——基于证据理论的多响应优化。与此同时也指出了各种方法优缺点及其适用范围。最后，本文以前面两大理论支柱为基础，提出了两个多响应稳健性设计优化模型——基于S/N的多响应稳健性优化模型以及DRSM与满意度函数结合的多响应优化模型。两种方法各有特点，文中对它们进行了比较研究。为了体现多响应稳健性设计优化与一般多响应有优化的区别，文章最后对一个V型铸造工艺的多响应实验设计进行了应用研究。分别运用本文提出的两种模型进行了多响应稳健性优化，并将其结果与一般多响应优化的结果进行比较和分析。得到了如下的结论：从总体上看，本文提出的两种模型比一般多响应优化方法获得的结果具有更好的稳健性。更多还原

【Abstract】 This thesis studies the methods of robust design optimization for multiresponse problems. As an important support tool in Continuous Quality Improvement (CQI), robust design optimization can be used in product design stage to improve product or process quality. Former studies were focused on single response problems. This doesn’t satisfy current customers’ needs because of the change of market circumstances. It is urgent to find solutions of multiresponse robust optimization.This thesis proposes a process to solve this problem by three stages: choosing a suitable robust design method, determining a right multiresponse optimization tool, and transforming general optimization to robust optimization. Firstly, the author studies two robust design methods: Taguchi Parameter Design and Dual Response Surface Model. This thesis analyses their weaknesses and advantages, and proposes some criteria of choice. Secondly, this thesis studies three traditional multiresponse optimization methods, and brings forward a new method: multiresponse optimization based on evidential reasoning theory. At the same time, weaknesses, advantages and applied scope of these motheds are discussed. At last, the author builds two models: multiresponse robust optimization model based on S/N ratios and multiresponse robust optimization model based on DRSM and desirability function, which ground on the former two theory foundations. Furthermore, comparative research between them is executed.Applied research of a multiresponse experiment in V-process is implemented to reflect the difference between general optimization and robust optimization for multiresponse. From this example, the following conclusion can be drawn: In general, the methods presented in this thesis are more robust than the methods of general multiresponse optimization.更多还原