-
摘要随着产品复杂程度以及顾客需求层次的变化，在实际的生产或者工艺过程中，往往需要考虑多个响应，因此多响应的稳健参数设计在质量改进活动中显示出越来越重要的地位和作用。同时，在实际的生产过程中，顾客一般会给定产品的规格限，因此为了提高生产的经济性，本文着重分析了考虑产品规格限的多响应稳健参数设计问题。 在产品的生产过程中，存在两种成本，一是产品因为不满足顾客的要求（位于顾客给定的规格限以外）造成的拒绝成本，例如，低于最低规格限造成的返工成本和超出最高规格限导致的报废成本。另一种是产品（满足顾客要求的合格品）输出特性的波动给用户和社会造成的损失成本。Taguchi 博士认为产品输出特性愈远离其目标值，造成的损失就愈大。同时，合格品所造成的质量损失成本又可以根据其具体的波动来源分为偏差成本、稳健性成本和预测质量成本三个部分。 本文以工艺（产品）设计阶段过程均值的优化问题为研究对象，在 Taguchi 单响应的二次质量损失函数的基础上，结合 Boylan的多响应质量损失函数方法和 Ko 的多响应质量损失函数方法，综合运用稳健参数设计、响应曲面法、遗传算法以及实证分析等技术手段，提出了一种既考虑顾客给定的规格限，又具体分析规格限内成本源头的新型优化模型。系统地研究了考虑规格限的多响应稳健参数设计问题，很好地解决了低稳健性和估计参数波动带来的优化模型有效性降低的问题。本文的主要研究内容如下： （1） 在加入规格限的同时分别刻画出规格限内的成本来源：本文综合运用 Boylan 的多响应质量损失函数方法和 Ko 的多响应质量损失函数方法将规格限外的返工成本、报废成本以及规格限内的偏差成本、稳健性成本、预测质量成本构建到一个优化函数中，使优化结果既考虑了顾客给定的规格限，增加了方法的经济性，同时又将规格限内的偏差成本、稳健性成本和预测质量成本分别刻画出来，有利于企业更好地诊断成本来源，有目标的降低生产成本。 （2） 同时考虑了稳健性和预测质量：在现存的经济性稳健参数设计研究中，一般只考虑响应值波动所造成的损失。本文结合 Ko 的多响应损失函数，将预测响应值波动所造成的损失也纳入到经济性的优化模型中，提高了优化模型的有效性。49166
- 上一篇：后组织公民行为的测量
- 下一篇：大学生消费心理特征的网络营销策略研究+SWOT分析
毕业论文关键词：多响应 经济性的稳健参数设计 响应曲面设计 质量损失函数
Title Economic Robust Parameter Design for Multi-response
Abstract As the change of the complexity degree of products and customer needs, in the actual production and technology process, more than one response need to be considered. So the robust parameter design for multi-response plays a more and more important role in the quality improvement activities. At the same time, in the actual production process, customers generally preset the specification limits. In order to improve the economics of the study, the paper investigates the robust parameter design problem from an economic point of view. In the production process, there are two kinds of costs, such as the rework cost and scrap cost caused by the rejection from customers because some products fail to meet the requirements of customers. In addition, some products meet the requirements of customers, but the variation of its output characteristics may cause loss to the user and society. Taguchi doctor considered that its output characteristics are farther away from the target, it causes more loss. At the same time, the cost of the products that meet the specifications can be pided into bias cost, robustness cost and quality of predictions cost according to their specific variation source. In the paper, taking the determination of the optimal process mean and response optimization design as the subjects of the research, based on the quadratic quality loss function of Taguchi, combing with the multi-response loss function of Boylan and the multi-response loss function of Ko, by means of Robust Design, Response Surface Methodology, Genetic Algorithm and empirical research. A new optimization method considering both the specification limits and detailed analysis of cost is proposed in this paper. This paper systematically studies the problem of parameter design for multi-response that considers specification limits and solves the problems of low effectiveness due to low robustness and low quality of predictions. Some main results in this paper are summarized as follows. (1) Analyze the source of the cost within the specification limits at the basis of considering the specification limits: in the paper, we build a function which concludes both rework cost and scrap cost outside the specification limits and bias cost, robustness cost and quality of predictions cost within the specification limits by combing with the multi-response loss function of Boylan and the multi-response loss function of Ko. The method considers the specification limits, which increase the economy of the method. In addition, the method pides the cost within the specification limits into bias cost, robustness cost and quality of predictions cost, which can help enterprises diagnose cost source and reduce the production cost purposefully. (2) Consider robustness and quality of prediction simultaneously: in the existing economic robust parameter design, there is only considering the loss caused by the response value fluctuation. In the paper, the method combines with the multi-response loss function of Ko, which also considers the loss caused by the predicted response value fluctuation, which improves the effectiveness of the optimization model.