【作者】 曾辉；

【导师】 苏海涛；

【作者基本信息】 南昌大学 ， 工业工程， 2012， 硕士

【摘要】 在实际的产品与工艺过程的优化设计中,往往需要考虑多个质量特性,这就是多响应的问题,多响应问题通常不存在一组特定的输入变量使得所有响应变量同时达到最优,多响应优化设计能够有效地提高产品的质量,并产生了巨大的经济效益,因此多响应优化设计在持续性质量改进活动中显示出越来越重要的地位和作用。本文以SX公司产品注塑过程的多响应问题为研究对象,通过对该公司产品注塑过程多响应问题的现状分析发现,该公司分析人员采取的是传统的多响应曲面方法,即利用最小二乘法(OLS)拟合响应变量与控制变量的模型,然后基于满意度函数确定最佳因子水平组合。同时通过查阅国内外文献发现,最小二乘法在因子个数较多时难以准确的拟合模型,且在优化过程中需要考虑多响应之间的相关性问题。鉴于以上分析,根据注塑过程中因子个数的情况,将注塑过程分为复杂注塑过程和简单注塑过程,由于两类注塑过程在多响应优化方法上存在差异,本文提出了两类注塑过程的多响应优化的思路：(1)基于ANN(人工神经网络)的复杂注塑过程多响应优化。计算各响应变量满意度值,利用ANN拟合响应变量的满意度值和控制变量的模型,并进行预测；采用PCA(主成分分析)方法将一系列相关的变量转化成不相关的变量,得到主成分序列；在确定主成分的优化方向后,运用TOPSIS方法确定最佳的因子水平组合。(2)基于SUR(似不相关回归)的简单注塑过程多响应优化。由于SUR方法在较好拟合模型的同时又能解决响应变量之间的相关性问题,因此,采用SUR方法拟合模型,确定各个变量的满意度函数和总体满意度函数,在此基础上,获得最佳因子水平组合。最后,将上述两种思路得到的最佳因子水平组合进行试验验证,并对优化前后的结果进行理论和实证上的对比分析。本文的第五章提出了一些促进多响应优化的管理建议。更多还原

【Abstract】 In the design of the actual product and process optimization, often need to consider multiple quality characteristics, which is multi-response problem, this problem usually does not exist a specific set of input variables to make all response variables at the same time to achieve optimal, multi-response optimization design can effectively improve the quality of the products and generate huge economic benefits, so the multi-response optimization design show an increasingly important position and role in the continuous quality improvement activities.This paper takes multi-response problems of SX company’s product molding process as the research object, through the current situation analysis of multi-response problem to find, analysis personnel taken the traditional multiple response optimization method, namely use least squares method (OLS) to fit the model of response variables and the control variables, and then based on the satisfactory degree function to determine the optimal factor level combination. At the same time, via referring to domestic and foreign literatures found, least square method cannot accurately fit the model in the case of more factors, and in the optimal process needs to consider the correlation between responses. In view of the above analysis, in this paper, according to the factor number in the injection molding process, the injection molding process is divided into complex injection molding process and simple injection molding process, at the same time, put forward two kinds of multiple response optimization thought.(1) Multiple response optimizations for Complex injection molding process based on ANN. Calculate satisfaction value of each response variables, use the ANN to fit the model of response variables satisfaction value and the control variable, and conduct to predict; then use the PCA method transfer a series of related variables into irrelevant variables, and get the principal component sequence; after determining principal components optimization direction, use TOPSIS method to determine the optimal factor level combination. (2) Multiple response optimizations for simple injection molding process based on SUR. As the SUR method can either accurately fit model, or solve the correlation between the responses, therefore, use SUR method to fit model, then determine the satisfaction function of each variables and overall satisfaction function, on the basis of these, obtain optimal factor level combination.Finally, conduct test validation for the best combination of factor levels, and compare the response’results on the theory and practice, and then find the response values have been optimized. The fifth chapter proposes management measures and recommendations to promote multi-response optimization.更多还原