复杂作用关系过程全局稳健性参数设计方法研究

发布时间：2018-05-25 07:53

本文选题：复杂作用关系过程 + 稳健性参数设计　；参考：《郑州大学》2014年硕士论文

【摘要】：稳健性参数设计是提高制造过程稳定性,减小产品质量波动的主要方法。随着技术水平的不断提高,众多行业出现了大量的作用关系复杂的制造过程,这些过程拥有多个极值点,而且其影响因子与响应输出之间存在高阶非线性关系。现有的稳健性参数设计方法由于实验设计方式、模型形式的限制,只能局部优化参数,不适用于复杂工艺过程的稳健性研究。如何在可行域全局范围内对质量特性的均值与方差进行有效建模,进而实现复杂作用过程全局意义上的稳健性优化,成为制造业质量改进的关键问题之一。本文从全局稳健性的角度探究复杂作用过程的参数优化问题,主要研究内容如下： (1)构建了基于支持向量机的单一响应模型。采用单一响应建模方式,研究可控因子变差和噪声因子波动规律对响应输出的影响,并结合因子的联合概率分布,以多重积分的形式呈现该影响规律；进而利用支持向量机优良的泛化性能,建立起复杂作用关系过程全局范围内响应均值与方差的连续变化模型。在拟合模型的基础上,通过取点将拟合值与实际值及双响应拟合值进行对比分析,证明了所提方法具有良好的拟合性能和预测性能。 (2)对建立的模型进行了稳健性参数寻优研究。提出利用可行域划分方法研究望大、望小问题的全局寻优过程；针对遗传算法提出使用均匀设计构造初始种群,针对序列二次规划方法提出使用可控因子组合角点和中心点作为初始点,然后使用改进后的算法对非线性问题进行最优值求解。 (3)利用所提建模优化方法进行仿真及实证研究。运用单一响应建模构建电感电阻串联电路和螺母生产过程的近似模型,使用所提优化策略对回归模型进行参数寻优,通过分析选取最佳因子组合方案,验证结果表明了所提建模方法和优化策略在复杂作用关系过程稳健性参数优化研究中的有效性和实用性。本文提出了适用于复杂作用关系过程的全局式稳健性参数优化思想及相应的实现方法和应用技术路线,研究成果拓展了稳健性参数设计的研究领域,对于制造业减少过程波动,提高产品质量具有显著的现实意义和较高的实用价值。
[Abstract]:Robustness parameter design is the main method to improve the stability of the manufacturing process and reduce the fluctuation of product quality. With the continuous improvement of the technical level, a large number of industries have a large number of complex manufacturing processes. These processes have multiple extreme points, and there is a high order nonlinear relationship between the impact factor and the response output. Some robust parameter design methods are limited by experimental design method and model form, only local optimization parameters can not be applied to the robustness study of complex process process. How to model the mean and variance of the quality characteristics effectively in the global scope of the feasible domain, and then realize the stability and robustness in the global significance of the complex process process. It is one of the key problems of quality improvement in manufacturing industry. This paper explores the parameter optimization of complex process from the perspective of global robustness. The main contents are as follows:
(1) a single response model based on support vector machine is constructed. Using a single response modeling method, the influence of the variation of controllable factor and the fluctuation law of noise factor on the response output is studied. Combined with the joint probability distribution of factors, the influence law is presented in the form of multiple integral, and then the excellent generalization performance of the support vector machine is built. A continuous variation model of the mean and variance in the global response process is established. On the basis of the fitting model, the fitting values are compared with the actual values and the two response values. It is proved that the proposed method has good fitting performance and predictive performance.
(2) the robust parameter optimization of the established model is studied. The global optimization process is studied by using the feasible domain division method. The initial population is constructed by using uniform design for genetic algorithm, and the control factor combination corner and the center point are used as the initial point for the sequence two programming method. Then, the improved algorithm is used to solve the nonlinear problem.
(3) using the proposed modeling optimization method to carry out simulation and empirical research. An approximate model of inductor resistance series circuit and nut production process is constructed by using single response modeling. The optimization strategy is used to optimize the parameters of the regression model, and the best factor combination scheme is selected. The results show the modeling method and the advantage of the proposed model. The effectiveness and practicability of chemical strategy in robust parameter optimization of complex interaction processes.
In this paper, the global robustness parameter optimization idea, the corresponding implementation method and the application technical route for the complex interaction process are proposed. The research results expand the research field of the robustness parameter design, and have significant practical significance and high practical value for the manufacturing industry to reduce the process fluctuation and improve the quality of the product.
【学位授予单位】：郑州大学
【学位级别】：硕士
【学位授予年份】：2014
【分类号】：TB472

【参考文献】