复杂产品设计中参数关联和等效简化方法及其应用

发布时间：2018-06-30 17:49

本文选题：关联强度 + 多源数据融合　；参考：《浙江大学》2014年博士论文

【摘要】：复杂产品的结构组成复杂、运行工况多变、设计变量数量众多、多性能目标相互耦合,使得产品的性能设计异常困难,难以获得产品整体性能最优的设计方案。模型复杂性和计算复杂性是复杂产品性能设计难点之一。本文在对复杂产品性能设计相关技术的研究现状进行总结和分析的基础上,针对设计变量与性能关联强度计算、设计模型解耦和设计模型等效简化等问题进行了深入研究,根据研究成果开发了大型空分装备性能设计系统,在杭氧8万等级空分装备设计中进行了应用验证。论文的主要研究内容包括：第一章概述了复杂产品性能设计的研究现状,综述了设计变量与性能间关联强度分析、设计模型解耦、模型等效简化、性能求解等关键技术,给出了本文拟解决的问题,介绍了本文的研究内容和组织结构。第二章提出了多源数据融合的设计变量与性能关联强度计算方法。使用响应面拟合法和灰色神经网络法实现了多源数据的融合分析及性能设计模型的构建。基于单因素敏感度信息熵的均值和标准差,将设计变量划分为强关联变量、弱关联变量和变关联变量。构建了变关联矩阵,在性能求解中实现了设计变量与性能间关联强度的精确计算及设计变量的自适应调节。第三章提出了设计变量与性能间多重耦合聚类分析与设计模型解耦技术。根据设计变量与性能的关联强度构建了包含设计变量与性能间耦合关联、变量间耦合关联和性能间耦合关联等多重耦合信息的综合耦合度矩阵。通过综合耦合度矩阵的降维分析和变量性能二元树的分支耦合度比较,实现了设计变量与性能的聚类分析。提出了聚合度的概念进行设计模型解耦效果评价,以聚合度为指标进行了设计模型解耦。第四章提出了相似组合模型的参数摄动等效简化技术。分析了相似组合模型的特点,根据同类变量的小扰动分析结果,进行了参数的奇异摄动分析。提出了相似组合模型等效简化的系数设置准则,包括组合模型划分准则、敏感子模型选取准则和修正系数求解准则。应用于精馏上塔开环响应和闭环响应下的产品纯度预测,保证了产品纯度预测结果的准确性。第五章提出了参数不确定性分析的模型等效简化技术。进行了各类参数不确定性的来源分析和定量表征,构建了性能不确定性度量函数。基于性能预测区间进行设计变量采样点选取,基于模型均方误差进行系统参数采样点选取,构建了性能不确定性驱动的等效简化模型。应用于换热器的传热性能求解,提高了换热器翅片结构设计结果的稳健性。第六章实现了大型空分装备性能设计系统的研制。将本文提出的理论方法应用于大型空分装备的性能设计,介绍了软件系统的体系结构,详细阐述了软件的各项功能。并将该软件应用于杭氧8万等级空分装备的设计,验证了研制软件的有效性。第七章对本文进行了总结,归纳了本文的主要研究成果和创新点,展望了今后的研究工作。
[Abstract]:Complex product has complex structure, variable operating conditions, large number of design variables and coupling of multi performance targets, which makes the product performance design difficult and difficult to obtain the best overall performance of the product. Model complexity and computational complexity are one of the difficulties in complex product performance design. On the basis of the summary and analysis of the research status of related technology, the problems of design model decoupling and design model equivalent simplification are studied. The performance design system of large air separation equipment is developed according to the research results, and the design of the 80 thousand grade air separation equipment is introduced. The main contents of this paper are as follows: in the first chapter, the research status of the performance design of complex products is summarized. The key technologies such as the analysis of the correlation strength between the design variables and the performance, the decoupling of the design model, the model equivalent simplification and the performance solution are summarized. The problems to be solved in this paper are given, and the contents of this paper are introduced and the contents of this paper are introduced. The second chapter puts forward the calculation method of the correlation strength between the design variables and the performance of multi source data fusion. Using the response surface fitting method and the grey neural network method, the fusion analysis of multi source data and the construction of the performance design model are realized. Based on the mean and standard deviation of the information entropy of single factor sensitivity, the design variable is divided into a strong pass. The variable correlation matrix is constructed. In the performance solution, the precise calculation of the correlation strength between the design variables and the performance and the adaptive adjustment of the design variables are realized in the performance solution. The third chapter puts forward the decoupling technique of the multiple coupling cluster analysis and the design model between the design variables and the performance. A comprehensive coupling degree matrix which includes the coupling correlation between the design variables and the performance, the coupling correlation between the variables and the coupling relationship between the properties is constructed. The cluster analysis of the design variables and performance is realized by the reduction of the dimension analysis of the integrated coupling matrix and the branch coupling of the two element tree with variable performance. The concept of degree of polymerization is put forward to evaluate the decoupling effect of the design model, and the design model is decoupled with the degree of polymerization. The fourth chapter puts forward the equivalent simplification technique of the parameter perturbation of the similar combination model. The characteristics of the similar combination model are analyzed. The singular perturbation analysis of the parameters is carried out according to the small perturbation analysis results of the same kind of variables. The equivalent simplified coefficient setting criterion of the similar combination model is proposed, including the combination model division criterion, the sensitive submodel selection criterion and the correction coefficient solution criterion, which is applied to the product purity prediction under the open loop response and closed loop response of the distillation tower, and the accuracy of the product purity prediction results is guaranteed. The fifth chapter puts forward the uncertainty of the parameters. The model equivalent simplification technology of qualitative analysis is used. The source analysis and quantitative characterization of various parameter uncertainties are carried out. The measurement function of performance uncertainty is constructed. Based on the performance prediction interval, the sampling points of the design variables are selected. Based on the mean square error of the model, the sampling points of the system parameters are selected, and the driving of performance uncertainty is constructed. The simplified model is applied to the heat transfer performance of the heat exchanger, and the stability of the design results of the fin structure of the heat exchanger is improved. The sixth chapter realizes the development of the performance design system of the large air separation equipment. The theory method proposed in this paper is applied to the performance design of the large air separation equipment, and the architecture of the software system is introduced and elaborated in detail. The software is applied to the design of the 80 thousand grade air separation equipment of Hangzhou oxygen. The validity of the software is verified. The seventh chapter summarizes the main research results and innovation points of this paper, and looks forward to the future research work.
【学位授予单位】：浙江大学
【学位级别】：博士
【学位授予年份】：2014
【分类号】：TQ116.11;TH122

【参考文献】