【作者】 常建涛；

【导师】 仇原鹰；

【作者基本信息】 西安电子科技大学 ， 机械电子工程， 2014， 博士

【摘要】 随着制造业信息化的推进，各种制造信息系统在离散制造企业中得到了广泛的应用，技术也不断成熟。这些异构制造系统的成功实施也为企业的业务需求做出了重要贡献。然而，由于各种制造信息系统在设计和建模的时候更多的只是关注信息系统自身功能，功能的设计和建模偏向企业的业务需求，而对于模型之间的集成考虑较少，致使相互之间的模型理解存在障碍，从而给系统集成带来了很大难度，形成了一个个“信息孤岛”。另外，在生产管理中的生产计划编制过程中，由于产量和产品总完工时间设定不准确而引起的生产计划编制不合理问题也日益突显，因此在解决“信息孤岛”问题的基础上，从多种异构制造系统集成框架模型中获取影响产量和产品总完工时间预测的各项因素，实现产量与总完工时间的准确预测是非常必要的。由于许多能够反映生产管理决策的关键信息不仅仅存在于某一个制造系统中，且是来源于多个制造系统的集成框架模型，从集成框架中获取的数据还存在抽象、繁琐、不易理解等特点，需要对这些数据进行可视化处理，因此，“信息孤岛”的存在也使得企业级领导无法看到企业信息化的综合成果。针对上述问题，本文的研究主要分为以下几个部分：论文对离散制造企业生产管理的原理流程进行了分析，在此基础上，结合国内某航空制造企业的特点，对原理流程的每个节点进行了优化，得到了适合该航空制造企业的生产管理流程优化实例，并给出了该流程实例的优点。为后续的集成框架建模、产量预测、总完工时间预测、生产管理信息可视化奠定了基础。针对各种异构制造系统单一运行、不能柔性集成问题，在生产管理流程分析和优化的基础上，研究了异构制造系统集成框架建模技术。提出了一套基于语义网关的异构制造系统集成框架的建模体系，分别运用UML(Unified ModelingLanguage)类图集成建模方法、IDEF0(ICAM DEFinition method)功能集成建模方法、GRAI(Graph with Results and Activities Interrelated)功能建模方法建立了异构制造系统的集成框架模型，研究了IDEF0功能模型与GRAI网格映射方法。在异构制造系统集成框架下研究了零部件层面、产品装配层面的信息模型表达方式，提出了一种生产管理集成框架下的统一信息模型，异构制造系统综合集成较为刚性的问题得到了解决，提高了集成度，方便了各类信息的挖掘和获取。针对生产计划编制过程中的产量设定不合理问题，在生产管理集成框架模型的基础上，提出了一种基于动态多元线性回归模型的生产计划产量预测方法。多元线性回归模型在集成框架模型基础上提取信息，即将生产管理过程中的制造资源、人力投入、制造工艺、产品报废等生产全周期的影响因素作为建模变量，将影响产量的相关因素尽可能多地包含在预测模型中，从而使得模型的预测结果与实际产量更为接近。运用“后推法”将建立的初始多元线性回归模型进行显著性辨别，剔除了对产量影响不显著的变量，建立了产量预测的改进多元线性回归模型，在此基础上进一步建立了动态改进多元线性回归模型。将该方法运用到某航空制造企业的产量预测中，通过模型预测产量和实际产量的对比，证明了动态改进多元线性回归模型在产量预测方面的实用性。针对产品总完工时间设定不合理问题，在生产管理集成框架模型的基础上，提出了一种基于AIGA-DBP(Adaptive Immune Genetic Algorithm-Dynamic BackPropagation)模型的产品总完工时间预测方法。该方法将生产管理集成框架模型中的完工时间历史数据和影响其预测的因素进行了全面分析，在此基础上建立了基于BP神经网络的总完工时间预测模型。接下来对BP神经网络模型的权值和阈值进行了动态改进，进一步运用AIGA算法对动态BP神经网络模型进行了优化，得到了AIGA-DBP预测模型。最后将该预测模型应用于某航空制造企业的总完工时间预测，预测结果与实际完工时间的对比表明，此方法能够准确地对完工时间进行预测。针对企业决策者难以从众多的异构制造系统中获取决策支持信息及信息的表现形式较为抽象、复杂、不易理解等问题，建立了一套基于生产管理集成框架模型的决策信息获取和可视化体系结构。该体系结构首先将生产管理过程中所涉及的异构系统建立数据集市，并根据企业需求开发了一套信息获取API(ApplicationProgramming Interface)。接下来研究了信息可视化技术中的抽象数据可视化、树状图可视化及动态流程图可视化，最后通过直方图、甘特图、折线图和动态流程图实现了企业领导层关心的生产管理决策信息的可视化。在某航空制造企业的应用表明，信息的传递效率和可视化程度得到了提高，并为企业领导的生产管理决策提供了有效的支持。更多还原

【Abstract】 With the development of manufacturing informatization and informationtechnology, a variety of manufacturing information systems has been widely applied inintermittent manufacturing enterprises. The successful implementation of theseheterogeneous manufacturing systems has made important contributions to businessneeds for discrete manufacturing enterprises. However, the integration of heterogeneousmanufacturing systems is a difficult problem, because in the design and modelingprocess, every manufacturing system only considers its own function and the businessneeds, so the integration function is ignored. For this reason, the information islandsappear. On the other hand, the irrational problems of production planning in productionmanagement also becomes increasingly serious, the reason is that the determination ofthe output and the makespan is inaccurate, so it is necessary to realize the prediction ofoutput and makespan accurately based on the solution of information islands problem.Due to the existence of information islands, enterprise-level managers can not feel thecomprehensive achievement of the informatization, the reason is that the decisioninformation of production management comes from the integration framework ofmultiple manufacturing systems but not a manufacturing system, and the data from theintegration framework is abstract, complicated and difficult to understand, so the datavisualization operation is needed. In order to overcome these problems above, theresearch contents of this paper include the following sections:The principle workflow of production management for discrete manufacturingenterprises is analyzed, and every node of the principle workflow is optimized based onthe requirement of an aviation manufacturing enterprise. The application example ofworkflow in this aviation manufacturing enterprise is obtained based on the principleworkflow optimization. The optimiazed production management workfolw can supplysupport for the research of integrated framework modeling, output prediction, makespanprediction and information visualization of production management.To overcome the flexible integration issue of heterogeneous manufacturingsystems, the integrated framework modeling technology is studied based on theproduction management workflow analysis. An integration framework modeling systemof heterogeneous manufacturing systems is presented. The integration framework modelof heterogeneous manufacturing systems is established via the method of IDEF0Functional modeling, GRAI Functional modeling, and the network mapping methodbetween IDEF0Functional modeling and GRAI functional modeling is studied. The information model of part level and assembly level are studied, and then a unifiedinformation model of production management is proposed. The flexible integrationproblem and information mining problem are solved by unified information model.To overcome the unreasonable setting of production output, a new approach topredict the output of production planning is presented. In this new approach, an initialmultiple linear regression model is established based on the related factors which canaffect the output of production planning, these factors are extracted from the integrationframework of PDM, ERP,CAPP and MES. Then the low significance factors areremoved by backstepping method, and a dynamic-improved multiple linear regressionmodel for output prediction in the next season is obtained. This new approach proposedtakes the factors in production lifecycle into account so that prediction results can bemore accurately. Finally, we demonstrate the performance of our model by comparingthe prediction results with actual output of production planning.A novel AIGA-DBP(Adaptive Immune Genetic Algorithm-Dynamic BackPropagation) model is developed for solving the problem of maximum completion timeprediction (makespan prediction). By analyzing the history data of makespan and itsrelated factors on enterprise level, the prediction model based on BP neural network isestablished, then the weight values and threshold values of the BP neural network modelimproved dynamically, and the dynamic BP neural network model is further optimizedvia AIGA, so the AIGA-DBP model is obtained. The proposed AIGA-DBP algorithm isapplied to an aviation enterprise’s makespan prediction. Computational experimentssuggest that the algorithm yields accurate results and minimum error. The usefulness ofaccurate makespan prediction on enterprise level as a tool for improving productionefficiency is highlighted.Due to the difficulty of decision support information access and the problem ofinformation’s abstract, complicated and difficult to understand, decision informationvisualization architecture is presented based on the integration framework of productionmanagement. In this architecture, the data mart of heterogeneous manufacturing systemsin production management is established, then the API of information access isdeveloped. The information visualization technologies of Abstract data Visualization,Tree visualization and dynamic workflow visualization are studied. Based on thevisualization technologies, the histogram, Gantt chart, line chart and dynamic flowchartof decision information in production management are realized. The application case inan aviation manufacturing enterprise shows suggests that this architecture can improvethe efficiency of information transmission and the degree of information visualization, and it can provide support for the decisions in production management.更多还原