云制造环境下资源建模及其匹配方法研究

发布时间：2018-06-24 11:28

本文选题：云制造 + 资源建模　；参考：《浙江工业大学》2014年博士论文

【摘要】：随着计算机网络技术的飞速发展,制造业的全球化、信息化日益突显,制造资源的共享度和利用率不断提高,网络化制造、制造网格等利用计算机网络技术的新型制造模式逐渐成为现代制造业的主要模式。近年来,随着云计算等“云理论”的发展,云制造的概念开始兴起。云制造是云计算在制造业的落地和延伸,它体现了“分散资源集中使用”和“集中资源分散服务”的思想,以“多对多”的服务模式,汇聚分布式资源进行集中管理,实现制造资源的高度共享和高效利用,提高制造企业的生产效益,为用户提供更高满意度和更环保的产品及服务。云制造是一种大规模网络化分布式制造,与网络化制造、制造网格等一脉相承,并具有三方面特征：云制造资源海量性、异构性、复杂性和粗粒度性；云制造的用户具有高度的参与性和多样性；云制造具有较强的自修复性。依据这三个特征,本文在国家自然科学基金(60970021)和浙江省科技计划项目(2007C21013)的支持下,着重研究了云制造环境下制造资源的定义、发布、匹配和选择问题,包括以下几方面： (1)针对云资源海量、复杂、异构的特征,提出了云制造体系结构、双层资源模型,以及基于资源云服务(RVCS:Resources Via Cloud Service)的资源封装、发布和发现机制。以WSRF体系架构为基础,提出了云制造体系结构,并详细分析了其中云端系统和云制造平台系统的逻辑层次和逻辑关系。在此基础上,根据云端和云制造平台不同的功能需求,提出了云端-云制造平台相分离的双层资源模型(BCMRM:Bilayer Cloud Manufacturing Resource Model):在云端建立面向资源基本属性的基础数据模型,在云制造平台建立面向资源服务属性的功能数据模型,并存在逻辑映射关系,实现了复杂、异构资源的统一数据模型。并且,根据云资源动态属性的功能特征和更新频率分为三层次两类别,即资源属性、服务属性和提供者属性三个层次,以及特征属性和状态属性两个类别。根据动态属性划分,引入具有实时侦听和属性甄别功能的连接器和云代理,进一步提出了基于RVCS的云资源封装、发布和发现机制,以便实现海量云资源数据的分布式存储和快速独立更新。通过对比实验,证明在大规模数据环境下,能够较好地提高资源发布和更新效率。 (2)针对云资源粗粒度性的特征,提出了一种基于多维可拓理论的云资源性能相似度计算方法。首先提出特征属性匹配的核心是云服务资源性能属性的匹配。通过物元模型描述云资源的性能属性和用户需求,并根据性能指标数分为一维、二维、三维和多维等四类性能模块,从而将云资源性能匹配问题转化为多维空间中点与多维体的可拓距计算问题,并进一步建立关联函数计算云资源性能模块的匹配度。同时,提出了自定义权重、结构权重和实例权重相结合权重确定方法,从而获得云服务资源性能属性综合相似度。通过实验,验证了该方法具有较好的最佳资源命中性,且降低了粗粒度云资源指标权重计算的复杂度。 (3)针对云制造用户高度参与性和多样性的特征,以Beth信任关系理论为基础,提出了一种基于用户自主预测评估和其他用户推荐评估的云资源QoS评估方法。首先提出了状态属性匹配的核心是云服务资源的QoS评估,并建立了一种基于预测评估和推荐评估的云资源QoS评估方法：对于预测评估,引入时效性经验因子和经验修正因子,优化了基于用户历史评价的预测模型；对于推荐评估,建立了基于历史经历的推荐用户粗选方法,和基于评价相似度与客观度的推荐用户群精选方法。最后通过变异系数,获得更为全面的QOS评估。通过对比实验,证明了预测评估能够充分反应历史评价且具有较好的灵敏性,满足了高度的用户参与性；用户筛选算法能够很好地过滤恶意评价和劣质评价,避免了用户多样性带来的干扰。 (4)针对云制造节点突发故障时替代资源选取问题,建立了基于三角模糊数互补判断矩阵的制造风险可拓评价模型,并进一步提出了云资源动态调整策略。替代资源的选取从制造风险和动态属性匹配度两个方面进行考量：首先在可拓物元模型的基础上建立云资源制造风险评价模型,通过三角模糊数互补判断矩阵计算获得风险指标权重,并给出了三种生产关系的组合风险计算方法；在动态属性匹配度计算中,采用变异系数和使用门限相结合的方法,解决了性能匹配度与QoS评估值的不可比问题和QOS评估的客观性问题。通过实验,验证了该方法能够较为全面地反映资源的制造风险和可替代性。
[Abstract]:With the rapid development of computer network technology, the globalization of manufacturing industry, the increasingly prominent information technology, the increasing sharing and utilization of manufacturing resources, network manufacturing, manufacturing grid and other new manufacturing modes using computer network technology have gradually become the main mode of modern manufacturing. In recent years, with cloud computing, "cloud theory" and so on. The concept of cloud manufacturing is beginning to rise. Cloud manufacturing is the ground and extension of cloud computing in the manufacturing industry. It embodies the idea of "centralized use of scattered resources" and "centralized resource decentralization service", with "multi to many" service patterns, centralized management of distributed resources, and high sharing and efficiency of manufacturing resources. Use, improve production efficiency of manufacturing enterprises, provide users with higher satisfaction and more environmentally friendly products and services.
Cloud manufacturing is a kind of large-scale networked distributed manufacturing, which is connected with networked manufacturing and manufacturing grid. It has three characteristics: cloud manufacturing resources are massive, heterogeneous, complex and coarse granularity; cloud manufacturing users have high participation and diversity; cloud manufacturing has strong self-repair and complex. Based on these three With the support of the National Natural Science Foundation (60970021) and the Zhejiang science and technology project (2007C21013), this paper focuses on the research on the definition, release, matching and selection of manufacturing resources in the cloud manufacturing environment, including the following aspects:
(1) in view of the mass, complex and heterogeneous features of cloud resources, the cloud manufacturing architecture, double resource model, and resource encapsulation, release and discovery mechanism based on RVCS:Resources Via Cloud Service are proposed. Based on the WSRF architecture, the cloud manufacturing architecture is proposed and the cloud system and cloud are analyzed in detail. On this basis, based on the different functional requirements of cloud and cloud manufacturing platforms, a double layer resource model (BCMRM:Bilayer Cloud Manufacturing Resource Model), which is separated from cloud manufacturing platform, is proposed. The basic data model for the basic resources of resources is established in the cloud, and the cloud system is used in the cloud system. The building platform establishes a functional data model of resource oriented service attributes, and has a logical mapping relationship. It realizes a unified data model of complex and heterogeneous resources. According to the functional features and update frequency of the dynamic properties of the cloud resources, it can be divided into three levels and two categories, namely, resource attributes, service attributes and provider attributes, and features, and features. Two categories of attribute and state property. Based on dynamic attribute partition, the connector and cloud agent with real-time listening and attribute discrimination are introduced. The mechanism of cloud resource encapsulation, release and discovery based on RVCS is further proposed in order to realize the distributed storage and fast independent update of mass cloud resource data. Large scale data environment can improve the efficiency of resource release and update.
(2) aiming at the coarse-grained characteristics of cloud resources, a method of computing the similarity of cloud resources based on multidimensional extension theory is proposed. Firstly, the core of feature attribute matching is the matching of performance attributes of cloud service resources. The performance attributes and user requirements of cloud resources are described by the matter element model, and the number of performance indexes is divided into one dimension. Four kinds of performance modules, such as two-dimensional, three-dimensional and multidimensional, are used to transform the problem of cloud resource performance matching into the extension distance calculation problem of multi-dimensional space middle point and multidimensional body, and to further establish the correlation function to calculate the matching degree of the cloud resource performance module. At the same time, the definition of custom weight, structure weight and weight of instance is put forward to determine the weight. The method is used to obtain the comprehensive similarity of performance attributes of cloud service resources. Through experiments, it is proved that the method has better optimal resource neutrality and reduces the complexity of the weight calculation of coarse grain cloud resource index.
(3) in view of the highly participatory and diversity characteristics of cloud manufacturing users, based on the Beth trust relationship theory, a cloud resource QoS evaluation method based on user independent prediction assessment and other user recommendation evaluation is proposed. The core of the state attribute matching is the QoS evaluation of cloud service resources and a prediction based on the prediction. The evaluation and evaluation method of cloud resource QoS assessment: for the prediction evaluation, the timeliness experience factor and the experience correction factor are introduced, and the prediction model based on the user history evaluation is optimized. The recommended user coarse selection method based on the history experience and the recommended user group based on the evaluation similarity and objective degree are established for the recommendation evaluation. Finally, a more comprehensive QOS evaluation is obtained through the coefficient of variation. Through a comparative experiment, it is proved that the prediction evaluation can fully respond to historical evaluation and have good sensitivity to meet the high user participation. The user screening algorithm can filter malicious evaluation and inferior evaluation well and avoid user diversity belt. The interference from it.
(4) in view of the alternative resource selection problem for cloud manufacturing nodes, a manufacturing risk assessment model based on triangular fuzzy number complementary judgement matrix is established, and the dynamic adjustment strategy of cloud resources is further proposed. The selection of alternative resources is considered from two aspects: manufacturing risk and dynamic attribute matching degree: first, extension On the basis of matter-element model, the risk evaluation model of cloud resource manufacturing is established. The weight of risk index is obtained by the triangular fuzzy number complementary judgement matrix, and the combined risk calculation method of three kinds of production relations is given. In the calculation of dynamic attribute matching, the performance matching is solved by combining the variation coefficient and the use threshold. The problem of matching degree and QoS evaluation value and the objectivity of QOS evaluation are discussed. Through experiments, it is proved that this method can reflect the manufacturing risk and substitutability of resources more comprehensively.
【学位授予单位】：浙江工业大学
【学位级别】：博士
【学位授予年份】：2014
【分类号】：TP393.09

【参考文献】