【作者】 陶飞；

【导师】 胡业发； Zhao Dongming；

【作者基本信息】 武汉理工大学 ， 机械设计及理论， 2008， 博士

【摘要】 当前针对制造网格(Manufacturing Grid, MGrid)的研究主要集中在MGrid的体系结构、关键技术、资源管理系统、某个行业中的应用原型平台、资源共享方式等方面,在抽象层面上进行了研究。而对构建MGrid资源共享的基础(MGrid资源服务数字化描述)和实现MGrid资源共享的关键技术(MGrid资源服务优化配置理论与技术)研究不足,没有具体的构建理论与方法。主要体现在以下几个方面:(1)MGrid系统中的用户如何向平台发布本企业空闲的资源或服务能力?(2)如何实现所发布的资源或服务的数字化描述?(3)成员之间如何相互发现资源或服务?(4)如何对所搜索到的资源服务的服务质量(Quality of Service, QoS)进行评估?(5)如何根据任务需求,实现资源服务优选与组合?(6)如何检测MGrid资源服务优化配置过程中出现的故障并消解?本文围绕以上MGrid资源服务优化配置问题,在深入研究MGrid内涵的基础上,提出了包括资源服务构造层、核心中间件层、用户中间件层、用户接口层、应用层在内的面向服务的MGrid五层体系架构。据此架构,设计了支持协同制造的MGrid运行平台。针对MGrid平台中资源服务优化配置需求,提出了MGrid资源服务优化配置整体解决方案并设计了实现系统。对实现该系统的资源服务数字化描述、匹配与搜索、QoS评估、优选与组合、故障检测与消解等基础理论与关键技术进行了深入研究。主要工作和研究成果如下:(1)讨论了MGrid资源的定义、给出了MGrid资源服务分类体系及相应的接口类。研究了MGrid本体构建方法。基于所提出的构建方法,参照OWL-S本体,建立了MGrid本体。在此基础上提出了基于MGrid本体和OWL-S的资源服务数字化描述方法。为MGrid资源服务优化配置提供了数据与信息支持。(2)针对MGrid资源服务数字化描述特点,将MGrid资源服务描述信息分为文字概念、句子、数值(包括数值区间和模糊数)、实体类(数据结构体)概念四类。分别提出了相应的资源服务描述信息相似度匹配算法。在此基础上设计了基于基本匹配、输入输出(I/O)匹配、QoS匹配、综合匹配的四步骤MGrid资源服务匹配与搜索机制,给出了具体实现算法并进行了实例应用验证。(3)为给用户和系统选择最佳资源服务,为资源服务优化配置提供量化参考依据,建立了MGrid资源服务QoS评估指标体系,给出了各QoS评价指标的评估模型和实现算法。重点研究了MGrid资源服务信任QoS(Trust-QoS)的评估模型、量化实现算法、信任值的实时动态更新算法,并进行了实例验证。(4)针对单一资源服务需求任务(SRSRTask)的资源服务优选问题和多资源服务需求任务(MRSRTask)的资源服务组合及优选问题,设计了资源服务QoS参数提取方法;提出了基于QoS的资源服务优选与组合方法,包括执行MRSRTask的组合资源服务基本构成模型及其QoS计算方法,MGrid资源服务组合路径生成方法和优选算法,并进行了实例应用验证。(5)提出了MGrid资源服务优化配置容错管理机制。定义了MGrid资源服务优化配置过程中可能产生的4类13种故障;设计了MGrid资源服务优化配置容错管理系统;给出了各类故障的具体检测方法及基于ECA(Event-Condition-Action)的故障消解策略;对所提出的容错管理机制进行了应用实验验证。结果表明能有效提高MGrid资源服务优化配置的可靠性。(6)开发了MGrid资源服务优化配置系统,对所提出的MGrid资源服务优化配置理论进行了验证;并以磁悬浮转子系统开发为应用验证对象,在开发的系统中实现了磁悬浮转子系统开发过程中所需的各类资源和服务的共享与优化配置。相关成果获得了三项国家软件著作登记版权。更多还原

【Abstract】 Existing works on manufacturing grid (MGrid) primarily concentrate on its concept, architecture, application prototype platform, application foreground, etc. But MGrid resource service digital description method----the foundation for MGrid resource sharing, and MGrid resource service optimal allocation----the basic theories and key technologies to realize MGrid resource sharing, are not effectively addressed till now. In order to realize MGrid resource service sharing and optimal allocation, the following problems must be addressed: (1) How to publish resource service into MGrid system and realize its digital description? (2) How to find out the qualified candidate resource service according to user’s requirements? (3) What is the quality of service (QoS) of the resource service and how to evaluate it? (4)How to realize resource service optimal-selection and composition according to different tasks’requirements? (5)How to provide failure-tolerance (including both failure detection and recovery) service during the process of MGrid resource service sharing and optimal allocation? The above problems are concluded as the problem of MGrid resource service optimal allocation. This paper emphasizes on above problems and aims to provide the basic theories and key technologies for MGrid resource service optimal-allocation (MGRSOA).In order to realize MGRSOA, after investigated the connotation of MGrid, a five-layered service-oriented MGrid architecture is put forwarded. Depended on the proposed MGrid architecture, an MGrid collaborative executing platform is proposed. Combined the requirements of resource service optimal allocation in MGrid collaborative executing platform, a synthetical MGrid resource service optimal allocation system (MGRSOAS) is proposed. The basic theories and key technologies to realize the MGRSOAS are studied in this paper, including MGrid resource service digital description, match and search, QoS modeling and evaluation, optimal-selection and composition, failure-tolerance (i.e., failure detection and recovery). The main contributions and works of this dissertation are as follows:(1) The definition of MGrid resource service is given out, associated with the corresponding resource service classes and their implementing API. A three-steps establishing method of MGrid ontology is proposed and the MGrid ontology is established. A new MGrid resource service digital description method based on MGrid ontology and OWL-S is put forward. A resource service digital describing document is given out which verifies the validity and utility of the proposed method.(2) MGrid resource service search and match mechanisms and its implementing algorithms are proposed. The describing information of resource service is classified into four categories: (a) word conception information, (b) sentence information, (c) number information (including number interval and fuzzy number), and (d) entity class (or data structure) information. The matching functions and algorithms of each kind of describing information are designed and proposed respectively. Based on the proposed describing information matching algorithms, the matching the basic information, including service name and service description, namely basic-matching; second, matching the inputs and outputs information of resource services, namely I/O-matching; third, matching the QoS information, namely QoS-matching; last, combining the above three matching results, and generating the general matching result, namely integrated-matching. The matching functions and algorithms of each phase are described in detail. The case study demonstrates the proposed methods and algorithms are valid and effective.(3) MGrid resource service QoS evaluation models and algorithms are proposed. In order to enhance the validity and success rate of resource service optimal-allocation in MGrid, provide high credible resource service abilities and results to user, the concept of resource service trust-QoS is presented, associated with the important roles it plays in MGRSOAS. A trust-QoS relationship model which is capable of capturing a comprehensive range of trust relationships exist in MGrid system is put forward. A two-layered resource service trust-QoS evaluation models are put forward, including intra-domain trust-QoS evaluation model and inter-domain trust-QoS evaluation model. The quantitative evaluation algorithms of trust-QoS degree value are proposed and detailedly described, as well as the real-time and dynamic updating algorithms of trust-QoS degree value.(4) In order to realize the optimal-selection of single resource service request task (SRSRTask), and composition and optimal-selection of multi-resource service request task (MRSRTask), the QoS information extracting methods are given out. QoS based MGrid resource service optimal-selection and composition methods and corresponding implementing algorithms are proposed, including four basis models for composite resource service and their QoS computing methods, generating methods of composite resource service executing path (CRSEP), MADM(Multiple Attribute Decision Making) and PSO (particle swarm optimization) based optimal-selection methods. The case study and simulation results indicate the proposed methods are valid and effective.(5) The potential failures that would generate during the process of MGRSOA are investigated. Thirteen failures are defined in detail, which are classified into four categories: (a) virtual link related failures, (b) resource service related failures, (c) task related failures, and (d) application related failures. The corresponding failure detection methods to each failure and ECA (Event-Condition-Action) based failure recovery mechanisms and methods are presented in detail. The implementation and simulation results indicate that the proposed approaches are sound on promoting the success rate and QoS of MGRSOAS(6) The MGrid resource service optimal allocation system is developed and implemented. Its feasibility and rationality is validated.更多还原