Research On Manufacturing Cloud Service Discovery In Cloud Manufacturing

Cloud Manufacturing is a new service-oriented networked manufacturing modewhich uses Cloud Computing for reference. It has become the hot spot in the field ofadvanced manufacturing. Manufacturing Cloud Service(MfgCS) is the virtualization ofphysical manufacturing resources. Since the Manufacturing Cloud ServiceComposition(MfgCSC) can allocate resources effectively and increase the utilizationratio of resources, it is significant to fulfill manufacturing tasks. MfgCSC can bedivided into three stages: task planning and decomposition, candidate service discovery,service composition and optimization. Because the candidate service set which isobtained via service discovery is the only one input for the following steps, theperformance of the whole MfgCSC obviously depends on service discovery. So theresearch of manufacturing cloud service discovery(MfgCSD) is important.Recently, the researches of Cloud Manufacturing mainly focus on conceptdefinition or architecture construction. The existing researches of service discovery aremainly specific to computing resources and web services. But manufacturing tasks aremore critical than computing ones, and manufacturing resources are more complicatedthan traditional computing resources. It means that the existing service discoveryresearch achievements cannot be applied to Cloud Manufacturing directly. So using webservice discovery framework for reference, the thesis studies three core parts ofMfgCSD including manufacturing cloud service description, service match and serviceselection.In terms of service description, we aim to solve the problem that the existingservice description methods could not provide comprehensive information of MfgCS.Firstly, a manufacturing resource classification model is established and manufacturingcloud service ontology is constructed by extending the OWL-S service top ontology. Onthe basis of the classification model and service ontology, a new manufacturing cloudservice description method is proposed. It focuses on both static and dynamic attributes,both functional and non-functional Qos attributes. The method provides integratedinformation for the following matching steps.Since the description parameters of MfgCS are more diversiform than traditionalweb service, the existing service matching algorithms are no longer suitable. The thesisinnovatively classifies service describing information into three categories which are word concept, sentence and number. The similarity matching algorithms(SMAs) of eachtype are separately researched. On the basis of SMAs, a multi-level MfgCS matchingmodel is proposed. This model contains four levels including basic match, functionalmatch, nonfunctional Qos match and integrated match. According to different types ofdescribing information, the service matching algorithm of each level is realized byadopting different SMAs. And each level adopts threshold to control quantity ofsuccessfully matched services. Moreover, to improving the efficiency of service match,a process of service discovery is designed. In the process, the pre-filtering is executedbefore service match. The rules of pre-filtering have fully considered classification ofmanufacturing resources and the features of business cooperation.In terms of service selection, the thesis firstly refines the Qos evaluation attributesof MfgCS and then establishes a Qos evaluation model. Considering that the historicalQos evaluation from service requestors could directly reflects their Qos demands forservices, but the evaluation are usually fuzzy, a fuzzy comprehensive evaluation methodbased on historical Qos evaluation for service selection is proposed.Finally, combined with the above researches, a simulation test environment isdeveloped. The comparison between proposed service matching algorithm andtraditional key words matching algorithm shows that the proposed service matchingalgorithm gets better recall and precision. The influence towards discovery efficiency byadopting different organization of service discovery process shows that the proposedservice discovery process can highly improve the discovery efficiency.