Research On Cloud Manufacturing Intelligent Service Discovery And Selection Approaches For Mechanical Products

Cloud manufacturing(CMfg)is a new manufacturing mode based on the idea of"manufacturing as a service" and cloud computing.Service discovery and selection problem is one of its hot and difficult problems.Current technologies can realize service discovery and selection.However,it can not realize intelligent service discovery and selection.Therefore,studying intelligent,efficient service discovery and selection technologies,developing intelligent,efficient service discovery and selection system has important theoretical research value and application value.Life cycle-oriented service types were studied.Principle of service decomposition and composition,intelligent service discovery scheme,service selection indexs were proposed.Intelligent service discovery and selection architecture for mechanical products was presented,which can realize the first k services discovery and selection with multi-service indicators by taking mechanical product images and service types as input.To realize intelligent service discovery model in intelligent service discovery scheme,intelligent service discovery algorithms based on CNN were studied.Firstly,the generation approach of attention map(P_LBP)for mechanical products was proposed,which was based on LBP.And enching layer(P_Net)was designed.A new CNN structure:P_VggNet,comprising the following parts:P_Net and VggNet-16 was proposed.Secondly,we employed the gradient disappearance performance of ReLU’ positive part,and less neuronal death performance of tanh’negative part.Tangent-based rectified linear unit(ThLU)was proposed.Different light intensity,camera angles and covers were considered when collecting images by camera.Through the above measures to collect mechanical product images,downloaded images online and intercepted images from videos,10 types of products’ images were collected(the mechanical product dataset),which were respectively bearings,studs,gears,springs,rollers,shearers,scraper conveyors,belt conveyors,roadheaders and hydraulic supports.The mechanical product dataset was preprocessed and evaluated.Based on the P_VggNet and ThLU,we trained the mechanical product dataset and achieved the intelligent service discovery model(neural network model).The test accuracy and test loss of it was respectively 95.38%and 0.1839.To realize the first k service composition schemes’ selection,firstly,aiming at optimizing the first k minimum/maximum service time(service cost,manufacturing capability and comprehensive capability),the objective functions and mathematical models for designing service composition schemes,production service composition schemes,product service composition schemes and product schemes were established.Secondly,as noted from the mathematical models,the service selection problem of CMfg was combination optimization problem with k-shortest/longest paths.To solve this problem efficiently,a method of transforming service composition schemes’ directed graph into standard directed graph,subshort path theorem and path extension method were proposed.Based on Dijkstra algorithm,subshort path theorem and path extension method,the k-shortest paths algorithm for CMfg was proposed.The k-longest paths algorithm and the k-shortest/longest paths algorithm for CMfg were proposed.The first k service selection problems in CMfg have been solved efficiently.The main parts of CMfg intelligent service discovery and selection system(system demands,system flow chat,system architecture,cloud platform,dataset and iOS client)were designed.An intelligent service discovery and selection system was developed,which can efficiently realize intelligent and efficient service discovery and selection of the first k services with multi-service indicators by taking mechanical product images and service types as input.