Research On The Key Technologies Of Workshop Planning And Scheduling In Virtual Cellular Shop-floor

The trends of increasing severe market competition and decreasing product life cycles in theglobal manufacturing era point at the need to develop flexible reconfigurable manufacturing systems,which should adapt themselves to single work piece, small-batch, diversified and make to ordermanufacturing environment. As the application of internet of things and manufacturing executionsystem in manufacturing field, the shop-floor has become more and more transparency, digital andviewable.Thorough summary and analysis of advanced manufacturing system mode, a new shop-floormanufacturing mode based on virtual manufacturing cell, internet of things and manufacturingexecution system, named virtual cellular shop-floor manufacturing mode, was developed. Thisdissertation focuses the production scheduling and control problem of shop-floor in this kind ofmanufacturing mode. An extensive study on the supporting technologies, reconfigurationmethodology of virtual manufacturing cell, production scheduling and control algorithm was carriedout. The main contents and achievements of the dissertations are as follows:1. In order to make the manufacturing system become transparency, digital, viewable, flexibleand reconfigurable, a new shop-floor manufacturing mode based on virtual manufacturing cell,internet of things and manufacturing execution system (MES), named virtual cellular shop-floormanufacturing mode, was developed. In order to realize the internet of things and automatic control inshop-floor, the feasibility of shop-floor internet of things was discussed and solving methods for a fewrelated problems were put forward. First, a shop-floor internet of things construction method based oncomputing grid was discussed, a MES construction method based on grid service was proposedaccording to characteristics of internet of things, and the architecture of MES system based on gridservice was constructed. And then, a method to realize real time monitoring of shop-floor and productorder with internet of things was proposed. With this method, machine tools was encapsulated as gridresources with WSRF at first, and shop-floor can be real time monitored through inquiring thedynamic information or notifications of grid resources; then the heterogeneous database system wasencapsulated as uniform grid service with OGSA-DAI and WSRF, and the information of orderproduction schedule can be integrated access easily. With the above method, not only the real timemonitoring of the shop-floor can be easily realized for managers of the factory, but also completiontime of orders can be accurately predicted for the customers.2. To adapt to characteristics of single work piece, small-batch, diversified and make to ordermanufacturing environment, a methodology which could be used to reconfigure VirtualManufacturing Cell for multiple product orders with different due dates was proposed. The methodology is divided into two phases: product clustering and virtual cell reconfiguration. In thephase of product clustering, an adaptive clustering algorithm based minimal spanning tree and geneticalgorithm was proposed. To compress the length of gene code and searching space of problem, a newcoding method based on minimal spanning tree is employed in the algorithm. The algorithm canautomatically estimate the optimal number of clusters without a-priori information. In the phase ofvirtual cell reconfiguration, a non-linear multi-objective integer programming model was constructedand a new parallel collaborative multi-objective particle swarm optimization (PCMOPSO) algorithmwas proposed. Finally, the algorithms were verified through examples.3. Aiming at the characteristics of scheduling problem in the virtual cellular shop-floor, themulti-objective scheduling problem with large dimensional searching space and scheduling problemwith resource shared by different cells ware studied respectively. To solve the multi-objectivescheduling problem with large dimensional searching space, a preference based multi-objectiveparticle swarm optimization algorithm (PMOPSO) was proposed. The preference information ofdecisions maker is incorporated into the algorithm to lead the searching direction. So that, not only thesearching space is compressed and the efficiency of the algorithm is improved, but also just a fewtrade-off solutions located in preferred area are obtained in a single run, and the hard work ofchoosing a satisfying solution from numerous non-inferior solutions is eliminated. In the algorithm, anew expression method of preference information based on importance relationship among objectivesand the value range of objectives or objective weights was proposed. With this method, not only thepreference of decisions maker can be easily specified, but also the range of searching area can beadjusted properly according to the requirements of decisions maker. In view of the characteristics ofpreference information, a new preference information handling method, which simulates the “vote” ofhuman society, was proposed. The method is intuitive, simple and easy to use. To solve the schedulingproblem with resource shared by different cells, a distributed collaborative multi-objective particleswarm algorithm was proposed, in the algorithm a new decoding method which simulates marketmechanism is employed to solve the resource conflict. Finally, the performance of the abovealgorithms was evaluated through simulations, and the results demonstrate the feasibility andefficiency of proposed algorithms.4. The multi-objective dynamic scheduling problem with lot-splitting for multiple product orders withdifferent due dates was studied. First, the traditional multi-objective dynamic scheduling problem wasstudied. To solve the multi-objective dynamic scheduling problem, a new multi-objective dynamicscheduling algorithm based on particle swarm optimization algorithm and rolling-horizon was proposed. In this algorithm, periodic and event driven rescheduling strategies were employed and thedynamic scheduling problem was decomposed into a series of continual and static schedulingproblems, then an improved multi-objective particle swarm optimization algorithm were applied tooptimize each of the static scheduling problems, and the decisions maker can choose a satisfyingsolution from many Pareto optimal solutions obtained in a single run. To compress the searchingspace, a new active multi-objective scheduling decoding method was employed in the algorithm. Tosolve the multi-objective dynamic scheduling problem with lot-splitting for multiple product orderswith different due dates, a novel multi-objective flexible size lot-splitting dynamic schedulingalgorithm based on particle swarm optimization algorithm was proposed. Local and global updatingstrategy are both considered in the algorithm, local updating strategy is adopted for those turbulencesthat happen in high frequency but have little effect on the scheduling; otherwise, global updatingstrategy is adopted. In the algorithm, a flexible size lot-splitting approach based on “cursors” was putforward. Combined the lot-splitting and the lot scheduling, a novel particle coding scheme wasproposed. So that the algorithm not only can split lots into flexible size sub-lots according to machineworkloads, but also can optimize the lots routing and sequencing simultaneously. The performance ofthe proposed algorithms was evaluated through simulations, and the results demonstrate the feasibilityand efficiency of the proposed algorithms.5. Finally, A MES prototype system for virtual cellular shop-floor was designed and developed. By itsexperimental application in industry pump and valve factories, the application problem of system isdiscussed.