Optimal Design Of Configuration And Multi Scale Prediction Of Reconfiguration Time For Reconfigurable Manufacturing System

Nowadays, international market environment is changing quickly and the requirementsof customers are more and more customized and diversified. But the existingmanufacturing systems basically have no reconfiguration capability which is difficult toadapt the rapidly changing market demands. RMS appears to solve the problem efficiently.RMS can improve the responsiveness to the manufacturing enterprises and decreaseinvestment cost by configuration rearrangement, replacement, upgrades and other means toadapt to market changes. The study for related theory and technique of RMS has animportant effect for the manufacturing enterprises to dominate the market and make full useof resources in the future.The configuration technology for RMS was researched based on the research statusanalysis and existing problems. The main research contents of this thesis are as follows：1. Related research results about RMS were summarized, and some basic conceptswere explained. The key features of RMS were analyzed in detail. The important enablingtechnologies to implement RMS were summarized. The reconfigurable manufacturinghierarchy was also analyzed in detail. The comparison between RMS and some advancedmanufacturing systems shows the advanced manufacturing systems are the appearance ofRMS at different levels; the comparison between RMS and traditional manufacturingsystems suggests RMS has a variety of advantages. Reconfiguration scales was divided intomachine reconfiguration, cell reconfiguration and system reconfiguration.2. In order to make RMS adapt to the fluctuations of production demand by theminimum number of reconfiguration in its full life cycle, design method for RMS based onthe balanced distribution of functional characteristics for machines was presented. With thismethod, functional characteristics were classified based on machining functions ofcutting-tools and machining accuracy of machines. Then the optimization objective was setas the total shortest mobile distance that all the work pieces are moved from one machine toanother and an improved genetic algorithm (GA) was proposed to optimize theconfiguration. In this algorithm, elitist strategy was used to enhance the global optimizationability of GA, and excellent gene pool was designed to maintain the diversity of population. 3. In the workshop that machining functions are evenly distributed, the problem ofoptimal manufacturing system configuration selection from the enormous number ofalternative configurations was proposed. Correspondingly, a configuration optimizationmethod which combines similarity coefficient method and improved taboo search algorithmwas raised to resolve it. Firstly, a mathematical optimization model was established tominimize the parts handling distance and the area of the minimization rectangles thatconclude all the equipment in one configuration. Then, similarity part families wereconstructed which was based on machining functions similarity, meanwhile consideringmachining time and lot size. Afterwards, taboo search algorithm was designed to resolvethe mathematical model. In the algorithm, the initial solution was generated by arrangingrandomly the sequence of machines that possess the same machining function andconstructing prohibited machines set. The method to generate neighborhood by exchangingtwo adjacent elements for two segments can guarantee the full connectivity ofneighborhood and avoid generating infeasible solutions.4. A comprehensive prediction method based on wavelet analysis theory andautoregressive prediction model was proposed for the selections of reconfiguration scaleand time point in the life cycle of RMS. Firstly, the dynamic performance of RMS wassimply described. Then, wavelet analysis was adopted to decompose the configurationperformance signal as the configurations change signals including workshop floor, cellfloor and machine floor. Autoregressive prediction model was used to establishconfigurations change prediction model. The predictive recursive formula was establishedthrough determining the orders and estimating parameters. The model can accuratelypredict the data within the next k steps. Based on the predictive data, enterprisers canreasonably carry on right scale reconfiguration at right time.