The Study Of Remanufacturing Process Routing Modeling And Machine Learning For Used Components

Remanufacturing is a kind of recycling process, which turns used products to“like-new” remanufactured ones. And it has great meaningful to relieve thecontradiction between resource shortage and wasting of resources, and protectenvironment. With the support of national Natural Science Foundation of China (NO.51105395), this paper emphasizes on the study of remanufacturing process planning(RPP), including classification of used components, remanufacturing process routing(RPR) and so on.First of all, the status and tendency of remanufacturing and RPP at home or abroadare analyzed. Based on the analysis of the uncertainties for used components, a kind ofRPP system structure is established, including the RPP, the key technology of RPP, andthe supporting layer of RPP. Then the impact of uncertainties to RPP is also analyzed.Secondly, based on the analysis of the common failures of used components, ansystem of classification indexes for used components are built, which considering thetypes of failures, the degree of failures and so on. Then on account of these indexes, afuzzy clustering method for used components is presented, and put this method intopractice of used spindles.Thirdly, with respect to the uncertain remanufacturing process in remanufacturingsystems, a Graphical Evaluation and Review Technique network (GERT network) isintroduced, and a GERT-based uncertain remanufacturing process routings model for isbuilt, in which all probable remanufacturing process routings of used components aredescribed, and the probability of every process and time distribution function of everyprocess time are discussed. The probability and expected time of the used componentsfrom one node to another node in the model can be acquired by signal flow graph andmoment generating function. Furthermore, this model is applied to the remanufacturingprocesses of used spindles.Fourthly, on account of the uncertainty of RPR in GERT-based model for usedcomponents, a Bayesian network based machine learning model for probabilitieslearning is established. And this approach is taken into practice of remanufacturing ofused lathe spindles.Finally, based on the remanufacturing practices of Chongqing machine tools groupCo., LTD, the GERT-based model for used spindles is simulated by Arena software to demonstrate the justifiability. And the results of Bayesian network based machinelearning model for used spindles are calculated by Bayes Net Toolbox.