| With the development of globalization in manufacturing industry and increasing fierce market competition, manufacturing enterprises need to enhance their ability rapidly to respond to the emergency form inside and outside. In the real manufacturing environment, there are many kinds of uncertainties, such as machine breakdown, additional order arriving and emergency order insertion. Because of the existence of uncertainties, the Job-shop state will become dynamic. So scheduling under a dynamic environment has become a key to solve actual scheduling problem.This thesis focuses on the dynamic scheduling problems, which are caused by the dispersed uncertainties mainly about machine breakdown, additional order arriving and emergency order insertion. The main researches are as follows:First, a traditional Job-shop scheduling problem model is built and solved it by a designed GA which accord to the model; then, dispersed uncertainties, such as machine breakdown, additional order arriving and emergency order insertion, are analyzed, meanwhile the rescheduling process is built according to this three situation; then, a neural network is structured according to the feature of JSP model and its training algorithm is designed; then, the theoretical basis for the training sample is established by the various JSP models produced by enumerating various uncertain events based on rescheduling process according to a certain rule and by the solution solved by GA corresponded to each model; then the training sample sets are established, which the input is designed according to the feature of JSP model and the output is designed according to scheduling scheme; then, the designed neural network is trained; finally, based on the trained neural network and the rescheduling process for the three dispersed uncertainties, the intelligent rescheduling model is constructed, which has the rapid response to uncertainty events and the performance of making sure the results are close to the optimal. |
PDF Full Text Request