The Research On Modeling And Scheduling Of FMS Based On Petri Net

Flexible manufacturing system is a typical discrete event dynamic system. The development of flexible manufacturing system scheduling and planning technology has a significant impact on improving the performance and efficiency of whole flexible manufacturing processe. The model and algorithm are two keys of flexible manufacturing system scheduling. Among them, model is used to solve the problem of flexible manufacturing system modeling, and algorithm is mainly used to solve flexible manufacturing system optimization and performance analysis.In this paper, the application of Petri nets is researched in-depth. First, a more comprehensive overview about how to use Petri nets modeling flexible manufacturing system is given. Second, a scheduling method combining genetic algorithm and Petri nets is proposed. A model of flexible manufacturing system is created by using timed Petri nets. Then genetic algorithm is used to schedule the model, and a near optimized result is got. In this method, the new encoding-decoding mechanism is adopted. The firing sequences of the Petri nets model are used as chromosomes. Each chromosome in population corresponding to the firing sequences is no longer required to meet the condition of reachability, but it is converted to the firing sequence meeting reachability through the decoding section, thus the operations of population initialization, crossover and mutation become very simple. The method combines the strongpoints of genetic algorithm and Petri nets, it can solve flexible manufacturing system scheduling problem better.For the large-scale multi-stage multi-product scheduling problem, in order to further simplify the system model and improve the algorithm performance, the colored Petri nets is used to model flexible manufacturing system. and the dynamic chain-like agent genetic algorithm is proposed. The algorithm combines the coding characteristic of genetic algorithm with evolution structure of agent system. The one-to-one correspondence between the encoding and feasible scheduling is achieved by new assignment rules. The population evolution is implemented by the operators of agents such as competition and cooperation with the dynamic neighboring environment and self-learning operator with their own knowledge. The simulation results of large-scale multi-stage multi-product scheduling problem show that the combination of dynamic chain-like agent genetic algorithm and new heuristic rules not only increases the diversity of the population but also improves the convergent performance, and it is an effective algorithm to solve large scale multi-stage multi-product scheduling problem.