Research On Distributed Flexible Job Shop Scheduling Based On Job Transfer And Emergency Events

Workshop scheduling is a major concern for most manufacturing enterprises,especially in situations such as multi-factory/workshop collaboration and emergencies,where it is extremely important to adjust scheduling resources to meet production plans.Such situations belong to the distributed flexible job shop scheduling problem(DFJSP),but research in this area is relatively scarce,and Hebei G Company frequently encounters this dilemma during actual production.Based on the company’s complex production situation,this study designed a multi-objective DFJSP mathematical model considering workpiece transfer(Distributed Flexible Job Shop Scheduling Problem with Multiple Objectives and Job Transfer,MT-DFJSP)and an emergency based multiobjective DFJSP mathematical model(Distributed Flexible Job Shop Scheduling Problem with Multiple Objectives and Dynamics,MD-DFJSP).The study proposes improved algorithms based on the NSGA-Ⅱ algorithm for each of them,and conducts corresponding experimental verification and comparison.The main research content of this article is as follows:(1)MT-DFJSP model was constructed that considers the optimization of maximum completion time,maximum factory load,and total energy consumption for workpiece transfer,and an improved NSGA-Ⅱ algorithm was designed.This algorithm designed an EGLR1 method based on energy consumption,which takes into account does the impact of energy consumption on the initial solution quality that is not considered in the GLR initialization method;In order to quickly obtain the optimal solution for the NSGA-Ⅱ algorithm,which only performs local search operations through mutation,a corresponding neighborhood search structure was designed based on three-layer encoding of factory allocation layer encoding,machine allocation layer encoding,and process sorting layer encoding;an elite library was established to preserve superior chromosomes to avoid the destruction of excellent individual genes during genetic manipulation.Combining the production scheduling of G company with Brandimarte examples,45 MT-DFJSP examples were constructed.Simulation experiments proved that the improved NSGA-Ⅱ algorithm performed better than the traditional NSGA-Ⅱ algorithm in 80% of MT-DFJSP examples.(2)An MD-DFJSP model was constructed that considers the maximum completion time and total energy consumption of unexpected events.Based on the aforementioned improved algorithm,the thesis modified the generation rules of factory selection layer encoding,designed the EGLR2 method,and changed the corresponding crossover and mutation operators.When a machine malfunctions,using a single rescheduling technique will increase the uncertainty of the rescheduling scheme.A new rescheduling scheme is generated using a mixed rescheduling strategy of right shift rescheduling and complete rescheduling;In order to avoid excessive completion time caused by inserting the order into a critical factory when inserting an order,the order is assigned to a non critical factory.High priority orders are rescheduled using a right shift,while other orders are completely rescheduled to generate a rescheduling plan for the corresponding factory.In the three emergency scenarios of MD-DFJSP,this method can generate rescheduling schemes based on the initial scheduling scheme,and has little impact on the maximum completion time and total energy consumption.(3)A production scheduling system based on the processing information of G company was designed and implemented.