Production Scheduling Method Based On The Non-cooperative Game Model Research

Job scheduling is classified a NP-hard problem and researchers have studied morethan forty years. The target of job scheduling is to search the optimal solution forminimizing complete time, which also called makespan, or decreasing processing timeto insure due date. For job shop scheduling and flow shop scheduling, how to solve thecompetition between different task and production resource is worth to study and lowcost is plus.This paper proposes a model based on game algorithm to solve the job schedulingproblem. In the non-cooperative game, machines in the manufacturing task are definedas players, strategies consist of all the feasible programs which are selected bydispatching rules for minimizing the mean flowtime, and the mean flowtime is used toconclude the payoff. So how to find the optimal solution of job scheduling istransferred to search the Nash equilibrium.Considering strategies which used to find the Nash equilibrium, different singlepriority rule is used to solve job shop scheduling and flow shop scheduling, and thenresults are compared and performance is discussed. Applying constructive method andcomputational intelligence algorithm, a neural network is used to obtain the feasiblestrategies. Proper priority rule which belong to the strategies insure each machine havehigh profit, which mean low mean flowtime and cost.In the experiment, comparing with integer liner neural network and ant colonyalgorithm to solve the two category scheduling problems, improvement of networkstructure is proved. To avoid local optimum, genetic algorithm is used to optimize theweight and structure of network. Data and results show that the modified network hasbetter performance.