Solving Multi-fuzzy Factors Vehicle Routing Problem With Improved Lévy Particle Swarm Optimization

With the increasing need of customers in diversity and personalization of products,Make-To-Order (MTO) and customization have been introduced in the process ofmanufacturing which is considered as mixed-model production. The mixed-model productionleads to frequently shifted production lines and plans. Under the requests of lean productionand JIT, the production management staff needs to schedule the production plans and materialdistribution solutions immediately. Manufacturing Execution System (MES), which isconsidered as an indispensable link between workshop and enterprise, could be useful andhelpful in assisting workers in management. In this paper, an intelligent model with multiplefuzzy parameters has been designed to depict the actual logistics process of manufacturing.The main fuzzy factors are fuzzy travel time of vehicles, fuzzy requirements and fuzzy duetime of workstations.The model constructed above is complex and sophisticated, so we need a more efficientalgorithm to solve it. Firstly, the search pattern of particle swarm optimization (PSO) isinvestigated and seems to be characteristic power law distribution. Secondly, the Lévy Flightsearch pattern, which is considered as an efficient way to search the unknown large space, iscombined with PSO, it leads to a new algorithm named LévyPSO. The roulette heuristicalgorithm is used to generate excellent initial population of LévyPSO. Based on these work,the LévyPSO could converge to the optimum value of the target problem, which isconstituted by several fuzzy factors. In the end, a system which is developed based onLévyPSO and Anylogic is implemented in manufacturing plants. It turns out that the systemcould improve the efficiency of production and reduce the cost of material distribution.