A Hybrid Genetic Algorithm For The Finite Horizon Economic Lot And Delivery Scheduling Problem With Flexible Flow Line For Manufacturing And Remanufacturing

With the development of society and economy, the environmental problem and resources shortage is becoming more and more seriously, to achieve efficient scheduling about component that need to be manufactured and recycled component which should be remanufactured on the flexible flow line will become the focus of many researchers. In the paper, we investigate the lot and delivery scheduling problem with two production sources, manufacturing and remanufacturing, where a single supplier produces multiple components on a flow line with multiple processors and delivers them directly to an assembly facility (AF). It is assuming that all parameters such as demand rates of the components are deterministic and constant over a finite planning horizon.The main objective is to find a lot and delivery schedule about component need to be manufactured and recycled ones to be remanufactured that would minimize the average of inventory holding cost (also including cost about the recycled ones), setup cost, and transportation cost per unit time. The uncertain quality of recycled components causing the remanufacturing stage to be different, thus the problem that we discuss is classified into two cases:recycled components using machine which belongs to final stage to process and recycled components using machine which belongs to middle stage to process.We develop a new mixed integer nonlinear program (MINLP), adopt an optimal enumeration metho, and use ILOG CPLEX to solve the problem. Due to the difficulty of obtaining the optimal solution in medium or large scale problems, we incorporate neighborhood search (NS) which could perform genetic search over the subspace of local optima into genetic algorithm, propose hybrid genetic algorithm (HGA) to find an optimal or near optimal solution for majority of the test problems within a reasonable time is developed.Particularly important to emphasize is that:selecting good initial population can largely increase the efficient of GA. If recycled components using machine which belongs to final stage to process, we adopt heuristics such as CDS, Palmer, RA, Gupta et, and combinate FAM (first available machine)rule, MFAM (modified first available machine) rule to form initial population and assign components to machines at stages with parallel machines; If recycled components using machine which belongs to middle stage to process, MFlowmult method is developed, combining with FAM rule to form initial population and assign components to machines at stages with parallel machines; the design of the start time recycled components is based on the principle that reducing idle time of the machine and lowering inventory holding costs of component.Then two proposed solution method are compared with the solution of ILOG CPLEX, the lower bound of the problem on randomly generated numbers respectively, computational results show that HGA is very promising for it can find efficient solution for majority of medium or large scale problems.