Optimal design of hybrid push/pull manufacturing system of fork/join queuing network topology using genetic algorithm.

摘要

The primary purpose of material control strategies such as MRP/push and Kanban/pull is to provide an efficient mechanism for determining the way production is triggered and inventory is controlled on the manufacturing shop floor. Both types of production control systems have advantages and disadvantages, thus it would be valuable to integrate both types to utilize their advantages.; This research explores the question of whether each part type should have its own junction point (less constrained) or whether there should be one common junction point for the overall system (easier to implement).; In this research a hybrid push/pull production control strategy is applied to Fork/Join Queuing manufacturing systems (FJQNs) with multiple part types. Certain parts of the system will be controlled using a push mechanism and the other parts will be controlled using a pull mechanism. The system is optimized by locating points of integration, and determining the optimal values of safety stocks for the push part and numbers of Kanbans for the pull part. The number of possible combinations of values of these variables exceeds 1014 in the moderate-sized motivating example in this research and so a solution strategy is required. A specialized genetic algorithm (GA) is used to obtain optimal solution for this large-scale problem. A modification of the GA chromosomes and crossover process is developed for the optimization. The optimization involves evaluations of the stochastic variables by a discrete event simulation model. Tunable software that integrates GA and computer simulation is developed. An algorithm based on the adjacency matrix of a directed graph is proposed to generate all push pull points of integration for any Fork/Join Queuing Network (FJQN).; Two motivating case studies from an aerospace manufacturer are presented. First a multiproduct multistage serial manufacturing system of a tube shop is analyzed. Second, a Landing gear assembly line, is modeled as a FJQN and analyzed.