Proceedings of the 2008 Winter Simulation Conference S. J. Mason, R. R. Hill, L. Moench, O. Rose, eds. OPTIMIZING INSPECTION STRATEGIES FOR MULTI-STAGE MANUFACTURING PROCESSES USING SIMULATION OPTIMIZATION Vahid Sarhangian Hamidreza Eskandari Mostafa K. Ardakani Abolfazl Vaghefi Dept. of Industrial Engineering School of Industrial Engineering Dept. of Civil Engineering Iran University of Science & Technology Tarbiat Modares University The Catholic University of America Tehran, 16844, IRAN Tehran, IRAN Washington, D.C. 20064, U.S.A. ABSTRACT This paper deals with the problem of determining the op- timal inspection strategy for a multi-stage production process using simulation optimization. An optimal inspec- tion strategy is the one that results in the lowest total in- spection cost, while still assuring a required output qual- ity. Because of the complexity of the problem, simulation is used to model the multi-stage process subject to inspec- tion and to calculate the resulting inspection costs. Simu- lation optimization is then used to find the optimal inspec- tion strategy. The performance of the proposed method is presented through the use of a numerical example. 1 INTRODUCTION Manufacturing systems are generally consisted of several stages, in which, raw materials pass through various op- erations and eventually transform into finished products. Each processing stage will produce a proportion of items that fail to meet the necessary requirements. A prelimi- nary idea to maintain the quality level is providing an in- spection station after the last stage. This is generally re- ferred to as outgoing inspection. However, with outgoing inspection, all efforts and costs invested in producing de- fective items during previous stages are wasted. It is more reasonable to place inspection stations after every major manufacturing process to insure that a specific quality level is being maintained. Efficient economic inspection strategies minimize the total inspection cost while ensuring the required output quality. In other words, a tight inspection results in higher product quality, but will also lead to higher costs of in- spection, scrap and rework. An optimum inspection plan balances these effects. After each manufacturing stage, a full or sampling in- spection can be performed. In full (100%) inspection, all of the items in a lot are inspected and defective items are replaced or reworked. However, in sampling inspection, a sample is picked and inspected from the lot and based on the defective items observed in the sample, the lot can be rejected or accepted. Accepted lots are released to the next stage and rejected lots are submitted for full inspec- tion. Good units replace all of the defectives found during the sampling or 100% inspection. Thus, in a multi-stage manufacturing system, the inspection strategy addresses the number and location of inspection stations and inspec- tion parameters (sample size, acceptance number) for each inspection station. Most multi-stage inspection models have focused on 100% inspection or screening, where defects occur inde- pendently and costs are proportional to the number of de- fects detected at each stage. Lindsay and Bishop (1964) showed that if the cost functions for inspection and re- working are assumed linear and the fraction of defective units for each stage is assumed to be fixed, then the 100% inspection will be more efficient than sampling. The simi- lar results were found by White (1969) in the case where the rejected items are repaired and replaced with good items. Raz (1986) reviewed the previous researches on multi-stage inspection allocation and found out that when the inspection costs are concave, then the optimal level at each stage is probably either 100% or 0%. On the other hand, Montgomery (2005) presented several situations in which sampling is most likely to be useful. For example, when the cost of 100% inspection is high and/or inspec- tions typically need destructive testing, sampling is pre- ferred. Heredia-Langner et al. (2002) presented a highly constrained multi-stage inspection problem where all stages must receive partial rectifying inspection and they solved their problem by using a real-valued Genetic Algo- rithm. Volsem et. al (2007) used simulation to model the multi-stage inspection problem and found the optimal in- spection strategy by an Evolutionary Algorithm. Their method is able to determine which type of inspection (0%, 100% or sampling) should be performed in each stage and the rigor of the inspections. However, in their research sampling parameters are considered fixed. Wu et al. (2001) pointed out in real cases; many un- avoidable factors (such as the wear of tools and the fluc- 1974 978-1-4244-2708-6/08/$25.00 ©2008 IEEE Proceedings of the 2008 Winter Simulation Conference S. J. Mason, R. R. Hill, L. Mönch, O. Rose, T. Jefferson, J. W. Fowler eds.