Metamodeling and multi-objective optimization for buffer allocation in multi-product serial production lines
Year: 2025
Type: article
Abstract: The allocation of capacity for the work in process is a complex optimization problem due to the stochastic nature of queueing systems; in addition, the variant with several classes of jobs has been scarcely addressed and so, there is a lack of proposals to quantify the buffer size in such situation. In this work we characterize the behavior of the throughput and cycle time as a function of buffer sizes in a serial production line with reliable stations and more than one job class. A simulation model built in ARENA software R16, a surface experimental design and regression analysis were used to obtain the approximate expressions of the throughput and the cycle time as a function of buffer size. The buffer sizes were obtained using a multi-objective optimization approach. The results show that, when all work classes are in the system, to maximize throughput, a greater buffer size must be allocated to slow stations; to minimize cycle time, the best allocation is a small buffer size at the fastest stations; the capacity allocated to each buffer is an indicator of slow and fast stations; finally the actual service time at a station is the result of the weighted service time of each work class. The proposal is replicable and flexible enough to add new variables; it will be useful for engineers and administrators who must plan the buffer size in systems that serve two or more work classes with random service times.
Authors: Salvador Hernández González, Idalia Flores de la Mota, Quetzalli Atlatenco Ibarra, Edgar Augusto Ruelas Santoyo
Language: English
Field: Decision Sciences
Domain: Social Sciences
Open Access status: diamond
OpenAlex ID: W4416525832
