A column generation algorithm for the Buffer Allocation Problem approximated by the Time Buffer concept

The Buffer Allocation Problem deals with buffer sizing in production systems. In particular, the objective is to find an optimal buffer size configuration achieving some target performance measure. It has two main tasks to accomplish: searching for an optimal system configuration and assessing the system performance. In real setting, simulation-optimization is the common technique to approach this problem. Recently, an approximated mathematical programming approach has been developed for BAP simulation-optimization. The approximation consists in modeling queues with temporal lags (time buffer) and allows to devise Linear Programming (LP) instead of Mixed Integer Linear Programming (MILP) models. This latter makes easier the solution process, which remains a challenge due to the models huge and complex dimension solution space topology. In this paper, we propose an exact method for the solution of the time buffer approximated BAP. Tests on randomly generated instances showed that the proposed method is computationally more efficient than standard LP solver in most of the cases.