A Genetic Algorithm for a capacitated lot-sizing problem with lost sales, overtimes, and safety stock constraints

Overview & Synopsis

This paper deals with a complex production planning problem with lost sales, overtimes, safety stock and sequence dependent setup times on parallel and unrelated machines. The main challenge of this work is to propose a solution approach to obtain a good feasible plan in a short execution time (around 2 minutes) for large industrial instances. We develop a genetic algorithm that combines several operations already defined in the literature to solve the problem. Preliminary numerical results obtained with our algorithm are presented and compared to a straightforward MIP resolution. The method appears to be an appealing alternative on large instances when the computational time is limited.