Scheduling batches in flowshop with limited buffers in the shampoo industry

In this paper we address the problem of planning a temporary storage area in a real production system. This temporary storage area is composed of parallel temporary storage units with distinct capacities. The storage operation of a job, also called a batch, has to answer time restrictions such as release dates, due dates, restricted family dependent setup times and time lags, and also a space constraint which is the capacity of the temporary storage unit. The goal is to schedule the batches on the storage units in order to minimize the total setup times and the maximum lateness. First, we model the problem on a single storage unit as a two-machine flowshop problem with a limited buffer capacity and we show that it is NP-hard. We also show that the particular case in which no lateness is allowed is solvable in polynomial time under special conditions on the buffer capacity, both for single or parallel temporary storage units. Next we provide three heuristics: a greedy algorithm, a hybrid heuristic based on Ant Colony Optimization and Simulated Annealing and finally a dedicated heuristic. The latter strongly exploits the structural properties shown in this paper. We provide experimental results which highlight the efficiency of the dedicated heuristic in comparison with the two other heuristics.

► We investigate the problem of scheduling batches on buffers with a non-negligible transfer and setup times in a real-life industry.
► We highlight a decomposition of the classical Finite Intermediate Storage (FIS) model into four sub-models.
► We provide a general bicriteria model for one buffer of this problem and proved its NP-Hardness.
► A special case of the two machine flowshop problem with a limited buffer can be solved in polynomial time.
► We provide three different heuristics for solving the industrial problem.