Flexible flow shop scheduling problems are computationally hard to solve. Their complexity further increases in practice due to highly customized products in to- day’s markets. Thus, this thesis studies the benefit of using marker-based genetic programming to warm-start mathematical solvers for multi-objective flexible flow shop scheduling with sequence-dependent setup times. The proposed hybrid ap- proach evolves dispatching rules and then converts them into a schedule that in- cludes job sequences and machine assignments. This schedule then initializes the solver’s decision variables. Performance is evaluated based on time-dependent and time-independent metrics. With a 20-minute solver time limit, our results show that up to 95% of the test runs benefit from a warm start. For large-sized instances, our warm-start approach improves the solution quality by 12.32% relative to cold starts. Our approach allows the industry to balance the trade-off between runtime and solution quality, since this approach provides quality solutions in a reasonable amount of time.     «

Flexible flow shop scheduling problems are computationally hard to solve. Their complexity further increases in practice due to highly customized products in to- day’s markets. Thus, this thesis studies the benefit of using marker-based genetic programming to warm-start mathematical solvers for multi-objective flexible flow shop scheduling with sequence-dependent setup times. The proposed hybrid ap- proach evolves dispatching rules and then converts them into a schedule that in- cludes job seq...     »