Efficient solution techniques for the optimization of scheduling and supply chain operations

This thesis addresses the development of efficient solution techniques for the optimization of supply chain, planning and scheduling problems. Decomposition solution strategies based on Lagrangean relaxation, Lagrangean decomposition, Lagrangean/surrogate relaxation, and subgradient and modified subgradient optimization are proposed and applied to a continuous flexible process network (CFPN) model. Problems of up to 63 time periods were solved. The comparison results show that strategies based on traditional Lagrangean relaxation and subgradient optimization provide a good initial approximation, while better solution values can be obtained from strategies based on the more sophisticated Lagrangean/surrogate relaxation and modified subgradient optimization.; Following the decomposition is the development of solution methods based on metaheuristics. A solution method based on the hybrid implementation of Branch & Bound approaches and the Tabu Search strategy is proposed. The method is applied to the optimization of the CFPN model. The results show that the method is capable of finding optimal solutions to problem scale of 7, 21, and 42 time periods, and good feasible solutions for larger instances of 63 and 84 time periods.; To cope with the increment in complexity in optimization problems under uncertainty, two scenario representations that rely on parallel and tree structures, are proposed. The number of scenarios increases with each time period progression in the tree structure, while the number of scenarios remains constant across all time periods in the parallel structure. Based on the two structures, three stochastic models are developed. The comparison between the three models shows that while the parallel structure may provide a simpler platform for the implementation of decomposition, the tree structure had better computational efficiency.; Lastly, an optimization model is developed for the planning and scheduling of single stage continuous multiproduct plants which was proven to be challenging and non-trivial in the literature. To improve its computational efficiency, two integer cuts are proposed and added to the model. A medium-term scheduling problem for a real-world polymer processing batch plant was efficiently solved; moreover, the model is compared with another model of similar nature and the results show that the former had better computational performance than the latter.