Approaches to supply chain coordination: decomposed and decentralised decision making models
2017-02-23T00:02:26Z (GMT) by
Modern supply chains have shifted from hierarchical, one-dimensional supply chains and increasingly operate in a global supply network. Supply Chain Coordination (SCC) focuses on optimising operations such as supply, manufacturing and distribution in the supply network of an enterprise. The major concern in such networks is to achieve coordination without compromising the autonomy of individual units or partner organisations. This area provides ample opportunities for research. Different modelling approaches based on operational and decision-making aspects of the supply chain can be used to address SCC issues. An integrated model (IM) is one which combines the constraints and objectives of different decision making units (DMUs) into a single, huge optimisation model. Integration allows the supply chain to solve all its sub-components (DMUs) simultaneously and hence, it can guarantee feasibility across DMUs compared to solving the DMUs separately. However, such models, being large and complex, are computationally difficult to solve. The solution, if any, need not be optimal for all DMUs. Often, the independent players in the system are reluctant to share all their competitive information in public. If the players are not willing to share complete information, then an attempt at an IM is extremely difficult. Therefore, the development of alternative approaches for SCC becomes increasingly necessary. In this thesis, new approaches for SCC are proposed using decomposed and decentralised decision making models based on Lagrangian relaxation (LR) and column generation (CG) methods. This is motivated by a real-world coal mining example, which is generalised to a multi-resource constrained scheduling problem. The transition of solution approaches from integrated approaches to decomposed approaches and then to decentralised ones is presented. The industry seeks a coordination approach that can deliver quality solutions in a reasonable amount of time without compromising their autonomy and their confidential and competent information. Therefore, decentralised decision-making will be the driving force in the future for supply chain coordination. A two-party coordinated production-planning and resource-scheduling problem involving a set of independent producers (multiple mines) and a shared resource manager (rail operator) is considered. The decisions in this SC are decomposed by relaxing the resource sharing constraints which link the DMUs. Decomposition approaches based on LR and CG are then developed. Several strengthening methods and stabilisation techniques have been implemented to improve the LR and the CG algorithms. The decomposition approaches are compared with the integrated approach to benchmark the performance of distributed decision making. Decentralised approaches are developed by further reducing the information-sharing and eliminating the central coordinator in the decomposed approaches. The role of information-sharing in a decentralised approach and how to quantify the usefulness of an information, are also addressed in this thesis. For the two-party case of the coal supply chain, two vital pieces of information are identified---the production capacity (at mines) and the resource capacity (at rail operator). The decentralised scheme proposed, based on LR, guarantees convergence and is found to outperform IM in terms of obtaining solutions in reasonable time. The value of information is also quantified using experimental results. The study is further expanded to a three-party case by including the second resource manager (the terminal in the coal supply chain). The three-party decentralised model was solved using a modified CG algorithm. A decentralised method called secure-sum has been implemented in this algorithm to compute the lower bound in all iterations. This algorithm is also strengthened with column management and other techniques. The computational results highlight the impact of an additional player and the value of information-sharing. The results show that the decentralised model could achieve better or equivalent solutions compared to that from the integrated model with significantly less information and interactions. In summary, the thesis proposes a scalable and robust, decentralised framework of decision-making for a generic multi-party supply chain, which is a better alternative to the integrated approach. It requires only minimal information sharing between the players and guarantees convergence by means of the underlying decomposition algorithm. The approach can be used even as the level of coordination (information-sharing) improves. The proof of the concept has been demonstrated using a large and complex multi-party coal supply chain. Thesis submitted in partial fulfilment of the requirements for the degree of Doctor of Philosophy of the Indian Institute of Technology Bombay, India and Monash University, Australia.