Dynamic optimisation of a production-inventory control system with two pipelines feedback

Interest in the need to have a strong upstream and downstream integration within production-inventory systems has arisen recently as a means to gain a competitive market position. As to the methods of achieving this, companies need to investigate the dynamic implications of ordering strategies and different feedback control mechanisms to optimise inventory levels and reduce negative consequences such as order amplification. The dynamics of a production-inventory control system consisting of a single product, one manufacturer and one retailer was studied. The order rate related costs and inventory level related costs were considered as the two main factors to model the total costs of the production-inventory system. The dynamic analysis of the production-inventory control system was addressed via control theory and simulation as follows. A novel dynamic model for a production-inventory control system has been proposed. The dynamics were modelled as a linear continuous-time control system. The proposed model considers an extension and improvement to the Inventory and Order Based Production Control System (IOBPCS) models, and it utilises a new feedback flow of information in order to improve the efficiency of order rate decisions. A procedure to select a best system configuration was developed. A Multi-Objective Particle Swarm Optimisation (MOPSO) approach for generating Pareto-optimal solutions was used to optimise the overall dynamic performance of the system and filter-out all undesired operational performance. Optimal solutions were selected based on competing criteria of minimisation of the variance ratio (Var) between the order rate and the consumption, and minimisation of the Integral of Absolute Error (IAE) between the actual and the target level of inventory. The optimal system configuration with minimum cost was then chosen as the best system configuration. The efficiency of the proposed model was evaluated under several scenarios. A comprehensive simulation-based comparison between the proposed model and a model published in the literature (APIOBPCS) was conducted. First, three different demand patterns were considered (literature published demand data, computer generated random data, and natural inspired data such as weather based). The Pareto curves of the proposed model in comparison with results from the APIOBPCS model reveal that the proposed model provides a systematically better performance than the APIOBPCS model under the same considerations. For example, under thesis assumptions, the simulation using literature published demand data showed that the proposed model is capable of achieving a 6% cost reduction compared to the APIOBPCS model. The comparison result using randomly generated computer data illustrated that the proposed model achieved a 4% cost reduction. The comparison utilising a nature inspired demand pattern demonstrated that the proposed model accomplished a 3.5% cost reduction. Lastly, to ensure a realistic scenario, another set of simulation-based experiments under four different operational scenarios (normal, mismatched lead time, capacity constraint and initial condition) were performed. It was found that, under the mismatched lead time operation, both models are affected. However, the proposed model offers better inventory response indicating greater robustness. Moreover, the variance ratios under capacity constraint for both models are reduced, and these simulation results do not show any superiority of any model over the other. Finally, the variance ratios under an initial condition for both models are increased, and the inventory responsiveness for the proposed model gives better performance in this case. Overall, the proposed model therefore provides an improvement over currents model in the dynamic optimisation of a production-inventory control system