Supply Chain Analysis Using Simulation, Gaussian Process Modelling and Optimisation

Abstract

This paper presents a simulation study on production and inventory control at the supply chain (SC) level and examines through a series of experiments the potential impact of the Hybrid Kanban-CONWIP production control strategy on the trade-off between the conflicting objectives of maximising customer service level and minimising Work-In-Process (WIP). Simulation based optimisation is computationally expensive approach to determining an optimal configuration of operational parameters for any stochastic system. In this paper we demonstrate an optimisation framework that will yield solutions with an accuracy that is reasonable for decision makers and computationally less expensive than simulation based optimisation. A simulation model for a centralised serial SC adopting Hybrid Kanban-CONWIP to process a single product was developed in order to explore the impact of some essential input factors on customer service level and average WIP through Design of Experiments (DOE), Gaussian Process Modelling (GP) and Metamodel-Based Optimisation using the Desirability Function. The precision of the results from this approach was determined by comparison to results from Simulation-Based Optimisation by means of Genetic Algorithms (GA). It has been shown that this framework will address the trade-off between accuracy and computational efficiency requirements of the decision maker.