Using system dynamics, neural nets, and eigenvalues to analyse supply chain behaviour. A case study

Abstract

This paper presents a new methodology to predict behavioural changes in manufacturing supply chains due to endogenous and/or exogenous influences in the short and long term horizons. Additionally, the methodology permits the identification of the causes that may induce a negative behaviour when predicted. Initially, a dynamic model of the supply chain is developed using system dynamics simulation. Using this model, a neural network is trained to make online predictions of behavioural changes at a very early decision making stage so that an enterprise would have enough time to respond and counteract any unwanted situations. Eigenvalue analysis is used to investigate any undesired foreseen behaviour, and principles of stability and controllability are used to study several decision configurations that eliminate or mitigate such behaviour. A case study of an actual electronics manufacturing company demonstrates how to apply this methodology and its real benefits for enterprises.