Detecting attractors in production systems by using system dynamics and neural networks
Daniel Thiel · 2002
This paper opens a new way for studying production system behaviours. It is based on networks of automata and particularly of Hopfield's neural networks model. All established causal schema based on Forrester's system dynamics principles, is converted into a neural network. The parallel computing which scrutinises all the combinations between the different model variables permits one to bring to light attractor behaviours (fixed points or limit cycles). The first simulation results based on the author's generic models of production systems, has permitted the author underline trends which some industrial companies could meet in the proximity of these attractor states. The author suggests a validation of these models and attractors by comparison with real observations in different companies.