The behaviour of learning production control

The increasing diversity of products and the fluctuation in customer demand requires new strategies of decentral production control. This particularly concerns versatile and customer order oriented shop floor production. The introduced control strategy uses, on one hand, neural networks to control the inventories on each work system. On the other hand, the simulated annealing approach ascertains adequate target inventories. The effectiveness of this continuously learning production control strategy is verified by a theoretical model and one practical case. The control quality improves successively and the system can react according to changing conditions.