Automated learning system for control and supervision of assembly systems

Automation has not been applied as widely in assembly as in other manufacturing areas, mainly due to the complexity of operations requiring control of many geometric and technological process parameters. This paper shows that automated learning can be applied to control and supervision of assembly systems for which only a qualitative process model exists. A qualitative process model is built from incomplete a-priori knowledge of processes, facilities and products including essential performance targets. Automated learning algorithms determine optimum process control and supervision strategies based on performance of the assembly system. The process model is updated in real time according to the process results. Critical parameters can be identified and supervision strategies optimized. A prototypical automated bonding system will serve as a practical example showing how automated learning can help determining and verifying control strategies.