Single Queue Management of a Job Shop as Implemented by a Data Flow Architecture

We investigated management of an automated job shop having an unpredictable job stream and randomly failing machines. The shop’s stochastic nature precluded a traditional scheduling approach. We devised an artificial intelligence system based on a distributed data flow control architecture. Machine utilizations in a simulated shop based on typical machine and job stream characteristics exceeded 93% for average machine downtimes below 16%. This conference paper briefly describes the architecture and experiment. It considers at length the technology required to implement the control system, and speculates briefly about the effect of this system on machine tool design.