Human-Computer Interaction in Decentralized Control and Fault Management of Dynamic Systems

Abstract Concepts of decentralized control and fault management of complex dynamic systems are described. The impact and possibilities of human-computer interaction are emphasized. Three hierarchical levels of control and overall management are distinguished, namely (1) decentralized control and decision making on the lower level, (2) coordinating process computer control, and (3) human supervision on the highest level. An experimental investigation of the human's role as a supervisor and fault manager in a computer-aided decentralized control situation is explained. Fault correction is a particular issue of these experiments. Soft and hard failures occur in the decentralized controllers. Certain failure situations may even lead to safe shutdowns. System recovery has to be obtained. The human supervisor may shift the control responsibility to higher levels which allows an online maintenance of the decentralized controllers. The results show lower average decision times for corrections of hard failures as compared to soft failures. The subjects seemed to prefer human operator participation in controlling the subsystems, mainly, because the resultant performance was higher when only one failure was present. The human strategies of fault correction found in this study have to be considered in the design of future computer-aiding algorithms.