Hybrid system diagnosis and control reconfiguration for manufacturing systems

A methodology for representing and analyzing manufacturing systems in a hybrid systems framework for control reconfiguration purposes in the presence of defects and failures at the product and system levels is presented. At the top level, a supervisory Petri net directs parts/jobs through the manufacturing system. An object-based hybrid systems model that incorporates both Petri nets at the event-driven level and differential equations at the time-driven level describes the subsystems. Rerouting capabilities utilizing this model at the product and operation levels were explained. Simulations were performed on a testbed model for optimal time and mode transition cost to determine the route for parts. The product level reconfiguration architecture utilizes an adaptive network-based fuzzy inference system (ANFIS) to map histogram comparison metrics to set-point adjustments when product defects were detected. Tests were performed on good and defective plastic parts from a plastic injection molding machine. In addition, a mode identification architecture was described that incorporates both time- and event-driven information to determine the operating mode of a system from measured sensor signals. Simulated data representing the measured process signals from a Navy ship chiller system were used to verify that the appropriate operating modes were detected.