Design and Experimental Evaluation of a "Fuzzy" System for the Control of Reactor Power

A rule-based, digital controller that incorporates fuzzy logic has been designed and experimentally demonstrated on the 5 MWt MIT Research Reactor. The potential value of fuzzy logic to the control of reactor power was assessed by comparing the performance of the fuzzy controller to that of one of identical structure that did not rely upon fuzzy logic. The fuzzy system perfomed better under degraded conditions, was less sensitive to noise, and required less movement of the actuator mechanism. Additional tests showed that the fuzzy controller adjusted the reactor power smoothly and without overshoot in response to changes in setpoint. Also, it functioned properly in the presence of externally-induced disturbances. Sensitivity experisents were conducted to identify the structural and design features responsible for the performance of the fuzzy system. Finally, the possible role of fuzzy logic in the control of industrial processes was explored.