Nuclear reactor control system design with sensor failure

Abstract

The problem of enhancing control system reliability is studied considering several reactor models in conjunction with various computational methods. A cons [not legible] gain feedback law is calculated to satisfy the sensor failure constraint which requires that if the output of an arbitrary sensor fails to zero, the failed closed-loop system remains stable. A control law which satisfies the sensor failure constraint can be obtained using a method for calculating a constant gain output feedback law. Two methods for treating the output feedback problem are developed. One method employs an intersection technique for solving multivariable polynomial inequalities which are generated using a symbolic computer language and the Lienard-Chipart criteria; the other method employs conventional minimization algorithms to obtain a negative value of a functional. This functional involves the spectral radius of the closed-loop system and provides for implicit control over eigenvalue placement. An observation of importance is that the usual Pressurized Water Reactor models with neutron precursors lack the property of complete controllability and observability. This complicates the problem of obtaining a feedback law. In this dissertation, six different reactor models are considered. At least one group of neutron precursors should be included in the dynamical equations; however, more than one group is not necessary. The linear open-loop reactor models with a positive moderator reactivity coefficient have a positive real component in their eigenvalue spectrum; conversely, those with a negative moderator reactivity coefficient do not have a positive eigenvalue. For open-loop models which do not have a positive eigenvalue, the sensor failure constraint can be satisfied by direct output feedback. However, the sensor failure constraint could not be satisfied with direct output feedback for the models with positive eigenvalues. Extended feedback dynamics are needed to satisfy this constraint for the models with a positive eigenvalue..