Reactivity Identification in the Course of Physics Startup of a Reactor Based on Neural Network

摘要

Neural network is used to identify reactivity. The identification result can be compared with those ofother methods to reduce the physics startup risk. Levenberg-Marquardt algorithm is employed to train thenetwork. The training samples stem from neutron kinetic equation of six groups. The characteristic of neutronmultiplication is analyzed. Simulation results indicate that the network is capable of spreading to other sampleswhich have not been learned.