Neural-network-based catastrophe avoidance control systems

A novel approach based on interpolative neural networks is proposed for catastrophic fault detection and isolation, and system reconfiguration to accommodate the fault. The neural networks are from a class of recently developed interpolative neural networks, based on a generalized Fock space. The technique is designed to make use of secondary control configurations, assuming only partial system operation, as may be obtained by simulation and test results at design time (information not used by current adaptive controllers).<>