Component based modeling and fault tolerant control of complex systems

This paper addresses the modeling and control of a large-scale dynamical system operating in the presence of a fault. A component-based modeling system is first introduced. In this model, components are modeled individually with the interactions between components represented by an interconnection structure. A functional model of the system, and the mapping from the structural model to this functional model, is then defined. Performance and stability criteria are specified on the functional model of the system. Next, a hybrid hierarchical control strategy is presented consisting of a Fault Detection and Identification (FDI) routine and a Fault Tolerant Control (FTC) strategy. The FDI routine uses a wavelet neural net to detect and identify a fault condition, based on training data from simulated fault scenarios. The FTC routine consists of a hierarchical accommodation strategy. In the presence of a fault, the high-level redistribution controller re-routes the available control authority taking advantage of any inherent redundancy in the system. The mid-level set point controller then determines set point trajectories which maintain stability of the restructured system, possibly at some degraded performance. Finally, the low-level algorithm adjusts local controller gains in response to the new set points generated by the mid-level controller. Simulation results are shown for an unmanned aerial vehicle with a stuck main rotor collective actuator. In this situation, main rotor rpm is used as a redundant control. These algorithms are being implemented in an Open Control Platform which accommodates rapid reconfigurability, interoperability, plug and play operation and openness.