Markov-parameter-based control reconfiguration by matching the I/O-behaviour of the plant

This paper describes the generalisation of a linear model-following approach to control reconfiguration after actuator or sensor faults. A static block is used in both cases, which imposes low computational demands on the implementation. It is first shown that the well-known pseudo-inverse method fails to yield a stable loop in many practical cases. The new design approach based on the recovery of the nominal plant I/O-behaviour described by the plant Markov parameters is introduced for actuator faults and extended to sensor faults. Conditions for reconfigurability are stated, and it is shown that an approximate solution can be used upon their violation. The methods are illustrated by a numerical aircraft yaw control example and verified by experimental results.