A method for the design of fault-tolerant systems in case of sensor and actuator faults

In this paper a new approach for fault-tolerant control (FTC) of systems subjected to multiple sensor and multiple actuator faults is presented. A model set is designed in such a way that the system in its normal or in each possible faulty mode of operation can be represented as a linear combination of the models in the model set. A possible way of selecting the model set in cases of sensor and actuator faults is presented. A recursive parameter estimation algorithm, in combination with a recursive Kalman filter, are used to produce estimates of the weights in the linear combination. These estimates are then used for reconfiguration of a generalized predictive controller. The proposed algorithm is illustrated in a case study using a linearized model of one joint of a real-life space robot manipulator, subjected to faults.