A new method of failure detection and isolation for sensors and actuators

An approach to failure detection and isolation is presented which is based on observer theory and the pole assignment technique in multivariable linear systems. Using a full order feedback gain matrix, the proposed scheme offers the advantage of ensuring that all observer poles are stable and that all eigenvectors are parallel to as many failure events as there are states. The method of solution for the evaluation of the feedback gain matrix, even in the presence of a direct input term to the output, is presented. An example using a 4-wheel steering vehicle is used to illustrate the theory presented. >