Идентификация зарождающихся неисправностей методом раздельного оценивания

Abstract

Originally the method of separate estimation developed by Friedland was used as an effective approach of computing complexity lowering of an expanded Kalman filter i.e. the filter which dimension is much higher than dimension of controlled system. The structure offered by Friedland represents pair of special filters which are isolated, but work in parallel, and linear combination of their exits approximates an exit of an expanded Kalman filter. In the most general case, this structure is not optimum in the sense that it is not completely equivalent to an expanded Kalman filter, being optimum only for perturbations of the determined type. In case of random disturbances the two-stage Kalman filter stops being an optimum filter because of connected with performance of very rigid restrictions which are difficult for executing in practice.