Mode identification for multi-mode switching systems based on multi-sampled data

A novel on-line mode identification algorithm is proposed for a class of switching systems. In order to cope with temporal output deviations due to the modeling errors for each mode, the temporally non-distinct dynamics between different modes, and the time-varying noise features, a multivariate hypothesis testing method is extended based on multi-model framework by using multi-sampled data. The probability of false estimation is significantly reduced by using the proposed algorithm. The proposed method is illustrated by a dead zone detection problem of a hydraulic valve system.