Abstract: In the feedback control loop, the adjustment ability of controller covers up the performance degradation of the actuator to some degree. A fault state evaluation algorithm based on a two-class non-negative matrix network is proposed to implement online fault state monitoring of the actuator, including fault classification and degradation assessment. The local static features of the samples are extracted, and a classifier model is established to form a network. The similarity is introduced to describe the dynamic characteristics between samples. To fulfill the actuator fault status assessment, the static distance and dynamic changes of the network output are merged to construct the fault discrimination index and degradation index. The effectiveness of the algorithm is proved by the experiment of double-tank tank liquid level cascade control and the single loop simulation based on DAMADICS platform.

Key words: actuator, non-negative matrix factorization, fault diagnosis, degradation assessment