Disturbance propagation analysis of complex chemical process based on directional mutual information

Oscillations and disturbances, which are usually caused by valve stiction or improper tuning controller, etc., are common problems in complex chemical process. A fault may propagate through the plant because of interlinking of processes and affect the performance of whole system. In this study, we proposed a new measure to evaluate causality between variables. First, the primary time series are coded in an ordinal way, which can be regarded as a symbolic procedure. Using the symbolic sequence, a mutual information like measure is obtained and a time delay is introduced to make the symbolic mutual information present directional characteristic. Hence, this asymmetrical measure can be used to construct a causality index to quantify the strength and direction of interaction. The performance of the method is illustrated by simulation on TE benchmark.