Recurrence Based Strategy for Fault Detection of Tennessee Eastman Process
Volume Title: 1
16 Azar Ave., Enghelab
This paper develops a methodology for detecting faults in the Tennessee Eastman process. In order to show capability of this method, Fault No.4 was selected. For this purpose, recurrence based methodology was utilized and two determinism and entropy parameters were calculated. Input parameters (time delay, embedding dimension, length of signal, minimum length of diagonal line, radius threshold) were optimized and calculated according to dynamics of process, based on which the recurrence plot was created. Variation of parameters was determined by 95% confidence interval. Results showed that the recurrence based method is able to detect faults of the Tennessee Eastman process.