A0188
Title: Variable contribution analysis in multivariate process monitoring using permutation entropy
Authors: Praise Otito Obanya - North-West University (South Africa) [presenting]
Roelof Coetzer - North-West University (South Africa)
Carel Olivier - North-West University (South Africa)
Tanja Verster - North-West University (South Africa)
Abstract: Permutation entropy (PE) - a statistical tool for the measurement of the complexity of a given time series - is used to estimate variable contributions to faults in an industrial process. Two sets of simulated industrial chemical process data, namely fault-free and faulty processes, are monitored using Hotelling's T-squared monitoring statistic. For the faults identified, PE is used to determine which variables contributed to those specific faults. Comparisons between the dynamics of the fault-free and faulty processes aid in the identification of the variables with the highest contribution to the specific faults. The well-known Tennessee Eastman Process is used to illustrate the application of PE for variable contributions. The results show that PE is an efficient analysis tool for estimating variable contributions to faults.
