%P 985-991
%T Comparison of ARX and ARMAX Decorrelation Models for Detecting Model-Plant Mismatch
%I Elsevier Ltd
%A F. Uddin
%A L.D. Tufa
%A S.M.T. Yousif
%A A.S. Maulud
%V 148
%D 2016
%R 10.1016/j.proeng.2016.06.536
%O cited By 7; Conference of 4th International Conference on Process Engineering and Advanced Materials, ICPEAM 2016 ; Conference Date: 15 August 2016 Through 17 August 2016; Conference Code:131138
%L scholars7437
%J Procedia Engineering
%X Process model is the kernel element of Model Predictive Control (MPC) system. It is always desirable to get a model as accurate as the actual facility or plant to reduce the built-in mismatch. With the passage of time, the mismatch between model and plant increases, which results in degradation of MPC performance. To rectify mismatches through plant re-identification is exorbitant and time consuming. Hence, mismatch detection is critical to isolate the faulty sub models to avoid complete re-identification. Badwe et al. proposed a method using partial correlation to isolate and detect plant-model mismatch which uses dynamic models in the decorrelation step. This study extends his work by comparing the performances of Autoregressive Exogenous Input (ARX) model and Auto-Regressive Moving Average with Exogenous Input (ARMAX) model for detection of model-plant mismatch. Wood and Berry binary distillation column is used as a case study to demonstrate the application of the ARX and ARMAX models in mismatch detection. Results show that ARMAX models provide higher accuracy with less model order as compared to ARX. This results in less computational complexity and less processing power required in the MPC, hence improving its efficiency. © 2016 The Authors.
%K Correlation methods; Distillation columns; Predictive control systems; Process engineering, ARMAX; Auto-regressive exogenous inputs; Autoregressive moving average; Binary distillation columns; Partial correlation; Plant model mismatches; Processing power; Re identifications, Model predictive control