Abstract

Data validation is important in chemical industries. Because of random and possibly gross errors in measurements, data reconciliation is needed to minimize the measurement errors. In this article, an adaptive method is presented for dynamic and linear reconciliation of process data for real time optimization (RTO) of the process. In this method, system model parameters are estimated by the recursive least square identifier. By using these parameters, errorless process data are generated through Kalman filter method. This method is implemented using the Simulink tool box of Matlab software package. In this study, a rigorous model for the sweetening process is developed. Data generated by a sweetening Process simulation is used for evaluation of adaptive data reconciliation. First, data is artificially contaminated to errors (white noise) in Simulink environment and then filtered by the proposed method. Comparison of the outlet data for the actual and filtered process data shows great improvement and the effectiveness of the adaptive method is demonstrated.