TY  - THES
T1  - Nonlinear data reconciliation with gross error detection and identification for steady-state and dynamic processes
A1  - Pilario, Karl Ezra S.
A2  - Muñoz, Jose C.
LA  - English
PP  - Quezon City
PB  - College of Engineering, University of the Philippines Diliman
YR  - 2015
UL  - https://tuklas.up.edu.ph/Record/UP-99796217611761473
AB  - Sensor measurements in a process network inherently contain random noise and/or gross errors. Thus, operational plant data should be pre-conditioned for process control, monitoring, and optimization. This work developed a strategy for nonlinear steady-state and nonlinear dynamic data reconciliation (DR) with gross error detection and identification (GEDI) for adjusting sensor measurements to satisfy process models. Steady-state DR-GEDI is posed as a mixed-integer nonlinear programming problem (MINLP) solved by a branch-and-bound technique. The branch-and-bound search tree is traversed by depth-first search (DFS), providing a systematic way of solving lesser relaxation NLPs. Each relaxation NLP is solved using a recent hybrid Nelder-Mead simplex and particle swarm optimization (NM-PSO) method suited for engineering problems. This strategy is modified for dynamic DR-GEDI using a moving horizon approach and a novel learning mechanism across horizons for data smoothing. The performance fo the DR-GEDI methods is tested for two highly nonlinear process networks under steady-state, and an adiabatic stirred-tank reactor with a first-order exothermic reaction under transient behavior. Results show that the adjusted steady-state measurements satisfy process models down to 10^-11 accuracy, and that the standard deviation of the smoothened responses improved by 80-90% from the standard deviation of the measured values. Interestingly, the algorithm was also able to spot what is called gross error equivalency, previously confirmed only in linear models.
CN  - LG 995 2015 E62 P55
KW  - Chemical processes : Data processing.
KW  - Error analysis (Mathematics)
KW  - Data reconciliation.
KW  - Gross error detection and identification (GEDI).
ER  -