UNCERTAINTY OF PREDICTED PROCESS PERFORMANCE DUE TO VARIATIONS IN THERMODYNAMICS MODEL PARAMETER-ESTIMATION FROM DIFFERENT EXPERIMENTAL-DATA SETS

A method is presented using Monte Carlo Simulation coupled with the La tin Hypercube Sampling technique and parameter correlations to determi ne the effect of different sets of experimental data on the uncertaint ies in model parameters and, thus, the uncertainties in predicted proc ess performance. Specifically, the estimation of binary interaction pa rameters for the NRTL equation was studied for three liquid-liquid ter nary systems: diisopropyl ether + acetic acid + water, 1,1,2-trichloro ethane + acetone + water, and chloroform + acetone + water. From the e xperimental data available for these systems, regressions were done an d the obtained parameters were used to perform simulations on a liquid -liquid extractor. The results show that the differences between the b inary parameters from different sets of experimental data lead to sign ificant differences in the uncertainty of predicted extractions in pro cess units. (C) 1998 Elsevier Science B.V.