Deterministic global optimization for error-in-variables parameter estimation

Standard techniques for solving the optimization problem arising in parameter estimation by the error-in-variables (EIV) approach offer no guarantee that the global optimum has been found. It is demonstrated here that the interval-Newton approach can provide a powerful, deterministic global optimization methodology for the reliable solution of EIV parameter estimation problems in chemical process modeling, offering mathematical and computational guarantees that the global optimum has been found. Although this methodology is typically regarded as being applicable only to very small problems, it is successfully applied here to problems with over 200 variables. It is a general-purpose technique and is applied here to a diverse group of problems, including examples in reactor modeling, in modeling vapor-liquid equilibrium, and in modeling a heat exchanger network.