Use of confidence region in the optimal design of a separation process in the presence of uncertainties

A large number of real-world problems nowadays require that decisions be made in the presence of uncertainties. Traditionally, the task of separation systems design neglects the influence of uncertainties. This work, however, explores the optimization strategy of a frequently encountered equilibrium-based distillation system design problem in the face of uncertainties. In particular, uncertainties arising from thermodynamic model parameters, which represent the vapor-liquid equilibrium (VLE) behavior of non-ideal mixtures involved in the system and are germane to the rational design of the separation system, are considered in the work. Joint confidence regions for the key thermodynamic model parameters estimated by the nonlinear least squares method have been incorporated in the analysis of the uncertain regions. We propose a novel two-stage stochastic programming solution scheme utilizing the optimum searching capacity of the simulated annealing algorithm in the first (design) stage and ASPEN Plus flowsheet simulator in the second (operating) stage in which Hammersley sequence sampling scheme has been implemented. A case study that investigates the optimal design of a two-column distillation system in an isopropanol (IPA) process will be demonstrated. Overall, this study provides us with a useful means to scrutinize the uncertainties and to optimize our design by quantifying the impact associated with them.