No abstract

in Wiley Online Library (wileyonlinelibrary.com).This article presents an algorithm developed to determine the appropriate sample size for constructing accurate artificial neural networks as surrogate models in optimization problems. In the algorithm, two model evaluation methods-crossvalidation and/or bootstrapping-are used to estimate the performance of various networks constructed with different sample sizes. The optimization of a CO 2 capture process with aqueous amines is used as the case study to illustrate the application of the algorithm. The output of the algorithm-the network constructed using the appropriate sample size-is used in a process synthesis optimization problem to test its accuracy. The results show that the model evaluation methods are successful in identifying the general trends of the underlying model and that objective function value of the optimum solution calculated using the surrogate model is within 1% of the actual value.

Carbon dioxide (CO 2) is one of greenhouse gases, which can cause global warming. One of studies to mitigate CO 2 emissions to the atmosphere is to convert CO 2 to valuable products (i.e., methanol). To make methanol production via CO 2 hydrogenation a competitive process, the optimal operating conditions with minimum production cost need to be considered. This paper studied an application of response surface methodology (RSM) in optimization of methanol production via CO 2 hydrogenation. The objective of this optimization was to minimize the methanol production cost per tons produced methanol. The sensitivity analysis was performed to determine the parameters that show significant impacts on the methanol production cost. Response surface methodology coupled with non-linear programming solver were used as the optimization tool. The results showed RSM was successfully applied to the methanol production via CO 2 hydrogenation process. The obtained minimum methanol production cost was $565.54 per ton produced methanol with the optimal operating conditions as follows. Inlet pressure to the first reactor: 57.8 bar, Inlet temperature to the first reactor: 183.6 • C, Inlet pressure to the second reactor: 102.6 bar, Outlet temperature of the liquid stream cooler after the second reactor: 63.5 • C, Inlet temperature to the first distillation column: 51.8 • C.

Aqueous amine is a solvent considered for carbon dioxide (CO2) recovery from the flue gas of a refinery gas turbine by chemical absorption/desorption process. The performance and the economics of this process depend on the choice of the amine absorbent, the concentration of the amine absorbent, the number of stages in the absorber and stripper columns, and the operating conditions. We used response surface methodology (RSM)a simulation–optimization technique, which uses local searches to estimate an appropriate direction to reduce the objective functionto optimize the amine-based CO2 capture process in a previous work [Nuchitprasittichai and Cremaschi Comput. Chem. Eng. 2011, 35, 1521–1531]. However, RSM does not provide any information about the quality of the obtained solution. In this paper, the RSM results are compared to those obtained by optimizing a global surrogate model of the system over the whole decision space with a global solver. We used an artificial neural network (ANN) as the global surrogate model. Depending on the accuracy of the global surrogate models, the solutions obtained using them can be shown to be global within the bounds of the data used to generate them. The comparison is used to assess the quality of the RSM results and their relative computational costs. Monoethanolamine (MEA), diglycolamine (DGA), diethanolamine (DEA), methyl diethanolamine (MDEA), triethanolamine (TEA), and blended aqueous solutions of these amines are considered in our analyses. The results reveal that the RSM algorithm yielded optimum solutions close to those obtained by the ANN approach for all solvents.

No abstract