The techno-economic analysis of chemical processes is usually based on steady state simulations. When dynamic processes are present, simplified or surrogate models are used to avoid the dynamic simulation of the overall system. Nevertheless, this usually results in loss of information when the simplified model is used or a large number of parameters when the surrogate model is applied. To circumvent this problem, a hybrid surrogate model is proposed that combines the low-cost prediction of the average behavior by a simplified model with the fine tuning provided by a multilinear look-up table. This concept was applied to the homogeneous alkaline transesterification of the soybean oil with ethanol in a batch reactor. The hybrid approach enabled a 70% reduction in the number of points of the look-up table compared to the pure interpolator model, for an accuracy tolerance of 0.01 mol/l.
Optimization of Chemical Processes Using Surrogate Models Based on a Kriging Interpolation
