Data-driven Process Design Exemplified on the Steam Methane Reforming Process

Process design based on physical models often faces computational problems with respect to convergence, especially if the underlying flowsheets are complex. The use of data-driven surrogate models promises to overcome these challenges. This contribution presents the development of surrogate models and their use for flowsheet simulation. A new sampling strategy consisting of a combination of adaptive and sequential sampling enables the selective placement of new sample points. It is shown, however, that this hybrid strategy does not necessarily lead to higher accuracies than a pure sequential sampling. Surrogates are built for selected key units of the steam methane reforming process, and their individual accuracies are analyzed. When the surrogates are combined to form flowsheets, the prediction errors show a tendency to damp from unit to unit. This proves the suitability of surrogate models for flowsheet simulations. The promising results of this paper pave the way for future work, such as the optimization of flowsheets or superstructure optimization.