Efficient Evaluation of Vacuum Pressure-Swing Cycle Performance using Surrogate-based, Multi-objective Optimization Algorithm

The performance evaluation of vacuum pressure-swing adsorption (VPSA) cycles with detailed, full-order models (FOMs) is typically a time-consuming and resource- intensive task. Popular state-of-the-art approaches include e.g., dynamic optimization approaches—see e.g., Swartz and Kawajiri (2019); and application of evolutionary algorithms, i.e., genetic algorithms, see e.g., Fiandaca et al. (2009). In this work, we present a strategy that combines two techniques in order to improve the computational efficiency of VPSA process modeling. Firstly, we improve the convergence in the calculation of cyclic steady state (CSS), using the FOM, by treating the calculation task as a fixed-point iteration (FPI) problem. We apply the FPI acceleration method by Anderson (1965) to improve FPI computations. Secondly, we embed the accelerated CSS computation into algorithm ‘Surrogate Optimization of Computationally Expensive Multiobjective Problems’ (SOCEMO) created by Müller (2017), which specifically addresses multi-objective optimization of expensive-to-evaluate black-box functions. This combination of approaches improves computational performance for Pareto frontier estimation that requires evaluation of process metrics at CSS, without sacrificing high-fidelity of the full-order VPSA model, and thus constitutes an attractive alternative to evaluate cyclic adsorption processes.

Keywords: surrogate-based multi-objective optimization, periodic processes