Design problem solutions for Bayesian optimisation of PVSA CO2 capture processes

This repository contains MATLAB .mat files detailing the results generated by solution of the PVSA design problems presented in Ward & Pini (2022): “Efficient Bayesian Optimisation of Industrial-Scale Pressure-Vacuum Swing Adsorption Processes for CO2 Capture”. Please see the published manuscript for full details of the problem definitions and solution approaches. The filename for each .mat file specifies which design problem the results correspond to. The structure of the filenames is as follows: “optimiser_material_problem_pressure.mat”. The optimiser is either the Thompson Sampling Efficient Multi-Objective Optimisation (TSEMO) or the non-dominated sorting Genetic Algorithm II (GA). The material is either zeolite 13X or ZIF-36-FRL. The problem can be either unconstrained purity/recovery optimisation, constrained productivity/energy optimisation or constrained capture cost optimisation. The pressure corresponds to the fixed evacuation pressure for which the PVSA design problem is solved (pL = 0.02 – 0.1 bar). For the TSEMO results files, the objective functions are stored in the array “Y”. Where for unconstrained optimisation the columns of “Y” are purity (%) and recovery (%), for constrained optimisation the columns are productivity (mol/m3/s) and energy usage (kWh/tonne) and for cost optimisation “Y” contains the capture cost ($/tonne). Each row of “Y” corresponds to a specific operating point which TSEMO evaluated during solution of the design problem. The decision variables which generate the model outputs stored in “Y” can be found in “X”. The columns of “X” are as follows: [t_ads (s), t_bd (s), t_evac (s), pH (bar), pI (bar), vF (m/s)]. The arrays “Ypareto” and “Xpareto” are the entries of “Y” and “X”, respectively, which form the Pareto front from the full set of simulations. For the GA results files, only the final Pareto fronts are provided. The optimal objective function values are found in “fval”, which has the same format as “Y” in the corresponding TSEMO solutions. The decision variables which generate each operating point in “fval” are found in “x”, which has the same format as “X” in the corresponding TSEMO solutions. The excel file “TEA_outputs.xlsx” contains a detailed breakdown of the cost optimal processes generated by solution of the design problems presented in Section 6.1 of the published paper.