Fig. 3: Evaluation of data-driven transport model for hydrocarbon mixture separations.
From: Data-driven predictions of complex organic mixture permeation in polymer membranes
(a-Torlon, b-Matrimid, c-DUCKY-9, d-DUCKY-10) Comparisons between experimentally-measured permeate concentrations and predictions. All experiments and predictions were performed at 22 °C. The pressures in the plots indicate the transmembrane pressure (i.e., the applied pressure at upstream side with an atmospheric pressure at downstream). The error bars represent the standard deviation of the permeate concentration predictions for each molecule, and the deviations are from uncertainty in the machine learning (ML) sorption model parameter predictions. The deviations in the total flux predictions are from the uncertainty in the ML diffusion model predictions (Supplementary Tables 3–6). Source data are provided in the Source Data file.
