(88f) Improved Characterization of Membrane Transport Properties through Advanced Data Analytics

Membrane processes have shown promise for addressing the critical needs for sustainability and energy efficiency. Recent material design to achieve separations of similar-sized molecules has evolved in the directions of precisely controlling the nanostructure of membranes and identifying chemical functionalities which accentuate desired transport properties. [1,2] A detailed understanding of the underlying thermodynamic and transport phenomena can elucidate the molecular interactions and mechanisms that affect the macroscopic transport properties of the membrane. [3,4] Motivated by this need, the development of membrane characterization techniques that explore the dependency of membrane performance on feed conditions can greatly accelerate the development of materials. [5] In addition, membrane characterization that elucidates underlying mechanisms provides essential information for scale-up, design, and optimization, facilitating the development of new separation systems.

Design of Experiments (DoE) methods enable optimization of computational and physical experiments that maximize the information gain and minimize time and resource costs. Classical ‘black-box’ DoE approaches (a.k.a. factorial, response surface), which decide the best design by the input-output relationship, do not (directly) incorporate membrane science knowledge; in contrast, model-based DoE (MBDoE) leverages high-fidelity models constructed from underlying physical principles that describe the experimental system. [6] The information collected from experiments can be applied to discriminate between scientific hypotheses, posed as mathematical models, and to improve the precision of parameter estimation. The emergence of techniques within MBDoE has great potential in the design of instruments and experimental conditions to better characterize the performance of separation devices as a function of solute concentration and in complex feed streams. However, to date, their application to problems in membrane science remains limited.

We recently proposed the Diafiltration Apparatus for high-Throughput Analysis (DATA), which enables a 10-times reduction in the time, realized with fewer experiments necessary to characterize membrane transport properties. [7] In follow-up work, we mathematically quantified these improvements in the form of information gain and further refine the static experimental conditions needed in DATA to characterize membrane transport properties. [8] In this talk, we apply the Fisher information matrix (FIM) analyses and MBDOE to further improve DATA. We highlight two non-ideal phenomena, namely “lag” and “overflow”, which occur when changing the operating pressure of the system. Guided by the tools of data science, we show that modeling these phenomena can leverage the additional data within the start-up process to elucidate the underlying physics, improve the parameter precision, and brings insights to design a time-varying applied pressure in DATA. A time correction for permeate product collected is also introduced to improve the model predictions. Moreover, our framework, which integrates data analytics and instrumentation design can be applied to investigate concentration-dependent membrane performance to further accelerate the development of materials. For example, we apply the improved DATA to explore the dependency of membrane transport parameters on feed conditions of a surface-charged membrane by ranking candidate models using information criteria.

References

1. Hoffman, J. R., & Phillip, W. A. (2020). 100th anniversary of macromolecular science viewpoint: integrated membrane systems. ACS Macro Letters, 9(9), 1267-1279.
2. Sadeghi, I., Kaner, P., & Asatekin, A. (2018). Controlling and expanding the selectivity of filtration membranes. Chemistry of Materials, 30(21), 7328-7354.
3. Geise, G. M., Paul, D. R., & Freeman, B. D. (2014). Fundamental water and salt transport properties of polymeric materials. Progress in Polymer Science, 39(1), 1-42.
4. Yaroshchuk, A., Bruening, M. L., & Zholkovskiy, E. (2019). Modelling nanofiltration of electrolyte solutions. Advances in Colloid and Interface Science, 268, 39-63.
5. Ghosh, R., & Cui, Z. (2000). Analysis of protein transport and polarization through membranes using pulsed sample injection technique. Journal of Membrane Science, 175(1), 75-84.
6. Franceschini, G., & Macchietto, S. (2008). Model-based design of experiments for parameter precision: State of the art. Chemical Engineering Science, 63(19), 4846-4872.
7. Ouimet, J. A., Liu, X., Brown, D. J., Eugene, E. A., Popps, T., Muetzel, Z. W., ... & Phillip, W. A. (2022). DATA: Diafiltration Apparatus for high-Throughput Analysis. Journal of Membrane Science, 641, 119743.
8. Liu, X., Wang, J., Ouimet, J. A., Phillip, W., Dowling, A. (2022) Membrane characterization with model-based design of experiments. In 14th International Symposium on Process Systems Engineering.