Mar 2026
Data-driven and Multiscale Systems Design of Sustainable Nanofiltration Processes
.png?sfvrsn=80b925f0_1)
Committee Members Information
- Ph.D. Advisor: Professor Gyorgy Szekely (KAUST Chemical Engineering)
- External Examiner: Professor Gonzalo Guillén Gosálbez, Professor Doctor in Chemical Systems Engineering Dept. of Chemistry and Applied Biosciences, ETH Zurich (ETHZ)
- Committee Chair: Professor Suzana Nunes (KAUST Environmental Science and Engineering)
- 4th Committee Member: Professor Carlos Grande (KAUST Chemical Engineering)
Abstract
Synthetic membrane technology is increasingly recognized as a key enabler of sustainable separation processes, offering energy-efficient and low-emission alternatives to conventional separation technologies across a wide range of chemical industries. In this dissertation, I connect materials and membrane research with process feasibility and sustainability implications through a combination of data-driven modeling, mechanistic process simulation, and systems engineering.
My research spans multiple design scales, from molecular descriptors to process superstructures and system-level assessments. At the molecular level, I developed transfer learning regimes for catalyst rejection prediction based on a curated homogeneous catalyst rejection dataset. At the membrane transport level, I introduced the Cascading Selectivity Principle as a unified framework for describing membrane and process selectivity and defined universal selectivity merits and efficiency indices as cross-technology descriptors for membrane performance evaluation. On the process level, simulation-based virtual screening of nanofiltration and conventional separation technologies enabled conservative estimates of the energetic and sustainability benefits of nanofiltration, demonstrating that nanofiltration can provide energetic advantages for up to 74% of evaluated separations. At the systems level, I developed multi-objective optimization frameworks to quantify trade-offs between product composition, recovery, and energy demand, and demonstrated that non-conventional multistage nanofiltration superstructures incorporating interstage dilution and recycling can deliver superior process performance.
The methodologies developed in this dissertation were demonstrated and validated across several industrially relevant domains of water-based and organic solvent nanofiltration. Machine learning-based catalyst rejection prediction was validated in an e-fuel production process and applied to the virtual analysis of pharmaceutical purification. Multi-objective optimization was further demonstrated for ion separation in industrial brine mining. Finally, high-throughput data-driven techno-economic screening of nanofiltration was extended from the general chemical space of nanofiltration to fourteen application domains, ranging from synthetic intermediates to natural product purification.
Overall, this dissertation integrates cheminformatics, data-driven modeling, and process systems engineering to advance the understanding, design, and optimization of sustainable liquid-phase membrane separation technologies across multiple scales and application domains.
-jpgf64a5d6c-a745-4a2a-882b-e4f50364e8d3.jpg?sfvrsn=439c23b6_2)