Paper SS-TuP35
Evaluating Influencing Factors on Nanofiltration and Reverse Osmosis Membrane Processes: A Computational Chemistry Approach to Modeling

Tuesday, October 29, 2013, 6:00 pm, Room Hall B

We have developed quantitative structure-property relationship (QSPR) models to assess the efficacy of water treatment processes for removal of endocrine disrupting compounds, pharmaceuticals, and personal care products (EDC/PPCPs). Data sets were developed from literature sources reporting nanofiltration and reverse osmosis membrane rejection. QSPRs were developed and implemented by relating compound properties to their removal in membrane water treatment processes. Properties were coded by descriptors capturing physical, chemical, and electronic molecular features. Descriptors were selected maximize model accuracy, sensitivity, and pre dictive ability. Individual descriptor importance was assessed using stability of descriptor weights across observations and sensitivity of model predictions across perturbations in descriptor values. Models were developed using Partial Least Squares and Support Vector Machine techniques, and were cross-validated and y-scrambled to verify that results were not serendipitous.

Initial modeling work was done to primarily investigate influential characteristics of the contaminant compound. Descriptor sets were then expanded to attempt to capture membrane influences, including membrane surface chemistry, as well. Modeling was also broadened to develop predictive models based on mechanistic data, instead of a process-specific percent rejection value. Mechanistic parameters of EDC/PPCP fate and transport in membrane processes can be determined from a combined film and hydrodynamic model, a pore surface diffusion model (PSDM) for adsorption, and a steric, electric, and dielectric exclusion (SEDE) model.