Efficient fitting of nanofiltration model parameters for a specified groundwater type by selecting suitable characterization data-sets

Abstract

Nanofiltration (NF) models can be useful to perform optimal designs of membrane systems and to estimate membrane performance for waters. There is a special interest in obtaining NF models with parameters based on measurable properties of the membrane and independent from the feed and operating conditions. However, many times, from a practical point of view, NF parameters can be directly fitted from experiments performed with salts in a range of compositions. The aim of this study is to select the better combination of experiments to yield a suitable fitting for the NF model Donnan steric-partitioning pore model with dielectric exclusion (DSPM-DE). In our case, the best fitting for a specific group of waters is searched (groundwater belonging to a Mediterranean region with moderate salinity). The first part of the work is devoted to study which combinations of salts and concentrations lead to higher information. Using known values of NF parameters, permselective results were computationally generated using the NF model for a huge number of different combinations of compositions and random parameter sets. Performance factors for permeate flux and rejection based on the comparison between the characterization groups and a control group were defined. The second part of the work focused on the experimental validation of the selection procedure. The results showed that there are characterization sets (composition and operating conditions) that yield higher fitting performance. These combinations of experiments should be the preferred ones, when direct fitting from experiments is going to be performed.