Experimentally Guided Neural Network and Statistical Forecasting of Membrane Water/Salt Selectivity with Minimal Mean Errors

Membrane life and performance are the key determining factors in the adoption of membrane-based processes for water treatment and separations. This work investigated various time series models following hold-out validation of experimentally generated water vapor flux and saltwater rejection rates. The membrane properties were optimized by incorporating nanomaterials to induce wetting and porosity and develop correlations between membrane properties and high fluxes. The fine-tuned Autoregressive integrated moving average (ARIMA), Prophet, Exponential Smoothing, and Neural Prophet models were trained on the experimental dataset (N= 434) collected over 36 hours to forecast for 72 hrs. The results demonstrate the suitability of the Exponential Smoothing statistical model for predicting and forecasting membrane performance with the lowest value of root mean square error (RMSE) at 0.006 and mean absolute error (MAE) at 0.007. This is attributed to the intrinsic features of attributed to its non-linear data fitting approach, which employs weighted averages to mitigate nonstationary behavior of data. The modeling approach proposed in this study could be a more efficient alternative to traditional experimental studies, potentially leading to significant cost and time savings in the research and development phase of membrane distillation processes.