Statistical regression and modeliing analysis for reverse osmosis desalination process

Experiment is conducted, across different units of a typical desalination plant, to build correlations between inputs/outputs. Steady-state statistical models of reverse osmosis (RO) are developed using stream characteristic data (flow rate, concentration and pH) over a period of time. Data-driven models are useful for deciding real time operational control strategies of desalination plant. The statistical analysis of correlations obtained indicates that permeate characteristics depend on feed stream flow rate by a second degree polynomial. Significance of regression was evaluated based on multivariate ANOVA analysis, on visual standardized residuals distribution and their means for confidence levels of 95% and 99%, clearly validating these models. Sensitivity of parameters is found from interaction and co-relation studies. These models will help in safe operation and control of RO-desalination plant. (C) 2014 Elsevier B.V. All rights reserved.