Distribution models for nitrophenols in a liquid-liquid system

In this study, a data-driven approach was implemented to provide prediction models for Dj of 2,4-dinitrophenol (DNP) and of 2,4,6-trinitrophenol (TNP) in a biphasic liquid system with a composition representative of the industrial processes. In the first step, screening tests were performed to identify the main variables influencing the experimental equilibrium weight fractions of nitrophenols in the aqueous phase wj,eA. Subsequently two independent data sets were built for development and external validation of prediction multivariate linear regression (MLR) models, at 30°C. The fitting results (R2 and Rad2⩾0.90) and the prediction results (Rpred,DNP2=0.931,Rpred,TNP2=0.908) confirmed the quality of the wj,eA models. Statistical significant predictive MLR models were also developed for Dj (which is related with wj,eA), at 30°C, with DNP evidencing a higher affinity for the organic phase (i.e. DDNP≈2DTNP).