Prediction of HETP for randomly packed towers operation: integration of aqueous and non-aqueous mass transfer characteristics into one consistent correlation

Height equivalent to a theoretical plate (HETP) calculations, essential for the design of randomly packed distillation columns were extracted from the open literature to generate a working database including over 2350 measurements (only total molar reflux data). The merging of mass transfer characteristics from non-aqueous and aqueous separation experiments has instigated the generation of a consistent correlation predicting HETP. Based on results presented elsewhere for absorption and stripping conditions (Ind. Eng. Chem. Res. 41 (2002) 4911), a set of artificial neural network (ANN) correlations for the gas–liquid interfacial area (aw) and the pure local mass transfer coefficients (kγ, γ=G or L) was proposed with the following dimensionless structures: aw/aT=f(ReL, FrL, EoL, I, χ, K) and Shγ=f(Reγ, Frγ, Scγ, χ). The gas–liquid interfacial area and the pure local mass transfer coefficients were extracted using a reconciliation procedure which combined actually measured interfacial areas with pseudo-interfacial areas inferred from the actually measured volumetric mass transfer coefficients (kLaw, KLaw, kGaw, KGaw—absorption and stripping) and HETP (distillation). The neural network weights of the two aw and kγ correlations were adjusted using a least-squared composite criterion simultaneously over the six mass transfer parameters’ databases. The optimized set of ANN correlations yielded an average absolute relative error (AARE) of 21.3% for the 2357 HETP measurements available. Likewise, the measured interfacial area and volumetric mass transfer coefficients (3770 data) were correlated with an AARE of approximately 26.5%, which undeniably proves the intimate correspondence of absorption and distillation mass transfer characteristics in randomly packed towers. HETP predictions remain as well in accordance with the physical evidence reported in the literature. [Copyright &y& Elsevier]