Pergamon PIT: S0273-1223(98)OO589-7 Wal. SeL Vol 38. No.6. pp. 295-302. 1998. lAWQ C 1998 Published by Elsevier Science Lcd. Printed In Orul Britaln. All rights raened 0273-1223198 $19'00 + 0-00 IMPROVED ONDA CORRELATIONS FOR MASS TRANSFER IN PACKED TOWERS Y. Djebbar* and R. M. Narbaitz** • S&:DD. Greater Vancouver Regional District, 4330 Kingsway, BBY. BC, Cmuu14, V5H4G8 •• University o/Onawa, 161. Louis Pasteur. Onawa, ON, CCUtDda, KIN 6N5 ABSTRACT A comprehensive database for VOC stripping in packed lowers WII gathered from 1.5 field and pilot-scalo studies. This database is used to develop I new correlation based on the Onda model. The development of this new correlation addressed the shoncomings of the Onda correlation which were identified in previous studies. The average error in the new correlation was less than 26% for both the development and the validation data. = 1998 Published by Elsevier Science Ltd. All rights reserved KEYWORDS Volatile organic compounds; packed tower; air stripping; Onda correlation; random structured and packing. INTRODUCflON An estimate of the mass transfer coefficient, KLa. is necessary for the designer to determine the dimensions of a packed tower to strip volatile organic compounds (V0Cs) from contaminated groundwater. This estimate can be obtained using a pilot-scale study which is expensive. A common alternative approach is to use mass transfer correlations. Tens of correlations have been developed during the past fUty years (Au• Yeung and Ponter, 1982), however, only a few of them have been tested for conditions that are of environmental concern. Among existing models, there is agreement in the recent literature (Staudinger et al., 1990; Roberts et al., 1985; Lamarche and Droste, 1989, Djebbar and Narbaitz, 1995) that the Onda model (1968) is the best KLa correlation for air stripping applications. The literature shows that this model yields satisfactory results for lab data, i.e., conditions that are si.milar to those under which this model has been developed. However, 30 to SO percent deviations predictions are. often reported in full-scale applications (Lenzo et al.. 1990; Staudinger et al., 1990; NarblUtz,. The lack of good predictive ability is due to: a) the extrapolative nature of b) the in the databases used to develop the correlations; c) the lUDltatlons of the parametric regression techmques used to generate these models; and d) the changing nature of mass transfer from one set of operating to which are not well understood and are poorly mUDlcked by current parametric models (DJebbar and NarbaItz, 1995; Bravo et al., 1992). 29.5