Solketal production in a solvent-free continuous flow process: scaling from laboratory to bench size

The extended use of biodiesel as a fuel from renewable sources involves an increase in glycerol linked to the transesterification process. Therefore, it would be desirable to find new valorization routes for this by-product. One of the most promising pathways to transform glycerol into added-value products is acetalization with carbonyl compounds. The aim of this work is to scale the process of solketal production from the acetalization reaction in a solvent-less continuous process on the bench scale. The thermodynamic and kinetic parameters were obtained in a batch reactor, using ethanol as the solvent and an ion-exchange resin as the catalyst. The experimental work on this scale was carried out under different operation conditions: acetone/glycerol molar ratio (from 2 to 10), temperature (from 298 to 333 K) and amount of catalyst added (% wt related to the initial glycerol weight from 1 to 5). Three typical mechanisms of heterogeneous catalysts were proposed as reaction rate equations (Langmuir-Hinshelwood, Eley-Rideal and low-range adsorption (LRA) mechanisms) in order to obtain an adequate fitting with the experimental data. The equation based on the LRA catalyst mechanism was selected as the most plausible model according to the minimized root mean square deviation calculated. In order to verify these results on a greater scale, an experimental study was carried out to validate this kinetic model in a bench-scale reactor in a solvent-less continuous flow process. This is the first time in which solketal production was tested on the bench or pilot scale in the absence of ethanol and, therefore, the stability of the operation and the catalyst were evaluated. From the comparison between the estimated and experimental conversion on the bench-scale, the proposed model was able to adequately predict the performance of the acetalization reaction under fixed conditions (from 313 to 333 K), with an error range of approximately 5% and 10% as the maximum error.