Kinetic Study and Model Assessment for n-Butyl Levulinate Production from Alcoholysis of 5-(Hydroxymethyl)furfural over Amberlite IR-120

Abstract

Alkyl levulinates such as n-butyl levulinate (BL) can play an important role in the fuel sector. Classically, BL is produced from the esterification of levulinic acid, but the butanolysis (or alcoholysis) of sugars requires less stages. The industrialization of this process requires the development of process flow diagrams and thus the knowledge of reaction kinetics and thermodynamics. To the best of our knowledge, there are no kinetic models for the production of BL from sugar butanolysis over heterogeneous catalysts. The kinetics of sugar monomer butanolysis is complex over heterogeneous catalysts due to several side reactions, for example, esterification and humin productions. Hence, developing a kinetic model strategy is vital by primarily focusing on the butanolysis of 5-HMF over Amberlite IR-120. The butanolysis of 5-(hydroxymethyl)­furfural (5-HMF) is the crucial step for the production of BL. The effects of temperature, catalyst, and 5-HMF loadings and catalyst deactivation on the kinetics of the different species were investigated. Then, different kinetic models were developed and tested by cross-validation. Special attention was given to vary widely the reaction temperature, catalyst, and 5-HMF loadings. To increase the accuracy of the estimation stage, independent experiments of levulinic and formic acid esterification, side products of 5-HMF dehydration, were included. It was found that the developed model with more reaction steps and by considering the reaction of humin production from 5-HMF as first order was the most reliable model.

Kinetic modelling: Regression and validation stages, a compulsory tandem for kinetic model assessment

“…Di Meno Di Buchianico et al [ 14 ] found that the reaction pathway displayed in Figure 1 was the most reliable. They assessed different models for the solvolysis of fructose to BL over Amberlite IR‐120 in an autoclave.…”

“…The previous article mentioned that the following species concentrations were used as observables: 5‐HMF, BMF, LA, BL, and BF. [ 14 ] Thus, different levels of noise were added to these species concentrations.…”

“…Table 1 shows the SSE_ i at different levels of noise. For LA concentration, the added noise was lower because in the work of Di Menno Di Bucchianico et al, [ 14 ] the analytical error was lower than the other species. SSE was defined as follows: where is the true concentration of species i and is the noised concentration of species i .…”

“…Some articles have shown the benefits of this approach compared to the holdout method, especially as a tool for model selection. [ 12–14 ]…”

“…For that, synthetic data were generated by simulating a kinetic model for the solvolysis of 5‐HMF over Amberlite IR‐120. [ 14 ] Then, different noise levels were added to the synthetic data, that is, concentrations. We did not remove the outliers because this article aimed to assess the different validation methods even when the noise was high.…”