Machine Learning Techniques Applied to Lignocellulosic Ethanol in Simultaneous Hydrolysis and Fermentation

Abstract

This paper investigates the use of machine learning (ML) techniques to study the effect of different process conditions on ethanol production from lignocellulosic sugarcane bagasse biomass using S. cerevisiae in a simultaneous hydrolysis and fermentation (SHF) process. The effects of temperature, enzyme concentration, biomass load, inoculum size and time were investigated using artificial neural networks, a C5.0 classification tree and random forest algorithms. The optimization of ethanol production was also evaluated. The results clearly depict that ML techniques can be used to evaluate the SHF (R 2 between actual and model predictions higher than 0.90, absolute average deviation lower than 8.1% and RMSE lower than 0.80) and predict optimized conditions which are in close agreement with those found experimentally. Optimal conditions were found to be a temperature of 35 °C, an SHF time of 36 h, enzymatic load of 99.8%, inoculum size of 29.5 g/L and bagasse concentration of 24.9%. The ethanol concentration and volumetric productivity for these conditions were 12.1 g/L and 0.336 g/L.h, respectively.

Production of enzymatic complex from agro-industrial biomass and its application in combustible ethanol

“…Research, Society and Development, v. 10, n. 6, e40410613705, 2021 (CC BY 4.0) | ISSN 2525-3409 | DOI: http://dx.doi.org/10.33448/rsd-v10i6.13705 The sugarcane bagasse used contained about 50% water, 30% cellulose, 7.3% hemicellulose, 11.2% lignin and 1.5% ash, according to previous characterization (method according to Browning (1967). The use of biomass as a carbon source for cultivation of microorganisms in solid media and production of enzymes and biomass for cellulolytic ethanol have been highlighted in previous studies (Fischer et al, 2017;Lopes et al, 2017). The results obtained in the present study point to biomass, roasting and ground coffee residue (RGCR), still little used in technologies for obtaining cellulolytic enzymes and renewable energy.…”

“…SSF cell concentrations were determined in a Neubauer Chamber. Propagation plate methodology (48 h incubation at 40 °C in Czapek medium) (Fischer et al, 2017) was sued to start SSF by estimating the optical density at 600 nm, correlated with the number of colonies obtained. The cellulase activity was determined by filter paper activity (FPA), following IUPAC standard procedures; result was expressed in international Units (Ghose, 1987).…”

“…Efforts have been made to improve processes that aim for the viability of cellulosic ethanol. Thus, enzymatic preparations in situ (Fischer et al, 2017;Lopes et al, 2017;Wang et al, 2020) and the simultaneous hydrolysis and fermentation processes represent a strategy to decrease operating costs, reduce inhibition and contamination, and facilitate recovery of the final product (Öhgren et al, 2007;Leduc et al, 2010;Gu et al, 2014;Guidini et al, 2014;Cuevas et al, 2015).…”

Polish Varieties of Industrial Hemp and Their Utilisation in the Efficient Production of Lignocellulosic Ethanol

“…Their results showed that the SHF process gave a higher concentration of ethanol (8.11 g•L −1 ). Fischer and others [33], in their research, dealt with lignocellulosic biomass and examined the SHF process, obtaining the ethanol concentration of 12.1 g•L −1 . SHF as an alternative process in an industrial bioethanol plant manifests both potential and limitations.…”

Ensemble models of feedstock blend ratios to minimize supply chain risk in bio-based manufacturing

“…To improve prediction performance, in this study, we developed ensemble machine learning (ML) models and applied them on this first dataset obtained from feedstocks available in Florida (dataset 1). Although ML has been widely adopted in many fields of research, including other bio-related fields [15][16][17], there have been very few applications of ML in deconstruction studies involving either a single or multiple feedstocks [18][19][20]. In this study, we tested the application of ML models such as single regression and 2-fold ensemble models integrated with four base learners in linear-weighted regression and nonlinear stochastic gradient boosting models (gbm).…”