Gueguim Kana, Evariste Bosco.Rorke, Daneal Carmine Solange.2020-03-312020-03-3120172017https://researchspace.ukzn.ac.za/handle/10413/17294Masters Degree, University of KwaZulu-Natal, Pietermaritzburg.The limitations of first generation biofuels have prompted the quest for alternative energy sources. Approximately 60 million tonnes of sorghum are generated each year, with 90% being lignocellulosic waste, which is an ideal feedstock for biofuel production. The recalcitrance of lignocellulose often demands harsh pre-treatment conditions and results in the generation of fermentation inhibitors, negatively impacting process yields and economics. In this study, an artificially intelligent model to predict the profile of reducing sugars and all major volatile compounds from microwave assisted chemical pre-treatment of waste sorghum leaves (SL) was developed and validated. The pre-treated substrate was assessed for bioethanol production using Saccharomyces cerevisiae. Monod and modified Gompertz models were generated and the kinetic coefficients were compared with previous studies on different substrates. To develop the Artificial Neural Network (ANN) model, a total of 58 pre-treatment process conditions with varying parameters were experimentally assessed for reducing sugar (RS) and volatile compound production. The pre-treatment input variables consisted of acid concentration, alkali concentration, microwave duration, microwave intensity and solid-to-liquid ratio (S:L). Response Surface Methodology (RSM) was used to optimise RS production from microwave assisted acid pre-treatment of sorghum leaves, giving a coefficient of determination (R2 ) of 0.76, resulting in an optimal yield of 2.74 g RS/g SL. A multilayer perceptron ANN model was used, with a topology of 5-13-13-21. The model was trained using the backpropagation algorithm to minimise the net error value on validation. The model was validated on experimental data and R2 values of up to 0.93 were obtained. The developed model was used to predict the profile of inhibitory compounds under various pre-treatment conditions. Some of these inhibitory compounds were: acetic acid (0-186.26 ng/g SL), furfural (0-240.80 ng/g SL), 5-hydroxy methyl furfural (HMF) (0-19.20 ng/g SL) and phenol (0-7.76 ng/g SL). The developed ANN model was further subjected to knowledge extraction. Findings revealed that furfural and phenol generation during substrate pre-treatment exhibited high sensitivity to acid- and alkali concentration and S:L ratio, while phenol production showed high sensitivity to microwave duration and intensity. Furfural generation during pre-treatment of waste SL was majorly dependent on acid concentration and fit a dosage-response relationship model with a 2.5% HCl threshold. VI The pre-treated sorghum leaves were enzymatically hydrolysed and subsequently assessed for yeast growth and bioethanol production using Saccharomyces cerevisiae BY4743. Kinetic modelling was carried out using the Monod and the modified Gompertz models. Fermentations were carried out with varied initial substrate concentrations (12.5-30.0 g/L). The Monod model fitted well to the experimental data, exhibiting an R2 of 0.95. The model coefficients of maximum specific growth rate (μmax) and Monod constant (Ks) were 0.176 h-1 and 10.11 g/L respectively. Bioethanol production data fitted the modified Gompertz model with an R2 of 0.98. A bioethanol production lag time of 6.31 hours, maximum ethanol production rate of 0.52 g/L/h and a maximum potential bioethanol concentration of 17.15 g/L were obtained. These findings demonstrated that waste SL, commonly considered as post-harvest waste, contain sufficient fermentable sugar which can be recovered from appropriate HCl-based pre-treatment, for use as a low cost energy source for biofuel production. The extracted knowledge from the developed ANN model revealed significant non-linearities between the pre-treatment input conditions and generation of volatile compounds from waste SL. This predictive tool reduces analytical costs in bioprocess development through virtual analytical instrumentation. Monod and modified Gompertz coefficients demonstrated the potential of utilising sorghum leaves for bioethanol production, by providing data for early stage knowledge of the production efficiency of bioethanol production from waste SL. The generated kinetic knowledge of S. cerevisiae growth on waste SL and bioethanol formation in this study is of high importance for process optimisation and scale up towards the commercialisation of this fuel.enSorghum leaves.Lignocellulosic pre-treatment.Fermentation inhibitors.Bioethanol production.Kinetic modelling.The potential of waste sorghum (sorghum bicolor) leaves for bioethanol process development using Saccharomyces cerevisiae BY4743.Thesis