Use of Artificial Intelligence to Experimental Conditions Identification in the Process of Delignification of Sugarcane Bagasse from Supercritical Carbon Dioxide

Abstract This study evaluated the use of artificial neural networks (ANN) as tools to predict and identify the experimental conditions used in the process of delignification of sugarcane bagasse using supercritical carbon dioxide (ScCO 2 ). The experimental conditions varied from 35 – 100°C to temperature, 75 – 300 bar to pressure and 0 – 100% to ethanol content, used as co-solvent in the extraction. HPLC and determination of insoluble lignin (Klason lignin) analysis were performed to evaluate the efficiency of the process. A database was constructed with the information of the experiments, dividing them into groups of training (70%) and test (30%). The models were obtained using toolbox of the MATLAB R2016b. In the developed neural network, the data obtained by the different techniques of analysis were used as neurons in the input layer and the percentage of insoluble lignin was used as neuron in the output layer. The performance of the neural network was evaluated by the correlation coefficient (R 2 ) and the error indexes (MSE and SSE). The process of sugarcane bagasse delignification using ScCO 2 showed good yields. The model developed from the neural network was satisfactory, since the R 2 value was 99.58% and the error index values were 0.176 to SSE and 0.0147 to MSE.