Modelling and prediction of cellulose yield using artificial intelligence models and techno-economic analysis of bio-based cellulose isolation from corncobs

Interest in the valorisation of corncobs for cellulose isolation has increased due to rising demand for affordable and sustainable bio-based materials. The study isolated bio-based cellulose using liquid hot water pretreatment and sodium hydroxide treatment. It investigated the modelling and prediction of cellulose yield using artificial neural network (ANN), adaptive neuro-fuzzy inference system (ANFIS), and response surface methodology (RSM) models, coupled with a techno-economic analysis of the cellulose isolation process. The techno-economic analysis was conducted for a simulated bio-based cellulose production plant using the Aspen HYSYS Process Economic Analyser. At 150 °C, 60 min, and 10.0% w/w, the ANN, ANFIS, and RSM models predicted a maximum cellulose yield of 54.2%, 54.0%, and 53.6%, respectively. The predictive performance of the ANN, ANFIS, and RSM models was assessed using statistical indicators, and the results demonstrated that the RSM model, with R² and root-mean-square error (RMSE) values of 0.9934 and 0.3338, respectively, slightly outperformed the ANN and ANFIS models. The total project capital, annual operating cost, and annual product sales were determined as US$22,320,000, US$36,390,000, and US$1,010,000, respectively. Likewise, the process plant’s net present value, internal rate of return, net return rate, payback time, and profitability index were obtained as US$35,460,000, 39.5%, 13.5%, 5.48 years, and 1.135, respectively, at a minimum selling price of US$4.50/kg of bio-based cellulose. The study revealed that the combination of artificial intelligence modelling and techno-economic analysis provided an effective framework for process optimisation, commercialisation, and sustainable bio-based cellulose production. Graphical abstract