prediction modelling to enhance anaerobic co-digestion process of ofmsw and bio-flocculated sludge using ann

Artificial neural networks (anns) simulate an anaerobic co-digestion process of organic fraction of municipal solid waste (ofmsw) and bio-flocculated sludge for a mesophilic lab-scale semi-continuous feed reactor. the operational, substrate quality and process control parameters such as organic loading rate, hydraulic retention time, ph, vfa/alkalinity ratio and total solids are input variables and methane yield and volatile solids removal are outputs for ann modelling. the lab-scale experimental results are used to develop a prediction model using fitting application for ann. the network architecture was optimized to achieve accurate predictions, resulting in a 5-19-2 architecture for methane yield and a 5-17-2 architecture for %vsremoval. the training was performed using the bayesian regularization (trainbr) algorithm, leading to high coefficients of determination (r2) of 0.953 and 0.978 for methane yield and %vsremoval, respectively. the results demonstrate the effectiveness of neural network-based modelling in capturing complex relationships within the methane yield process, facilitating accurate prediction of crucial output parameters.