Artificial Neural Network Modeling of Palm Oil Mill Effluent (POME) Treatment Using Plant-Based Bio-coagulant and Bio-flocculant

Increasing palm oil production in the country has led to a large amount of palm oil mill effluent (POME) generated in the mill. Treatment of POME using plant-based materials like fenugreek seeds and aloe vera gel seems to be a good option to treat POME. Fenugreek seeds and aloe vera gel were utilized as a natural coagulant and flocculant, respectively. The 3 input factors of the process, namely pH, coagulant, and flocculant dosage were used as the inputs, while turbidity, total suspension solids (TSS), and chemical oxygen demand (COD) were used as the outputs of the model. 11 training algorithms were used for the development of backpropagation feedforward neural network (BFNN). Conjugate gradient backpropagation with Polak–Ribiére (CGP) learning rate backpropagation training algorithm is optimized to be the best training algorithm, having 17, 5, and 20 hidden neurons for the percentage removal of turbidity, TSS, and COD, respectively, as the optimal values. The error factors obtained by the CGP training algorithm are as follows: mean squared error (MSE) values of 49.42, 55.23, and 20.94, mean absolute error (MAE) values of 4.89, 5.01, and 3.21, and mean absolute percentage error (MAPE) values of 0.13, 0.12, and 0.22 for turbidity, TSS, and COD, respectively. The overall coefficient of determination, R2, values as follows for turbidity, TSS, and COD, respectively: 0.9886, 0.9832, 0.9667.