The artificial neuron network for photocatalytic degradation of acid orange 7 using Cerium Oxide (CeO2) / Wan Nur’ain Awanis Wan Sa’ari ... [et al.]

The presence of dyes in water resources contributes to the accumulation of dyes in fish and other aquatic life. Azo dye toxic compounds mix with bodies and penetrate fish and other aquatic species that are taken up by humans with prolonged health effects. In order to overcome this problem, scientists discovered the photocatalytic process, which is one of the most effective solutions for eliminating organic compounds in wastewater. However, developing an automated dye wastewater treatment plant is very difficult because the condition (e.g. concentration, pH, etc) of dye waste changes severely, depending on the type of the dye. Hence in this research an artificial neural network was developed to
predict the adsorption efficacy of CeO2 photocatalysts. A network was trained by using the experimental data reaction time and pH as the input while the degradation of AO7 as output. The reflective input-response correlation is predicted via a feed forward neural network with hidden layers trained by Lavenberg-Marquardt method. The optimum number of neurons was decided by using trial and error methods. The simulation performance of ANN models was evaluated by using the Root Mean Square Error (RSME) and the coefficient of determination (R2). ANN predicted high accuracy in which R2 is 0.99835 while MSE is around 0.35014.