ARTIFICIAL NEURAL NETWORK-BASED MODELLING OF METHYLENE BLUE PHOTODEGRADATION USING HEXAGONAL PRISM-SHAPED NIOBIUM OXIDE AS A HETEROGENEOUS CATALYST

Abstract

Photodegradation is great and green method for taking away stubborn organic bad chemicals from water. This work uses a computer program that learns from examples to guess and make better the Methylene Blue (MB) breaking down by light using hexagonal prism-shaped Nb₂O₅ and Nb₂O₅.nH₂O catalysts. All 136 test points were used, with 82 for Nb₂O₅ and 54 for Nb₂O₅.nH₂O to make links between way things are done and how well they break down. The ANN model was trained using 17 neurons for Nb₂O₅ and 13 neurons for Nb₂O₅.nH₂O, obtained mean squared errors (MSE) of 22.164 and 4.61; and regression coefficients (R²) of 0.98 and 0.99, respectively. Validation proved to be very accurate with relative errors less than 10%. Nb₂O₅ showed a linear decrement in degradation efficiency which implied controlled reaction kinetics; Nb₂O₅.nH₂O showed an exponential decrement which implied surface saturation effects. Additionally, dye concentration influenced degradation efficiency, with excessive dye hindering catalyst activation. Analysis of UVA and UVC radiation sources revealed that UVA facilitated faster degradation, whereas UVC provided sustained photocatalytic activity over extended durations. These findings highlight ANN-based modelling as an effective tool for optimizing photodegradation conditions.