Accurate prediction and further dissection of neonicotinoid elimination in the water treatment by CTS@AgBC using multihead attention-based convolutional neural network combined with the time-dependent Cox regression model

Chao Zhang,Xiaoyong Li,Feng Li,Gugong Li,Guoqiang Niu,Hongyu Chen,Guang-Guo Ying,Mingzhi Huang

DOI: https://doi.org/10.1016/j.jhazmat.2021.127029

IF: 13.6

2022-02-01

Journal of Hazardous Materials

Abstract:Imidacloprid (IMI), as the most widely used neonicotinoid insecticide, poses a serious threat to the water ecosystem due to the inefficient elimination in the traditional water treatment. Chitosan (CTS)-stabilized biochar (BC)-supported Ag nanoparticles () are applied to eliminate the IMI in the water treatment effectively. Batch experiments depict that the modification of BC by CTS and Ag nanoparticles remarkably improve its adsorption performance. The pseudo-second-order and Elovich models have good performance in simulating the adsorption processes of and BC. This indicates that the chemical adsorption on real surfaces plays the dominant role in the adsorption of IMI by and BC. In addition, the multihead attention (MHA)-based convolutional neural network (CNN) combined with the time-dependent Cox regression model are initially applied to predict and dissect the adsorption elimination processes of IMI by The proposed MHA–CNN model achieves more accurate concentration prediction of IMI than traditional models. According to influence weights by MHA module, biochar category, pH, and treatment temperature are considered the three dominant environmental variables to determine the IMI elimination processes. This study provides insights into roles of environmental variables in the elimination of IMI by and the accurate prediction of IMI concentration.

environmental sciences,engineering, environmental