Modelling the phytoremediation of formaldehyde from indoor air by Chamaedorea Elegans using artificial intelligence, genetic algorithm and response surface methodology

Abstract

Recently, indoor air quality has become a great concern for human health. Volatile organic compounds (VOCs), are the most common pollutants in the indoor air. Some VOCs such as formaldehyde are toxic and carcinogenic. A lot of researches have proved that potted plants through phytoremediation can help to reduce these pollutants, but modelling on these systems is not developed adequately and without creating the appropriate model, there will be no ability to predict the performance of phytoremediation. In this study, Formaldehyde removal from indoor air by Chamaedorea Elegans was modelled through Artificial Neural Networks (ANN) and Response Surface Methodology (RSM), and optimized by genetic algorithm (GA). Here, the formaldehyde concentration, leaf surface area, light intensity and relative humidity were considered as input variables. The experimental design was conducted using full factorial design method. The RSM results showed that the second-order nonlinear equation was in high accordance with the experiments. A Multilayer Perceptron ANN was developed and network performance was evaluated with several training algorithms, in which, the Levenberg Marquardt (LM) algorithm showed the best performance. Finally, the optimum value of the variables was determined by GA. The results indicated that the ANN model exhibits a higher degree of delicacy and accuracy compared to the RSM.