Predicting the biosorption capacity of copper by dried Chlorella pyrenoidosa through response surface methodology and artificial neural network models

Abstract

The present work aimed to model the copper biosorption capacity by the dried microalgae Chlorella pyrenoidosa by response surface methodology (RSM) and artificial neural networks (ANN). A concise material characterization and biosorption equilibrium isotherm tests were also carried out. The material characterization showed different functional groups associated with the presence of lipids, proteins and polysaccharides. The Toth equilibrium isotherm model showed that the maximum biosorption capacity was 44.59 ± 1.49 mg/g. The RSM model was built from the Box-Behnken design and the ANN models were built using an automated network search tool, varying different parameters such as transfer functions and number of neurons. The best ANN was a multilayer perceptron neural network with 7 hidden neurons, using the logistic and exponential as the hidden layer and output layer activation functions, respectively. Moreover, this network presented a 71% greater error than the RSM. Despite of that, the experimental validation of the models showed that both RSM and ANN could be used for predictive purposes.