The Use of Neural Networks for the Prediction of Zeta Potential of Kaolinite

The sign and the magnitude of the zeta potential must be known for many engineering applications. For clay soils, it is usually negative, but it is strongly dependent on the pore fluid chemistry. However, measurement of zeta potential time is time-consuming and requires special and expensive equipment. In this study, the prediction of zeta potential of kaolinite has been investigated by artificial neural networks (ANNs) and multiple regression analyses (MRAs). To achieve this, ANN and MRA models based on zeta potential measurements of kaolinite in the presence of salt and heavy metal cations at different pH values have been developed. The results of the models were compared with the experimental results. The performance indices, including coefficient of determination, root mean square error, mean absolute error, and variance, were used to assess the performance of the prediction capacity of the models developed in this study. The obtained indices make it clear that the constructed ANN models were able to predict zeta potential of kaolinite quite efficiently and outperformed the MRA models. Results showed that ANN models can be used satisfactorily to predict zeta potential of kaolinite as a rapid inexpensive substitute for laboratory techniques.