Artificial neural network prediction models of heavy metal polluted soil resistivity

Abstract

Heavy metal contaminated soil is on the rise in the whole world with the boost of industrialisation and urbanisation. As such, the heavy metal pollution in lands is of attracted considerable concern to professionals and planners. The detection of contaminant plumes in the subsurface is an important aspect of geotechnical and geo-environmental engineering practice. In this regard, a commonly used method involves conducting electrical resistivity measurements is employed to qualitatively delineate suspected contaminated sites subsurface contamination. However, resistivity measurements alone will lead to some degree of ambiguity in the results, and give only qualitative information about the changes in the chemical composition of the soil-pore fluid. In order to generate much confidence in using the resistivity, the better computational algorithms (viz. artificial neural networks) that should be capable of incorporating the interdependence of several parameters must be employed. Artificial neural network (ANN) presents an oversimplified simulation of the human brain and is accepted as a reliable data modelling tool to capture and represent complex relationships between inputs and outputs. With this in view, efforts were made to develop ANN models that can be employed for predicting electrical resistivity by employing different soil properties such as contaminants, ionic concentrations (n), moisture content (w), porosity (φ) and saturation (S_r). To demonstrate the efficiency of the ANN models, the results obtained were compared with those obtained from experimental investigations and empirical relationships, which are reported in the literature. In addition, the performance indices such as coefficient of determination, root mean square error, mean absolute error and variance were used to assess the performance of the ANN models.

Keywords:

• Heavy metal polluted soil
• resistivity
• laboratory investigations
• predicting model
• ANN

Acknowledgment

The authors would like to express acknowledgment to the editor and anonymous reviewers for their valuable comments and suggestions.

Additional information

Funding

Majority of the work presented in this article was funded by the Key Project of Natural Science Foundation of China (Grant No. 41330641), the "Twelfth Five-Year" National Science and Technology Support Plan (Grant No. 2012BAJ01B02).