Predicting the amount of toxic metals and metalloids in silt loading using neural networks

Material deposited on road surfaces, called road dust, are known to contain different toxic elements. According to particle size, there are different fractions. Particles with an aerodynamic size less than or equal to 75 µm are called silt loading. As a result of exhaust and non-exhaust emissions from motor vehicles, silt loading deposited on the road surface contains toxic metals, non-metals, and metalloid like Cr, Ni, Zn, Cu, Co, Cd, Pb, and As. Through different pathways, these toxic elements can easily get into the soil, surface and ground water, plants, animals, and the human body. The high risk of contamination and the extent of toxic effects determine the need for their control and health regulation and systematic monitoring. Specific laboratory equipment is used to perform multiple measurements of toxic metal ions. The procedure is heavy and time-consuming due to the difficulties associated with stopping road traffic during sampling in large settlements and the standard elemental analysis technique ICP-MS that is usually applied. The paper proposes a method for predicting the amount of toxic elements in silt loading using artificial intelligence. The paper proposes the use of neural networks, using previously collected experimental data as a training base. The high prediction accuracy that is obtained (As-95.304%, Cd-99.616%, and Pb-98.832%) shows that the proposed prediction could successfully replace the standard elemental analysis.