Abstract

There are several correlations in the literature that allow an estimate of the soil unit weight for natural soils, but when dealing with materials whose actual specific gravity of solids is outside the range of natural soils for which the correlations were developed, doubts arise, as occurs in the interpretation of tests on mining tailings. Therefore, the present paper aims to evaluate the application of a previously developed approach supported by machine learning techniques for estimating soil specific weights for mining tailings. This approach was developed considering a more comprehensive range of the specific gravity of solids. So, this work relies on a database with results of CPTu tests carried out in different mining tailings deposits from Brazil to estimate specific weights. The values of the specific weights obtained from the machine learning model were compared with literature data, presenting a suitable fit. The research demonstrates that artificial intelligence can contribute positively to the estimation of reliable design parameters and add security to the development of designs of mining tailings containment structures.