Empirical correlations between uniaxial compressive strength and density on the basis of lithology: implications from statistical and machine learning assessments

Uniaxial compressive strength (UCS) is a crucial mechanical parameter in the mining, construction, and petroleum industries. However, determination of the UCS is very tough, expensive, time-consuming, and destructive, requires expert workers for sample preparation, and cannot be determined in the field. As a result, prior researchers have employed different indirect proxy tests to estimate the UCS indirectly. Among these indirect tests, determining density (rho) is the cheapest, simplest, non-destructive, and does not require sample preparation; also, rho can easily be determined in the field. Therefore, the correlation between UCS and rho has been rigorously studied in this paper. A total of 800 data points from 26 previous studies were incorporated and lithology based characteristic simple regression (SR) equations for six rock types (pyroclastic, sandstone, shale, carbonate, plutonic and volcanite) have been proposed. UCS can easily be estimated using the proposed regression equations for the six rock types, which will be helpful in geotechnical and geological engineering projects. The lithological control on the correlation for each rock type has also been validated using principal component analysis (PCA) and descriptive statistics. The obtained database was also used to classify the six rocks on the basis of UCS and rho as per International Association of Engineering Geologist (IAEG) classification. Soft computing method of artificial neural network (ANN) was also used to estimate the UCS using two ANN models (ANN-1 and ANN-2). Finally, the estimated values of UCS from SR and ANN models were analysed in 1:1 measured vs. estimated plot and statistically assessed.