In recent years, micro rebound hardness testing has received increasing attention in
rock engineering due to its simplicity, precision, and cost-effectiveness. The Leeb rebound
hardness (HL), as measured with the Equotip device, is mainly used for the prediction of
uniaxial compressive strength (UCS). A review of the literature reveals that, in establishing
correlations with the UCS, the calculated arithmetic mean HL values are used in the majority
of applications. However, the arithmetic mean is not a robust test statistic since it is
influenced by potential outliers, which is typical for natural materials like rocks. Especially in
the case of relatively small number of observations, the arithmetic mean can only be used in
the presence of normal distribution (outliers excluded). Thus, the presumption of normal
distribution in HL data sets may not always be appropriate. In this study, by employing the
widely used Shapiro-Wilk, kurtosis and skewness test statistics, normality analyses were
performed on HL data sets belonging to 24 different rock samples. On the other hand, the
same rock samples were subjected to UCS testing to check the accuracy of the established
prediction equations. The results of the statistical tests indicated that approximately 58% of
the presently considered HL data sets exhibited non-normal distributions. The UCS prediction
accuracy of the normally distributed HL data sets (R2was significantly higher than those of
non-normally distributed ones (R2Based on this outcome, researchers are encouraged to use
robust test statistics for evaluating HL data to achieve better UCS prediction performances.
Anahtar Kelimeler: Micro Rebound, Leeb Hardness, Equotip, Uniaxial Compressive Strength
|