Assessment of Bohme surface abrasion resistance of natural stones by simple and nonlinear multiple regression methods

Abstract

Natural stones are widely used in the construction sector and the physico-mechanical features are very important for them in terms of areas of use. Abrasion is an important phenomenon for causes to decrease in aesthetical and mechanical properties. The Bohme surface abrasion (BSA) test is the widely used test method to determine abrasion resistance, but it is time-consuming, very difficult, and expensive to apply. Because of such difficulties, this study has been conducted to indirectly determine the BSA value of natural stones by some practical methods. For this purpose, BSA, Point load index (Is50), Shore Schlerescope hardness (SSH), Brinell hardness (HB), Rock impact hardness number (RIHN), Coefficient of rock strength (CRS), Crushability index (CI), Impact strength index (ISI), Water absorption (WA), Unit weight (UW), and Porosity (Pg) properties of a total of up 119 rocks, mainly of Turkish rocks, were determined. The main objective of this article is to develop some estimation models for the BSA of different rocks. Two different methods have been utilized in the modeling of data, including simple regression analysis (SRA), and nonlinear multiple regression analysis (NMRA). In addition to the SRA performed using all data sets, the SRA analysis was deepened by classification according to origin, porosity, and strength. A large number of estimation equations were obtained by NMRA performed by applying three different models. The actual BSA values obtained from the laboratory study and the BSA values obtained from the models analyzed with SPSS software were compared using ANOVA analysis and various graphics. The most successful SRA equations derived using the entire dataset were obtained with HB, Is50, and SSH parameters, respectively. In general, there is a significant correlation between various rock properties and BSA, except CRS and CI. For NMRA models constructed with at least three independent variables, the highest predictive power belongs to the equations using (Is50 / HB / UW), (Is50 / HB / CRS), and (Is50 / UW / WA / Pg) parameters, respectively. Rock abrasion has a complex nature. The results of this study strongly support that it is an accurate approach to seek a solution to this complex problem by using parameters such as hardness, strength, and physical properties in the same model.