Mining and quarrying activities cause vibrations that depend on blast-induced excavation. Today, competitive factors such as transportation costs and an expanding population result in much of the quarrying of raw materials falling within city limits or at least occurring very close to settled areas. Such situations can lead to various legal restrictions. Buildings and other sensitive structures are often present near limestone quarries, in which blast-induced excavation may be necessary. Because such structures may be affected by blast-induced vibration, their characterization is important in determining the effects on nearby built structures. However, due to differences in geological conditions and blast patterns, the amount of vibration that occurs without blasting in a given zone can be estimated by evaluating the results of vibration measurements from trial blasting. It is important that the vibration values that occur during the ongoing blasting are below the legal limit for areas near settlements. The prediction models are generated by binary regression analysis, based on the data set obtained from the trial blasting. The coefficients of the regression function are specific to the study field. These coefficients characterize the geological structure and other blasting pattern variables. However, for binary regression analysis, different approaches have been accepted for the reduction of variables in the literature. In this study, 22 blasting trials in a limestone quarry were monitored to produce 39 measurements of vibration directed towards a residential area located 396 m west of this quarry. The generated data set was evaluated using prediction models based on binary regression analysis, which have been previously accepted in the literature (USBM, Ambraseys-Hendron, Langefors-Kihlstrom and Indian Standards). From this evaluation, the USBM and Ambraseys-Hendron prediction models are found to have realistic estimates close to the values from the empirical data used in the prediction of blast-induced vibration. Correlation coefficients in the cross-validation of USBM and Ambraseys-Hendron estimation models were found to be 78%. Additionally, the correlation coefficients of cross-validation of the Langefors-Kihlstrom and Indian Standards estimation models were found to be 70%. The Langefors-Kihlstrom and Indian Standards estimation models are quite successful in predicting the charge per delay, which is a parameter included in the dataset, but if the amount of explosives per delay is increased, the results of these estimation models are far off from reflecting the real situation.
Read full abstract