The conclusions including: the Kuznetsov and kansake models are still applicable to the prediction of unit explosive consumption in bench blasting tunnel engineering, and the Kuznetsov model considers the influence of rock mass characteristics, which is more practical than the kansake model. In addition, the Kuznetsov model can predict the average blasting fragment as the basis of the distribution model of blasting fragment. Compared with the lower bench, the upper bench has higher single blasting consumption, higher content of fine particles and smaller fragments. The Kuz-Ram model has a good prediction effect for the small blasting fragments that below the average value. For the large blasting fragments that above the average value, the KCO model has a better prediction effect, and it has more accurate for the prediction of the largest blasting fragment. This paper analyzes the applicable conditions and scope of the prediction model for rock fragment, which provides a basis for the unit explosive consumption and fragment distribution of tunnel blasting. The methods for predicting the specific consumption of explosives and the fragmentation of rock mass during blasting originated from open-pit blasting. However, the stress state of the rock mass in stepped tunnels is different from that in open-pit mines, so it is unknown whether the above prediction methods are applicable to stepped tunnel blasting. Based on statistical data from the bench blasting in the Tianjiangli tunnel, several commonly used calculation methods in open-pit mines were used to predict the specific consumption of explosives and the distribution of fragments, and the predicted results were compared with the actual measurements. The results show that the Kuznetsov model and the Kansake model are still applicable to predicting the specific consumption of explosives in stepped tunnel blasting, and the Kuznetsov model takes into account the influence of rock mass characteristics, making it more practical than the Kansake model. In addition, the Kuznetsov model can predict the average value of the blasting fragmentation and serve as a basis for the fragmentation distribution model. The specific consumption of explosives is higher on the upper bench, resulting in a higher content of fine particles and smaller overall rock mass after blasting, while the specific consumption is lower on the lower step, resulting in a lower content of fine particles and a larger overall rock mass after blasting. The Kuz-Ram model is better in predicting the blasting fragmentation of small rock masses with a block size below the average value, while the KCO model is better for predicting the blasting fragmentation of large rock masses with a block size above the average value, and the KCO model can accurately predict the maximum blasting block. This article analyzes the applicable conditions and scope of the prediction models and provides a basis for the specific consumption and fragmentation distribution in tunnel blasting.

blasting  /  specific consumption of explosives  /  fragment distribution  /  model  /  block size