In order to assess the capacity of ground support systems when submitted to dynamic loading, simulated rockburst tests utilizing blasting have been performed for many years in different countries with limited success. In general, the blasts need to be carefully designed in order to reach the goal; however, different blast layouts (e.g. blasthole angle, burden) have been used based on researcher’s experience without conducting detailed analyses, the exception being a field test by CSIR. Recently, field trials have been conducted at the LKAB Kiirunavaara underground mine with some unexpected results which show that either the whole tested panel was destroyed or only a few fractures were formed without any ejections being observed. The aim of this paper is to investigate the failure mechanism in the simulated rockburst tests and improve the blast design by back-analyzing the test results using a coupled numerical modeling technique. The blast was simulated by using finite element method (LS-DYNA) and the dynamic interaction between the blasting generated waves and the opening was simulated by using discrete element modeling (UDEC) with the dynamic input from LS-DYNA. The numerical modeling showed that blasting can create both radial fractures radiating from the blasthole and fractures parallel or sub-parallel to the surface of the tested panel caused by reflected tensile stress waves. By comparing the results of the numerical modeling with the measured data, it is shown that the collapse failure was mainly controlled by the creation of a cone-shaped area formed by radial fractures and the burden seems to be a critical factor. In order to obtain fractures caused by reflected tensile stress waves and reduce blasting induced radial fractures, 2 parallel blastholes are suggested with larger burden (> 5 m) for future tests. Furthermore, the limitation of the current numerical modeling has also been discussed. The coupled numerical technique has shown its advantage when simulating blasting as well as interaction between waves and opening and it can thus be used as a tool for extrapolating results from simulated rockburst experiments if detailed geological structure and ground support system can be incorporated in the model and the model can be well calibrated.