Columnar dangerous rock mass is widely developed in many high and steep mountain areas around the world. It often collapses, disintegrates and produces debris flow, which is disastrous. The collapse process of the columnar dangerous rock mass is very similar to the collapse of granular column. In this paper, we report the results of an experimental investigation of the flow induced by the collapse of a column of granular material over a horizontal surface.Two different setups are used, namely, a channelized granular column collapse (i.e., two-dimensional) and an unchannelized granular column collapse (i.e.,three-dimensional), allowing us to compare channelized and unchannelized collapses flows. The experimental data suggest that our experimental findings were markedly different from those reported by previous authors (i.e., include the channelized and unchannelized collapse flows showed differences in energy conversion and dissipation). In channelized collapse flows, the maximum vertical speed appears in the free fall regime, while, the maximum speed in the vertical direction of unchannelized collapse flows appears in the spreading regime. During the whole collapse process, i.e., in channelized and unchannelized collapse flows, the conversion of potential energy and kinetic energy does not occur uniformly, and the maximum kinetic energy of the channelized collapse flows is higher than that of the unchannelized collapse flows, and compared with the unchannelized collapse flows, the dissipation energy in the channelized collapse flows is lower. A series of experiments was performed to predict the behaviour of different granular columns (characterized by different initial aspect ratio (a), varying from 1 to 4).The data obtained from 2D experimental model and 3D experimental model have certain amount of difference, such as the particle runout distance (d1),the maximum central height (h2), and the deposition angle (i.e., β1, β2). These differences show that the 2D experimental model does not fully represent the 3D conditions (i.e., the role of side-walls on the channelized collapse flows characteristic is non-negligible). Accordingly, care must be taken when validating 3D models with 2D experimental data. The movement of the tower dangerous rock masses with collapse failure mode could be evaluated using this channelized and unchannelized granular column experimental results.