This paper presents a lumped mass model to describe the run-out and velocity of a series of large flume tests,which was carried out to investigate some propagation mechanisms involved in rapid,dry,dense granular flows and energy transformation when the flows encountered obstacles and reoriented their movement directions.Comparisons between predicted and measured results show that the trend of predicted velocities was basically matched with that of measured ones.Careful scrutiny of test videos reveals that subsequent particles with a higher velocity collided with slowed fronts to make them accelerate. However,this simple model cannot reflect collisions between particles because it treated released materials as a rigid block.Thus,the predicted velocity was somewhat lower than the measured velocity in most cases.When the flow changed its direction due to the variation in slope inclination,the model predicted a decrease in velocity.The predicted decrease in velocity was less than the measured one within a reasonable range of 10% or less.For some cases in which a convexity was introduced,the model also predicted the same trend of velocities as measured in the tests.The velocity increased greatly after the materials took a ballistic trajectory from the vertex of the convexity,and reduced dramatically when they finally made contact with the base of the lower slope.The difference between prediced and measured decrease in velocity was estimated to be about 5% due to the landing.Therefore,the simple lumped mass model based on the energy approach could roughly predict the run-out and velocity of granular flows,although it neglected internal deformation,intergranular collision and friction. This paper presents a lumped mass model to describe the run-out and velocity of a series of large flume tests, which was carried out to investigate some propagation mechanisms involved in rapid, dry, dense granular flows and energy transformation when the flow encountered obstacles and reoriented their movement directions. Comparisons between predicted and measured results show that the trend of predicted velocities was basically matched with that of measured ones. Careful scrutiny of test videos reveals that subsequent particles with a higher velocity collided with slowed fronts to make them accelerate. , this simple model can not reflect collisions between particles because it processed released materials as a rigid block. Thus, the predicted velocity was somewhat lower than the measured velocity in its most cases. When the flow changed its direction due to the variation in slope inclination, the model predicted a decrease in velocity.The predicted decrease in velocity was less than the measu red one within a reasonable range of 10% or less. For some cases in which a convexity was introduced, the model also predicted the same trend of velocities as measured in the tests. velocity is greatly after the materials took a ballistic trajectory from the vertex of the convexity, and reduced dramatically when they finally made contact with the base of the lower slope. The difference between predicated and measured decrease in velocity was estimated to be about 5% due to the landing.Therefore, the simple lumped mass model based on the energy approach could roughly predict the run-out and velocity of granular flows, though it neglected internal deformation, intergranular collision and friction.