Abstract: Rock collapse has been a hot issue in the study of geological hazards for many years, and it is difficult to prevent because of its sudden disintegration, which is a serious threaten to human life and property safety. The rock collapse is caused by the dynamic failure of the system instability, so it can be more effective to apply the dynamic monitoring index in early warning. By introducing time-domain dynamic monitoring indicators, the whole process of rock collapse is monitored in real time. Through vibration amplitude, kurtosis index and other indexes, the failure precursor in the detachment phase is analyzed. As the early warning method based on detachment precursor recognition has better timeliness in the early warning of rock collapse, it can realize the early warning of rock block collapse 55 s in advance. The experimental results show that the time-domain dynamic index monitoring can identify the obvious incompatible dynamic response before rock failure, and the variation coefficient has obvious advantages in identifying the oscillation feature. The early warning of collapse disaster can be realized by identifying this oscillation feature. Furthermore, the vibration velocity of rock mass before the collapse is 2.1 times that in the stable phase, and the large impact energy of rock mass at the time of failure is one of the main reasons for the high-speed rockslide. By analyzing the kurtosis index and other time-domain dynamic indexes, the reasonable identification of the failure precursors can be achieved. The study provides new data support and enlightenment for the early warning of rock collapse, the mechanism of rock collapse and the rock movement characteristics after failure.