The monitoring and forecasting of hazardous rock mass collapse on slopes have always been a critical yet underdeveloped area in geological disaster prevention research. An automatic sensing mechanism was devised for acquiring, computing, and transmitting minor tilt angles and strong vibration accelerations of tension-splitting rock mass. A micro-core-pile geological disasters monitoring sensor has been devised, enabling low-power long-term monitoring. Through on-site monitoring and analysis of tension-splitting rock mass collapses, it was found that these rock masses exhibit a precursor of collapse characterized by accelerated tilt deformation accompanied by an increase in the frequency and amplitude of strong vibrations. It was revealed that there is a significant exponential relationship between the cumulative tilt deformation and the tilt deformation rate during the accelerated tilt phase immediately preceding collapse, and a linear correlation exists between the reciprocal tilt rate and the remaining time before collapse. Subsequently, a ‘reciprocal tilt rate method’ was established for predicting the time to collapse, and an algorithm for real-time application of the prediction model based on MEMS tilt angle sensor data characteristics was developed. These research findings can have a positive promoting effect on the monitoring and early warning of collapse disasters.