THEORY AND METHOD OF MONITORING AND EARLY WARNING FOR SUDDEN LOESS LANDSLIDE—A CASE STUDY AT HEIFANGTAI TERRACE

Most of the loess landslides induced by irrigation own obvious sudden characteristics. The deformation and displacement during slope failure process are small and the time of duration is short, which is of great risk. Due to such loess landslides undergo a short time in accelerated deformation stage, it is difficult for traditional monitoring methods, such as GNSS system and crack gauge, to obtain complete monitoring data in accelerated deformation stage and to predict the sudden landslide occurrence. With respect to this problem, a self-adaptive frequency conversion acquisition monitoring method is designed to monitor the deformation of sudden loess landslides, which adjust automatically the frequency sampling according to the speed of landslide deformation. To meet the needs for risk mitigation and management of slope sudden failure, it is of practical significance to develop a self-adaptive frequency conversion acquisition monitoring method and establish a real-time automatic early warning system. The new artificial intelligence by the authors' institute can obtain entire monitoring data in accelerated deformation stage and to predict the sudden failure occurrence time. Taking deformation rate threshold and the improved tangent angle as the early warning parameters of comprehensive warning model, a four-level early warning criterion is established. The real-time automatic early warning of the landslide is realized through the self-developed "real-time monitoring and early warning system of geological hazards". The early warning information is released in the local group defense information platform, which provides a direct gauge for disaster prevention and emergency avoidance. Since 2017, it has been successfully warned six times of loess slope sudden failure on the Heifangtai terrace, which avoided heavy casualties and achieved remarkable disaster prevention and mitigation effect.