Applications of AIoT Techniques and ARIMA Model for Large-scale Landslides Decision System

• Main Authors: Shei-Chen Ho, 何學承
• Other Authors: Miau-Bin Su
• Format: Others
• Language: zh-TW
• Published: 2019
• Online Access: http://ndltd.ncl.edu.tw/handle/7n9r3b

博士 === 國立中興大學 === 土木工程學系所 === 107 === Abstract This study was to build an integrated system based on the research area of a massive landslide area applying AIoT technique (AI, IOT) and ARIMA model. Theory and results of the system architecture can be described in two parts as below: 一、 Establishment of IOT automatic system The establishment of landslide area IOT monitoring instrument and management reference value, including applying Time Domain Reflectometry (TDR) to monitor heterotaxis and inclination of ground surface as well as the basis of slope monitoring system formed by rain gauge and groundwater stage gauge, has become the monitoring instrument providing the basis for critical decisions. In order to integrate the analysis of management reference value in the whole area, real-time operational analysis applying a cloud server will provide the effect of real-time notification and early warning. 二、 ARIMA analysis model In a typical landslide area, underground water level would rise during storm, which would undermine the stability of side slope and may further cause landslide. By analyzing the correlation between rainfall and underground water level based on the time series established using the rainfall and underground water level data from established fully automatic monitoring station in landslide area, applicable rainfall reference value was obtained by further designing storm with reference to existing underground water reference value. Autoregressive Integrated Moving Average (ARIMA) model proposed by Box and Jenkins in 1970 was applied as research methodology, to establish the correlation between rainfall and change in underground water level and rainfall, respectively, and analyze representative conversion function between rainfall and slippage in sliding soil mass, which could be used for predicting change in stability. Potential change in underground water level in the research area was predicted using typhoon event data and the optimal conversion function, which could also be applied in analyzing the stability of side slope, in order to facilitate designing supporting system in landslide area. Research results show that by applying AIoT technique and ARIMA model, early warning of landslide event can be made possible, but management reference value may vary significantly depending on in-situ hydrogeology, engineering renovation and selected event conditions. However, ARIMA model indeed can be applied to understand the correlation between rainfall and underground water level, and can be applied in predicting change in underground water level in landslide area subject to extreme weather with reference to side slope stability analysis and real-monitoring, in order to facilitate the application of landslide decision making system. In the future, when 24-hour quantitative precipitation data become available from the Bureau of Meteorology, analysis model can be incorporated to understand potential change in underground water level and side slope safety coefficient in landslide area in the next 24 hours, so as to make early warning possible.