Hotelling의 T-square 통계량을 이용한 강우유발 사면붕괴 예측

Abstract

A new technique is presented to detect unstable slope behavior, based on Hotelling`s T2 analysis of pore pressure and water content obtained during flume tests using granitic and gneissic weathered soils. Three sets of pore pressure-water content values were simultaneously obtained during each test, and T2 statistics at the 90.0% and 95.0% confidence levels were calculated based on the correlations between values. The results show that unsuccessful detection of some local failures of the flume slope depended on the sensor position. In the case of global slope failures, anomalous behavior was detected between several hundred and several thousand seconds before the event as T2 statistics exceeded the confidence interval 90%. Hotelling`s T2 analysis provides a single control criterion because it enables correlations between diverse measured values within the same slope; the criterion also includes stepwise criteria for a forecasting and warning system based on confidence levels.