Application of Gaussian Processes on Slope Failure Predication along Taipei County Mountain Road

• Main Authors: Chien-Tsung Li, 李建宗
• Other Authors: Jian-Ye Ching
• Format: Others
• Language: zh-TW
• Published: 2006
• Online Access: http://ndltd.ncl.edu.tw/handle/gbwc85

碩士 === 國立臺灣科技大學 === 營建工程系 === 94 === This study proposes a process for evaluating the potential and probability of slope failure along the mountain road in Taipei County. It is aimed to simulate the procedure of dealing with slope maintenance problem in practice. Two steps are considered including the evaluation of failure potential and failure probability. The probability of failure is determined based on the inherent factors of the slope and the triggering factor of rainfall. The statistical models of Gaussian processes and discriminant analysis together with 89 failed slope cases and 111 not-yet-failed slopes are used to establish the process. Although the former is more complicated than the latter, the accuracy of the results predicted by Gaussian processes is more satisfactory. So it is more appropriated to be used to predict where and at what potential and probability of a slope to fail. The goal of this research is to provide the high potential sections of collapse, give the reference maintained as the road in the relevant unit. The opportunity that collapses happens in the slope when the rainfall, can offer to the relevant unit as the system setting up prewarning and taking precautions against natural calamities. However, it is still difficult to predict when a slope failure will occur so far. More rainfall records are needed to get a better idea about the possible amount of rainfall at a specific slope site during a typhoon.