Prediction of Ground Anchors Load For Artificial Slopes Using Evolutionary AI Model-Case Study of The New Taipei Side Ring highway WuChong Creek Case

• Main Authors: Po-Kun Tsai, 蔡柏坤
• Other Authors: Min-Yuan Cheng
• Format: Others
• Language: zh-TW
• Published: 2017
• Online Access: http://ndltd.ncl.edu.tw/handle/c6przw

碩士 === 國立臺灣科技大學 === 營建工程系 === 105 === Development in the flat area of New Taipei City is approaching saturation. To respond to the rapid development of the city and reduce the gap between urban and rural areas, building traffic facilities on hillsides has become inevitable. However, after land excavation, filling, and preparation, hillsides become susceptible to the influence of various internal and external factors and tend to slide. Therefore, this study investigated artificial slopes secured by ground anchors and retaining structures. Assuring the stability of ground anchors and retaining structures is a crucial topic for administrative and maintenance units, because the results can directly affect the safety of road users. Through a literature review and factor monitoring, this study compiled and analyzed preliminary factors that can affect the load of ground anchors. Correlation analysis was conducted on the preliminary factors and output variables with statistical software. Factors that affect the load of artificial slopes were selected objectively as the input variables of the prediction model, and the load of ground anchors was used as the output variables. Subsequently, multiple evolutionary inference models were used to perform database learning, training, and testing, through which the optimal mapping relationships between input and output variables were identified. The inference model with the highest prediction accuracy was then obtained. The prediction and comparison of various artificial intelligence reasoning models, the results show that the overall accuracy of "SOS-LSSVM" is the best. The mean absolute percent error (MAPE) for the test is 9.53%, this is a precise prediction. Therefor, this model effectively replaces the prediction of traditional subjective experience, it can be used in prediction of artificial slopes without load cell, and so fast and accurate understanding of the ground anchor force, therefor, it can be a reference for subsequent contingency processing.