Research On The Collapse Risk Predication Of Deep Excavation Of Subway Station Based On SVM

With the rapid development of subway in China, deep excavation has been widely used in the construction of subway. However, the construction of deep excavations has high risk which can bring big loss. Therefore, it’s a hot topic in research to control the risk to improve the safety and reduce the cost during excavation constructions. For example, the collapse of deep excavations is a typical pit accident which can lead to heavy losses of life and property. Therefore, the research on the collapse of deep excavation plays a key role in the deep excavation engineering. This paper is supported by The Chinese Academy of Engineering’s major program "Research on the Safety Risk Management Regulations System of Cilvil Engineering in China","Research on the key technologies for the deep-excavation constructions in the JuZiZhou station of Changsha Metro line2" and focuses on reducing the deep excavation collapse risk. It includes following three aspects:(1) This paper collects148cases of the collapses of deep excavations. It does statistical analysis of the cause of these accidents based on hierarchy of influences in construction accidents, from which, this paper established risk-factor relationship graphs which provided good foundation to forecast deep excavation accidents.(2) This paper makes full use of the statistical analysis results in chapter two, characterizes the main factors of the accidents according to the features of these factors, and then quantifies these main factors to build a training set and test set. SVM is then trained and tested by the training set and the test set. It’s verified that the SVM model provides accurate prediction of deep foundation pit collapses. Actually, this SVM model can be expected tobe used for risk predications in other projects.(3) As a real test, the SVM model is used to predict the collapse risk of the deep excavation in the Middle Xiangjiang Road subway station of line2in Changsha, and it turns out that the probabilistic risk is C-level. It is also graded3by using the method of equivalent risk loss, and the comprehensive risk level is grade Ⅲ. Finally, it obtains the degree of importance of each factor which may lead to the collapse of deep excavation. PRCA risk management model is proposed to control the construction risk, at the same time the performance evaluation is used to evaluate the effectiveness of the risk management. The initial risk is reduced from grade3to grade2after carrying out the risk control measures, and this indicates the feasibility of the risk control measures.