Bayesian Networks Analysis on Soil cement Mixing Wall

Abstract:

Based on an analysis of the advantages and the disadvantages of event trees (ET), fault trees (FT) and Bayesian networks (BN) respectively, a conclusion is drawn that the three methods should be applied jointly, and the way to construct a BN is improved. This way consists of three steps. First, the FT whose top event and middle events are the output event and the event sequence of the ET respectively is built. Then, the constructed FT is transformed into a BN directly. Finally, with the advantages of the calculation and analysis, the BN is used to make predictions as well as diagnostics. When there is any failure in the non root node event in BN, the corresponding root node events should be controlled by the descending order of their posterior probability values. By applying the proposed method to the soil cement mixing wall in a deep excavation, the failure probability of the top event and the main cause are figured out easily and the practicability of this method is proved.