| description abstract | A shield tunnel may suffer various structural defects during its operational period. Different factors can contribute to defects on site, such as cracks, spalling, leakage, or offset, and a rational understanding of the failure path as a function of these defects is not clear at present. This paper aims to identify the cause of defects in shield tunnels using a new data fusion–based dynamic diagnosis. A literature review indicated that three main causes are abnormal load, installation error, and structure decay. These factors were taken into consideration in the proposed method. Both continuous and discrete Bayesian networks were constructed to integrate different types of data and to develop a reliable explanation for the occurrence of the tunnel defects. With in situ real-time monitoring data, the probability distributions for tunnel deformation and internal force were calculated using a continuous Bayesian network. A dynamic diagnosis of the defects was done by updating the monitoring data nodes and defect information in a discrete Bayesian network. A case study of the diagnosis of defects illustrated the method. Tunnel defect occurrence and the effects of multiple defects and changes in monitoring data had different influences on the diagnostic result. Based on the case study, it was concluded that the data fusion diagnosis method provides an efficient method for engineers to find and quantify the main causes of tunnel defects. | |
