Multi-Scale Model Of Dam Safety Condition Monitoring Based On Dynamic Bayesian Networks

In order to monitor dam safety condition better, a Dynamic Bayesian Networks (DBN) model is developed to overcome the shortcomings of the ordinary monitoring methods in this paper. Ordinary methods include comprehensive assessment methods and numerical simulation methods. Comprehensive assessment methods have shortcomings such as weight determination, scale difference, variables correlation, etc. In addition, comprehensive assessment methods cannot describe the multi-scale characteristics of monitoring data and dynamic property of large dam. Numerical simulation methods need complex mathematical theory, mechanics methodologies and high performance computer. DBN is a novel model with the consideration of correlations, delay and multi-scale characteristics of the variables such as deformation, seepage, stress and water load and temperature loads, as well as duration at every state. And the new model is also a simpler method with less experience, computational complexity and fewer experiments comparing with numerical simulation. Case study shows that better effect has been achieved in dam safety condition monitoring because the model can take the specific properties of dam into account. The main novel technical contributions of this paper are as follows: applying DBN to establish a multi-scale dynamic model on large engineering conditions monitoring at the first time; providing a new concept of latent durational-state time and analyzing its meaning of dam in the physical sense; using mechanics simulation analysis to verify the new model.