改进贝叶斯方法在桥梁状态评估中的应用
Application of improved Bayesian method in bridge state assessment

为了更有效地对桥梁状态进行准确评估，以现有评估体系中的层次分析理论为基础，根据桥梁的不同部位及其功能将其整体进行逐层依次分解，将每个子构件转换为贝叶斯概率网络中的节点，使得桥梁各子构件的状态得分与贝叶斯网络节点的状态概率一一对应，进而确定某一时刻桥梁的整体状态；同时，结合时变效应，考虑贝叶斯方法中先验信息对后验信息的影响，对桥梁不同时刻的状态信息进行更新、传递与评估，按照时间序列建立完整的评估体系，提出一种基于改进贝叶斯理论的桥梁状态评估方法，并通过某座桥梁的评估实例，分析该桥不同年份级别评定的状态概率分布。研究结果表明：采用改进贝叶斯方法训练后，桥梁不同年份对应的状态级别概率均有明显提高，评估准确度从50%~60%提高到95%左右；另通过对比桥梁不同年份的状态级别概率分布曲线可知，随着运营时间的增加，曲线的主要包络区域级别呈现下降趋势。反映出桥梁在运营过程中产生累积损伤，对应的状态等级呈现下降趋势。与此同时，曲线最高点对应的概率呈现增大趋势。也反映出采用这一方法进行桥梁评估更加直观、可靠，其评估结果更符合实际。
To more effectively make an accurate assessment of bridge state, based on the theory of analytic hierarchy process (AHP) in existing system, the whole bridge was decomposed hierarchically according to its different parts and functions, and each sub-component was converted into a node in Bayesian probability network. Therefore, the state score of each sub-component was set to be corresponding to the state probability of Bayesian network node, and then the whole state of bridge at a certain time was determined. At the same time, combined with time-varying effect and based on Bayesian inference which considered the influence of the prior information, a complete assessment system was established in chronological order by updating, transmitting and evaluating the state information of bridge at different moments, and a bridge state evaluation method based on Bayesian theory was proposed. Through the evaluation example of a bridge, the state probability distribution of bridge assessed in different years was analyzed. The results show that the probability of the corresponding state level in different years has been improved significantly after training with the improved Bayesian method, and the assessment accuracy has been increased from 50% or 60% to 95%; by comparing the probability distribution curves of bridge state level in different years, the main enveloping region level of the curve shows a decreasing trend with the increase of operating time, indicating that a cumulative damage is produced during the operation of bridge, and the corresponding state level shows a descending trend. Meanwhile, the probability corresponding to the highest point of the curve presents an increasing trend, indicating that the use of this method to evaluate bridge is more intuitive and reliable, and the evaluation results are more in line with the actual situation