摘要
为克服既有安全评估方法时效性差、主观性强、难以量化等缺点,解决已有人工神经网络缺乏可解释性和物理意义、图解可视化弱等问题,以贝叶斯推理与效用计算为依据,提出一种结合贝叶斯决策网络(Bayesian decision network,简称BDN)的桁架结构安全评估方法。首先,通过分析桁架传力路径,将桁架体系拆分为多个子结构,再以子结构为不确定性节点,同时增加决策节点和效用节点,通过有向弧连接以完成BDN网络拓扑定义;其次,计算节点变量的条件概率表,确定节点状态参数并计算概率参数;然后,以影响因素为自变量、以效用值为因变量建立函数关系,完成对效用函数的计算;最后,以一榀试验钢桁架为验证对象,通过绘制体系状态评分图,完成对钢桁架当前状态的决策和评估。结果表明:结合BDN可以刻画钢桁架在某荷载组合作用下的状态变化过程,桁架的安全和失效评分曲线交点即为某个状态的临界荷载;通过推理、决策得到的桁架体系状态概率与试验现象基本吻合,可为类似结构体系的安全评定提供参考。
Existing safety evaluation methods for civil structures suffer from poor timeliness,strong subjectivity,and difficulty in quantification.Current arificial neural network has been criticized for lacking interpretability and physical meaning,as well as weak graphical visualization capabilities.To address these issues,a safety evaluation method for truss structure is proposed using Bayesian decision network(BDN),which rely on Bayesian reasoning and utility calculation.First,based on an analysis of the load transfer mechanisms,the truss system is divided into several substructures,each is defined as a chance node.Meanwhile,decision nodes and utility nodes are added into the network,and all the nodes are connected with directed edges to define the BDN topology.Second,the nodal conditional probability tables are subsequently computed without assuming specific load distributions,and the tables are used to determine the probability and state parameters of the nodal variables.Then,the utility function is established on the influence factors and the utility values,which are the independent and dependent variables for the function,respectively.Finally,the state score curve is generated to support decision-making and assess the current condition of the structure.The results show that BDN can describe the state variation of the steel truss under a given load combination.The critical load of a specific state is identified at the intersection point of the safety score curve and the failure score curve.Based on reasoning and decision,the inferred state probability of the truss system is consistent with the experimental observations.Thus,the proposed method provides a reference for safety evaluation of similar structures.
作者
郑金林
方圣恩
ZHENG Jinlin;FANG Sheng'en(School of Civil Engineering,Fuzhou University Fuzhou,350108,China;National&Local Joint Engineering Research Center for Seismic and Disaster Informatization of Civil Engineering,Fuzhou University Fuzhou,350108,China)
出处
《振动.测试与诊断》
北大核心
2025年第6期1136-1142,1273,共8页
Journal of Vibration,Measurement & Diagnosis
基金
国家自然科学基金资助项目(52178276)。
关键词
安全评估
贝叶斯决策网络
桁架结构
决策节点
效用函数
safety evaluation
Bayesian decision network
truss structure
decision nodes
utility function
