基于多因素数据融合的PC斜拉桥安全评估

摘要

Taking a cable-stayed bridge in Shenyang as an engineering background, referring to the existing theoretical research results of bridge influencing factors and using a combination of numerical simulation and data fusion,the study on the application of multi factor data fusion in the safety assessment of cable-stayed bridge is conducted.Based on the study of existing bridge safety assessment methods,the neural network and the rough set theory are combined to evaluate the safety of PC cable-stayed bridge.The bridge bearing capacity grade is divided into five levels according to the current requirements of the bridge.The Kohonen neural net-work is utilized to cluster the established safety assessment database.The specific evaluation criteria corre-sponding to the five security grades are given.Rough neural algorithm,which integrates rough sets and neural networks,is applied to the data fusion of the database.The attribute reduction function of rough sets is utilized to reduce the input dimension of neural networks.The neural network is trained,studied and predicted.The trained neural network is applied to the safety assessment of cable-stayed bridge, and the 6 specific attribute values of the bridge are directly input.The safety status of the bridge is obtained,and the corresponding man-agement suggestions are given.%以沈阳某斜拉桥为工程背景,借鉴现有的桥梁影响因素理论研究成果,采用数值模拟和数据融合相结合的方法,开展了多因素数据融合在斜拉桥安全评估中的应用研究.在研究现有桥梁安全评估方法的基础上,把神经网络和粗糙集理论结合起来应用于PC斜拉桥的安全评估.将桥梁承载能力等级按照满足目前使用要求的程度划分为五级,利用Kohonen神经网络对建立起的安全评估数据库进行聚类,从而给出了五类安全等级所对应的具体评定指标;将粗糙集和神经网络相集成的粗神经网络算法应用到该数据库的数据融合当中,利用粗糙集的属性约简功能减少神经网络的输入维数,对神经网络进行训练学习和预测;将训练好的神经网络应用到斜拉桥的安全评估中,直接输入该桥对应的6个具体属性指标值,得到该桥目前所处的安全状况,并给出了相应的管理建议.