【作者】 牛杰；

【导师】 宗周红；

【作者基本信息】 东南大学 ， 土木工程， 2018， 博士

【摘要】 桥梁结构健康监测与安全评估一直是国内外的研究热点和关注焦点之一,由于桥梁结构荷载及环境的不确定性和随机性、监测检测信息的不完备性、计算模拟方法的简化等均会影响到损伤识别的准确度和精度,桥梁的早期损伤诊断已经成为制约桥梁结构健康监测发展的瓶颈之一。本文基于有限元模型修正和模型确认技术,考虑参数不确定性的影响,尝试建立桥梁结构的概率损伤识别的理论与方法,并探索桥梁结构早期损伤诊断的有效途径。主要工作如下:(1)提出基于单元模态应变能变化率的损伤指标,并验证其应用于桥梁结构确定性损伤识别和概率损伤识别的有效性;(2)以灌河大桥为工程背景,研究其不确定性的量化与传递规律;(3)提出基于PSO优化的BP神经网络的有限元模型修正及确认方法,并以灌河大桥为工程背景,完成其有限元模型修正和模型确认;(4)考虑参数不确定性,进行基于模型确认的灌河大桥概率损伤识别;(5)探索基于概率盒理论的斜拉桥损伤识别。得到主要结论如下:(1)考虑参数不确定性后,基于单元模态应变能变化率的损伤指标可较好地识别出简支梁模型单损伤和多损伤的损伤位置及损伤程度,具有应用于实桥结构损伤识别的潜力。(2)建立了灌河大桥的环境温度、车辆荷载及结构响应的不确定性量化模型,温度不确定性可采用傅里叶级数拟合的年趋势函数结合正态分布的日温差形式表达;车辆荷载的车重服从高斯混合分布概率模型,车间距服从对数正态概率模型;主梁结构目标响应基本服从正态分布。不确定性量化结果可用于灌河大桥的不确定性传递、有限元模型确认和概率损伤识别研究。(3)提出了基于PSO优化的BP神经网络的大跨度桥斜拉桥有限元模型修正及确认方法,修正后灌河大桥模型的各阶频率重合度指标均达80%以上,计算频率和实测频率的相对误差最大不超过5.09%,且模型的可信度在95%以上;确认后的有限元模型可以进一步用于灌河大桥的概率损伤识别研究。(4)基于确认后的灌河大桥模型,考虑不确定性的灌河大桥损伤识别表明:竖向一阶损伤指标能较好地识别出主梁损伤位置,其指标基本服从正态分布,且均值可较好地识别出主梁损伤的程度;BP神经网络法的拉索随机损伤识别精度和效率均优于响应面法。(5)滤波方法可以很好地分离环境温度、车辆荷载等不确定性对结构响应不确定性的影响,进而可建立结构响应不确定性的概率盒模型,为灌河大桥基于概率盒的损伤识别研究提供基础。(6)提出了基于概率盒的灌河大桥不确定性损伤识别方法并加以验证,模拟损伤识别结果分析表明:损伤单元的竖向一阶损伤指标概率盒可以充分表达考虑灌河大桥不确定性时的损伤识别效果;基于概率盒的损伤识别方法适用于桥梁结构不确定性损伤识别研究,在实桥结构的损伤识别和损伤预后领域有较好的潜在应用前景。更多还原

【Abstract】 Bridge structural health monitoring and safety evaluation has always been one of the research hotspots and focus of attention at home and abroad.The load and environment of bridge structures have uncertainties and randomness;the information of monitoring and inspection is incomplete;and the computing and simulation methods are simplified;etc.,so that the accuracy and precision of damage identification can been affected.The early damage detection of the bridge structures has become one of the bottlenecks restricting the development of bridge structural health monitoring.Based on the finite element model updating and model validation technology,and considering the influence of parameter uncertainty,the probabilistic damage identification theory and method of bridge structure are established.And an effective way to early damage diagnosis of bridge structure is explored.The main works are as follows:(1)The damage index based on the change rate of element modal strain energy is explored,and its effectiveness in the application of deterministic and probabilistic damage identification for bridge structures is discussed;(2)Taking Guanhe Bridge as the engineering background,the uncertainty quantification and transmission are studied;(3)A finite element model updating and validation method based on PSO optimized BP neural network is proposed,and is used in Guanhe Bridge model updating and validation;(4)The probabilistic damage identification of Guanhe Bridge based on model validation is carried out by considering the uncertainty of parameters;(5)The damage identification of cable-stayed bridge based on probabilistic box theory is explored.The main conclusions are as follows:(1)By considering the parameter uncertainties,the damage index based on the change rate of element modal strain energy can identify the damage location and damage degree of simply supported beam model with single or multiple damages,and has potential for being applied to the damage identification of real bridge structures.(2)The uncertainty quantification models of the environment temperature,vehicle load and structural response of Guanhe Bridge are established,and can be used in the uncertainty transmission,finite element model validation and probabilistic damage identification of Guanhe Bridge.(3)A finite element model updating and validation method of long-span cable-stayed bridge based on PSO optimized BP neural network is proposed.The contact ratios of each order frequency of Guanhe Bridge finite element model after updating are above 80%.The maximum relative error between the calculated frequency and measured frequency is less than 5.09%.And the model credibility is more than 95%.The finite element model after validation can be further applied to the probabilistic damage identification of Guanhe Bridge.(4)Based on the Guanhe Bridge model after validation and considering the uncertainties,damage identification results of Guanhe Bridge indicate that: The vertical first-order damage index can recognize the girder damage location.The index are basically obeying normal distribution,and the mean can better identify the degree of the girder damage;When processing random damage identification of the cables,the accuracy and efficiency of BP neural network are better than the response surface method.(5)The filter method is a good way to separate the uncertainties of the environment temperature and vehicle load which can influence the structural response uncertainties.The probability box models of the structural response uncertainties are established,which can be used to provide the basis of damage identification of Guanhe Bridge based on probability box theory.(6)A damage identification method of Guanhe Bridge based on probabilistic box theory is proposed and verified.The analysis of simulative damage identification shows that: The vertical first-order damage index can express fully the damage identification effectiveness of Guanhe Bridge when considering the uncertainties;The damage identification method based on the probabilistic box theory is suitable for the uncertain damage identification study of bridge structures,and it has a good potential application prospect in the field of damage identification and damage prognosis of real bridge structures.更多还原