【作者】 焦峪波；

【导师】 刘寒冰；

【作者基本信息】 吉林大学 ， 道路与铁道工程， 2009， 硕士

【摘要】 在我国中小跨径桥梁中,钢筋混凝土简支梁桥占有很大的比重。这些桥梁在服役过程中,由于环境荷载、疲劳效应、材料老化等影响,不可避免的产生损伤破坏。以往的损伤模型研究大都着眼于简单等直梁单元,而对于多片简支梁的研究甚少。因此,对这类广泛应用的桥型损伤识别的数值模拟不仅可以验证方法的适用性而且对实现该特定桥型实际的损伤识别具有重要的指导价值。本文针对实际工程中广泛应用中的多片简支梁桥的损伤、老化等问题,提出了一种用于其损伤识别的三步法。该方法综合运用频率变化、曲率模态差及神经网络对结构损伤进行识别。结构固有频率的降低可以判定损伤的存在,曲率模态差指标可以进行损伤位置的判定,而神经网络技术可以识别相关单元的损伤程度,从而避免了单纯使用某种方法所固有的缺点。最后将该方法用于一五片简支T梁桥的损伤识别,结果表明该方法可以有效的识别结构的损伤位置及损伤程度,并且具有一定的抗噪性。更多还原

【Abstract】 Bridge is the transport hub of one nation,the safety of it is essential to the normal operation of railway and highway.The structure inevitably has damage because of environmental load, fatigue effect and material aging during its operation.So the evaluation of the safety state and residual life is becoming the urgent research topic.The progress is called damage detection or health monitoring.The damage will cause changes of the structural dynamic characteristics,and the characteristics can be measured from in-situ test,so it is feasible to detect the damage from measuring the changes of the dynamic characteristics.In recent years,this damage detection method based on modal analysis has been widely applied,although it is difficult to conduct damage assessment of complex structures.This method can be classified into several direction depending on the species of modal data such as the damage detection based on frequency,modal shapes, strain mode, frequency response function and other advanced algorithm including neural networks,genetic algorithm and so on.The damage detection results based on traditional calculation method are usually not very satisfactory because of its complexity,lower calculation speed,no convengence and possibly local optimal solution.In order to solve these problems,the domestic and foreign scholars successfully combined the computational intelligence and damage detection,and achieved good effect.In this thesis,firstly we conduct a comprehensive discussion regarding the theory,methods,research status.On this basis,we emphatically discussed the method based on neual networks and dynamic signature,we also conducted the numerical simulation of the bridge structure.Finally we proposed a three-step damage identification method for simply supported T-beam bridge with multiple girders,and the numerical simulation of a bridge with five girders verified its effectiveness.The first chapter:On the one hand,we discussed the research background and significance.On the other hand,the research status,the exiting problems and the development trend of the damage detectin of bridge structure were described.Finally,we conduct a summary review of this thesis.The second chapter:In this chapter, firstly we introduce the basic theory about damage detection and describe the eigenvalue problem of structural dynamics and the finite element modal of damaged structure.Secondly,we classify the methods of damage detection into three kind,including the damage detection based on model updating,modal analysis and computational intelligence.The modal updating method belongs to inverse problem of mathematics,it is statically indeterminate and need increase the number of constraint to solve it,the traditional solution methods include optimal matrix modified method, sensitivity method, characteristic structure configuration method and the mixture method.The modal analysis method is based on the characteristic that the damage of structure will induce the change of modal character,so it is feasible to find some damage fingerprint to conduct the damage detection.The computational intelligence method usually include the neural networks method,the genetic algorithm method and the wavelet analysis method.At the end of this chapter the numerical simulation of a three-span bridge confirm the effectiveness of the method based on modal analysis.The third chapter:At the beginning,the theory of damage detection based on neural netwoks is described.The numerical simulation of RC-beam bridge shows that neural netwoks possess good memory, interpolation and extrapolation ability,it is feasible to use neural networks for damage detection.The results also show that the results depend on the quality of training patterns and if we use neural networks as the only method to conduct damage identification,the number of the training patterns will be huge and unfeasible for practical use of real bridge.The fourth chapter:In this chapter,we propose a three-step damage detection method for simply supported bridge with multiple girders.First,the damage is detected using the sensitive property of frequency to damage.Second,the damage is roughly located from the curvature modal shape difference index.Finally,the location and severity of damage is exactly determined by neural networks.Subsequently,we conduct the numerical simulation of a simply supported T-beam bridge with multiple girders to confim the feasibility of this method.The results are as follows:the first curature modal shape can successfully locate the unique damage successfully, moreover the damage effect of the element near fulcrum is better than the element near mid-span.In the progress of detection for damage with two or three position,we should comprehensively consider the first three curvature modal shapes in order to avoid the erroneous and missing judgment.The detection precision of damage is decreasing with the noise level,so in practical,we should reduce the effect of errors and improve the measurement accuracy of modal shapes.The fifth chapter:conclutions and prospects.we summarize the work of this paper.更多还原