【作者】 王锦文；

【导师】 瞿伟廉；

【作者基本信息】 武汉理工大学 ， 结构工程， 2004， 硕士

【摘要】 现代的许多土木结构正不断向大型化、复杂化方向发展，高层、大跨结构不断出现；同时，我国在役建筑物中有许多结构物存在着不同程度的损伤或结构已经进入服役后期，急需损伤鉴定与维修加固。对于重大土木工程结构，容易实现和量测的是结构的动力响应。利用结构的动力响应识别结构的物理参数，并进而评价结构性能、判别结构的损伤以及实现结构的有限元模型修正已成为新世纪土木工程领域的一个前沿研究热点。 要对建筑结构进行动力响应预测、振动控制和结构状态评估及健康监测，首先必须详细了解结构的动力特性。结构的动力特性和结构物理参数直接相关，这些动力特性可以通过有限元分析得到其理论值，也可以由实验模态分析得到其实测值。由于模型描述实际结构的不精确性，结构模态参数的实测值与有限元理论值之间常常存在较大的差异，因而迫切需要解决的问题便是如何对有限元模型进行修正，使得模型模态特性的理论值趋近于实测值。神经网络方法因其具有非线性映射能力强、计算速度快、容错性好等优点，非常适合于结构有限元模型修正。但如何改变实验数据不完备、如何选取结构损伤标识量、如何选定合适的神经网络类别等问题没有得到很好的解决。 由于钢结构的模型误差主要体现对杆件连接刚度而非对杆件的物理尺寸的把握上，故本文以深圳市民中心屋顶网架结构为对象，针对其节点固结系数，选择节点连接刚度为损伤标识量，通过研究结构模态变化与节点连接刚度变化的关系，建立了有限元模型修正的神经网络模型，通过理论模态分析对深圳市民中心屋顶网架结构的有限元模型进行了修正。 本文依据风洞试验数据模拟了静风压的空间分布，依据脉动风谱模拟了网架结构的风荷载时程，并对有限元模型进行了有限元瞬态动力学时程分析。建立了以固结系数表示的三维空间有限元模型，根据单元刚度矩阵中的固结系数的变化，建立固结系数与结构动力特性的关系，应用BP神经网络，进行结构节点固结系数的识别；研究了用于振动测试的加速度传感器的优化配置与网架结构模态参数的实侧方法，通过在风荷载作用下有限元瞬态动力学时程分析检验了有限元模型修正的精确性。 根据本文的研究可以得知，BP神经网络可以很好的用于结构有限元模型的修正;模态分析和时程分析相结合的方法，能够很好的保证有限元模型修正的神经网络模型的可靠性。更多还原

【Abstract】 With the tendency for the structures becoming larger and more complex than before, many new structures spring up in modern time. At the same time, there are various damage in many existing structures as they have been used for several decades needing to be surveyed and strengthened. Therefore, the structural damage diagnosis which makes use of structural dynamic response to identify the physical parameter, sequentially to assess the performance of the structure and to realize the finite element model updating has become the advancing front of civil engineering researches.To forecast structural dynamic response, realize vibration control state evaluation and health monitor, the first thing is to know the structural dynamic characteristic in detail. Structural dynamic characteristic is related to structural physical parameter, and its theoretical value can be gained by finite element model (FEM) analysis, while its practical value can be achieved by experimental modal analysis. Since the structural dynamic characteristic will be changed with its damage, the great difference of structural frequency exists between the theoretical value and the practical value. The problem we are confronted with is how to correct structural FEM to make the theoretical value equal to the practical value. Neural network technique is adapted well to the FEM correction for its strong non-linear mapping ability rapid computation and anti-interference capability. But there are still some problems being open such as selection of neural network, determination of structural damage indicator and incompletion of measurement.As the steel structural model error mainly relied on the control to the member bar’s link stiff and its physical dimension, this paper take the roof structure of Shenzhen citizen center as subject, through its fixity factor, to select the joint link stiff as the damage indicator. The FEM correction of truss structure based neural network can be established with the change of the structural modal and the joint link stiff, thus the FEM for the roof structure of Shenzhen citizen center are corrected by the theoretical modal analysis.In this paper, the wind loading time history of the truss structure is simulated according to the data of wind tunnel experiment and pulsating windspectrum, and the transient dynamic analysis is used on the FEM. Secondly, the space FEM of truss structure that based fixity factor is determined, and the relation between fixity factor and structural dynamic characteristic is established with the change of fixity factor. The fixity factor is identified by BP neural network The FEM correction method of truss structure based neural network technique is developed in this paper, and transient dynamic analysis is used on the FEM. Thirdly, the optimizational layout for acceleration transducer and the actual measured method for the modal parameter of the truss structure are also discussed in this paper and the accuracy of the FEM correction is verified by the finite element transient dynamic temporal analysis,According to the research of this paper, BP neural net can be well adapted to structural FEM correction. The combination of modal analysis and the transient dynamic analysis can guarantee the reliability of the FEM well.更多还原