基于自回归模型和主成分分析的结构损伤识别方法研究

发布时间：2018-06-19 08:51

本文选题：结构损伤识别 + 自回归模型　；参考：《哈尔滨工业大学》2013年硕士论文

【摘要】：土木工程结构在其服役期内，难免会受到自然环境的侵蚀和人为因素的侵害，这些因素会导致结构出现不同程度的性能退化或者损伤。如果不能及时发现和修复这些结构隐患，可能会导致灾难性的后果。因此，结构损伤识别发展成为一个具有重要理论意义和应用前景的领域。本文利用时间序列分析的自回归模型，结合主成分分析对结构进行损伤识别研究，重点研究去除环境因素影响的损伤识别方法。主要内容包括以下几点：综述了环境因素对结构动力响应的影响以及基于结构振动响应的损伤识别理论和方法。介绍了自回归模型的基本理论，比较了定阶准则和参数估计方法的精度，确定了BIC准则结合递推最小二乘估计法能在保证精度的前提达到快速计算，适合用于在线监测。采用基于自回归模型的损伤识别方法：基于自回归系数的欧式距离判别法和基于模型残差的Itakura距离判别法，通过桁架模型试验，验证了两种方法的有效性，指出了这两种方法易受环境因素干扰。提出基于自回归模型与主成分分析的结构损伤识别方法。对结构加速度数据建立自回归模型，利用主成分分析提取并消除环境分量，利用重构的自回归模型系数向量构造损伤指标进行损伤识别。通过桁架模型试验，验证该方法能识别出结构的小损伤，，且不受环境因素干扰。提出了自回归模型与核主成分分析相结合的损伤识别方法。利用核主成分分析提取作为参考样本的自回归系数向量的特征，以特征空间为索引，寻找与待检自回归模型系数向量具有相同特征的样本向量，通过比较两个向量的相似性来识别损伤。试验研究表明，该方法同样能消除环境因素的干扰，准确识别出结构的损伤。利用佛山平胜大桥的实际监测数据，验证上述两种方法的有效性。结果表明核主成分分析比主成分分析更适合于实际工程应用。
[Abstract]:In the service period of the civil engineering structure, it is unavoidable to be eroded by the natural environment and the human factors. These factors will lead to the deterioration or damage of the structure to varying degrees. If the structural hidden danger can not be found and repaired in time, it may lead to disastrous consequences. Therefore, the development of structural damage identification becomes one. In this paper, this paper uses the autoregressive model of time series analysis, combined with principal component analysis to study the damage identification of the structure, and focuses on the damage identification methods to remove the influence of environmental factors. The main contents include the following points:
The influence of environmental factors on the dynamic response of structure and the theory and method of damage identification based on structural vibration response are summarized. The basic theory of autoregressive model is introduced, the accuracy of the order criterion and parameter estimation method is compared, and the BIC criterion combined with the recursive least square estimator can be used to achieve rapid calculation in the premise of ensuring the accuracy. It is suitable for on-line monitoring.
The damage identification method based on autoregressive model is adopted: the Euclidean distance discrimination method based on Autoregressive coefficient and the Itakura distance discrimination based on the model residuals. The validity of the two methods is verified by the truss model test. It is pointed out that these two methods are easily disturbed by the environmental factors.
A structural damage identification method based on autoregressive model and principal component analysis is proposed. The autoregressive model is established for the structural acceleration data, and the environmental components are extracted and eliminated by principal component analysis. The damage identification is constructed by using the coefficient vector of the reconstructed autoregressive model. Through the truss model test, it is verified that the method can be identified. Small damage to the structure and not disturbed by environmental factors.
A method of damage identification combined with autoregressive model and kernel principal component analysis is proposed. The kernel principal component analysis is used to extract the characteristics of autoregressive coefficient vectors as reference samples, and the feature space is used as index to find the same vector with the same characteristics as the coefficient vectors of the unchecked autoregressive model, and the similarity of the two vectors is compared. Experimental results show that the method can also eliminate the interference of environmental factors and accurately identify structural damage.
The effectiveness of the two methods is verified by the actual monitoring data of the Foshan Ping Sheng Bridge. The results show that the nuclear principal component analysis is more suitable for practical engineering than the principal component analysis.
【学位授予单位】：哈尔滨工业大学
【学位级别】：硕士
【学位授予年份】：2013
【分类号】：TU317

【参考文献】