基于统计模式识别的空间钢结构损伤预警
发布时间：2018-08-30 12:00
【摘要】： 大跨空间钢结构多用于体育馆、会展中心等城市标志性建筑,其结构体型庞大,造型独特,形式新颖。为了保证结构的安全性,需要对大跨空间钢结构进行健康监测和损伤预警。本文主要研究如何由加速度时程响应提取结构损伤敏感指标并分离环境变化和监测信号噪声的影响,统计判别结构的工作状态,主要研究内容如下: 研究了由加速度时程响应建立ARMA模型。理论分析了ARMA模型的适合阶次,阐明了ARMA模型的阶次由结构参与振动的振型数决定。提出了基于遗传算法的ARMA定阶方法,首先根据AR模型的阶次确定ARMA模型阶次的范围,然后应用遗传算法搜索ARMA模型的阶次。K8型网壳的数值算例表明,所提方法减少定阶运算量,能够实现ARMA模型的快速定阶。 分别研究了以AR系数和脉冲响应为结构损伤敏感指标,表征结构的工作状态,并提出了基于主成分分析和假设检验的统计判别方法。简支梁和K8型网壳的数值算例分别表明,以AR系数为结构损伤敏感指标,可以预警简支梁的微小损伤和网壳结构的比较严重的损伤,该指标的敏感性随着结构复杂程度增加有所降低;以脉冲响应为结构损伤敏感指标,能够较好的预警网壳结构的微小损伤。 基于“水立方”结构的数值模型,研究了脉冲响应作为结构损伤敏感指标对监测信号噪声的鲁棒性。数值算例表明,该指标能够预警“水立方”结构一定程度的损伤,并对信号噪声有较好的鲁棒性。提出采用支持向量机回归得到脉冲响应以温度和风速为自变量的函数关系,以支持向量机的训练残差建立统计过程控制图,分离环境因素对结构工作状态的影响。
[Abstract]:Large span space steel structures are used in iconic buildings such as gymnasium and exhibition center. Their structures are large, unique in shape and novel in form. In order to ensure the safety of the structure, it is necessary to carry out health monitoring and damage warning for long span space steel structures. This paper mainly studies how to extract structural damage sensitive index from acceleration time history response, separate the influence of environment change and monitoring signal noise, and judge the working state of structure statistically. The main contents are as follows: the ARMA model is established from the acceleration time history response. The suitable order of ARMA model is analyzed theoretically. It is clarified that the order of ARMA model is determined by the number of modes in which the structure takes part in the vibration. The ARMA order determination method based on genetic algorithm is proposed. Firstly, the range of ARMA model order is determined according to the order of AR model, and then the numerical example of searching ARMA model order. K8 reticulated shell shows that the proposed method reduces the number of order determination operations. The fast order determination of ARMA model can be realized. AR coefficient and impulse response are used as damage sensitive indexes to characterize the working state of the structure, and a statistical discriminant method based on principal component analysis and hypothesis test is proposed. Numerical examples of simply supported beam and K8 type reticulated shell show that using AR coefficient as the sensitive index of structural damage, the small damage of simply supported beam and the severe damage of reticulated shell structure can be forewarned. The sensitivity of the index decreases with the increase of structural complexity, and the pulse response is used as the sensitive index of structural damage, which can be used to predict the small damage of latticed shell structure. Based on the numerical model of "water cube" structure, the robustness of impulse response as a structural damage sensitive index to monitoring signal noise is studied. Numerical examples show that the index can warn the damage of "water cube" structure to a certain extent and has good robustness to signal noise. The function relation of impulse response with temperature and wind speed as independent variables is obtained by using support vector machine regression. The statistical process control chart is established by training residual of support vector machine, and the influence of environmental factors on the working state of structure is separated.
【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】：2010
【分类号】：TU391;TU312.3
本文编号：2213010
[Abstract]:Large span space steel structures are used in iconic buildings such as gymnasium and exhibition center. Their structures are large, unique in shape and novel in form. In order to ensure the safety of the structure, it is necessary to carry out health monitoring and damage warning for long span space steel structures. This paper mainly studies how to extract structural damage sensitive index from acceleration time history response, separate the influence of environment change and monitoring signal noise, and judge the working state of structure statistically. The main contents are as follows: the ARMA model is established from the acceleration time history response. The suitable order of ARMA model is analyzed theoretically. It is clarified that the order of ARMA model is determined by the number of modes in which the structure takes part in the vibration. The ARMA order determination method based on genetic algorithm is proposed. Firstly, the range of ARMA model order is determined according to the order of AR model, and then the numerical example of searching ARMA model order. K8 reticulated shell shows that the proposed method reduces the number of order determination operations. The fast order determination of ARMA model can be realized. AR coefficient and impulse response are used as damage sensitive indexes to characterize the working state of the structure, and a statistical discriminant method based on principal component analysis and hypothesis test is proposed. Numerical examples of simply supported beam and K8 type reticulated shell show that using AR coefficient as the sensitive index of structural damage, the small damage of simply supported beam and the severe damage of reticulated shell structure can be forewarned. The sensitivity of the index decreases with the increase of structural complexity, and the pulse response is used as the sensitive index of structural damage, which can be used to predict the small damage of latticed shell structure. Based on the numerical model of "water cube" structure, the robustness of impulse response as a structural damage sensitive index to monitoring signal noise is studied. Numerical examples show that the index can warn the damage of "water cube" structure to a certain extent and has good robustness to signal noise. The function relation of impulse response with temperature and wind speed as independent variables is obtained by using support vector machine regression. The statistical process control chart is established by training residual of support vector machine, and the influence of environmental factors on the working state of structure is separated.
【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】：2010
【分类号】：TU391;TU312.3
【引证文献】
相关硕士学位论文 前1条
1 黄振兴;震后桥梁结构时频域损伤诊断研究[D];西南交通大学;2012年
，本文编号：2213010
本文链接：https://www.wllwen.com/guanlilunwen/huizhanguanlilunwen/2213010.html
