【作者】 陈涛；

【导师】 黄焱；

【作者基本信息】 天津大学 ， 船舶与海洋工程， 2019， 硕士

【摘要】 海洋平台作为海洋油气资源开发的基础性设施,是海上生产作业和生活的基地。为了保障海洋平台结构的安全服役,避免重大的恶性事故发生,必须对海洋平台结构在服役期内进行必要的检测和安全评估,海洋平台结构模态特征实时提取就是以此目标出发而衍生出的一项重要监测手段。目前,常用的模态特征提取方法包括峰值拾取法、自然激励技术(Natural Excitation Technique)、ITD(Ibrahim Time Domain)法、随机子空间方法(Stochastic Subspace Identification)等。其中,随机子空间方法(SSI)因其具有计算模型参数少,无迭代运算,收敛性好,运算快捷,鲁棒性好等特点,受到广泛关注。本文旨在发展一套基于海洋平台的协方差驱动随机子空间模态实时提取方法。首先,基于某大型导管架式海洋平台有限元模型,验证了随机子空间方法理论与算法的可行性,并分别对该方法涉及的计算参数:系统阶数、Hankel矩阵维度以及计算数据长度进行讨论分析,初步筛选出最佳参数配比。其次,针对传统随机子空间算法在有限测点下无法反映结构整体特征以及对现实结构不敏感等问题,提出了一种新的Hankel元素重构方式,算法构建与实际空间结构保持的高度相关性,既提高了运算的准确度,消除了部分虚假模态,又极大地的提升了计算效率。然后,针对传统随机子空间方法在计算大型复杂结构时存在虚假模态过多的现象,建立了一种基于信号累积特征的虚假模态辨识方法与剔除准则,有效识别出数模结构的前三阶真实模态。既有效的剔除了虚假模态,也极大的增加了响应数据的利用率。最后,对于传统随机子空间方法在计算现实海洋工程结构的模态信息时,出现的响应信号中干扰成分(非平稳信号源)太多以及所消耗的时间成本太大等问题,提出一种增强数据凝练计算效率的信号平稳处理方法。通过把信号平稳处理后的响应数据在单种规则下做模态识别计算,不但能准确识别出结构的模态信息,而且计算效率也大幅提高。通过对目标平台有限元模型响应信号及现场数据进行计算处理,验证了该方法改进后具有很好的适用性和鲁棒性。改进后的随机子空间方法不但可以为结构模型优化、性能评估、损伤预报、以及结构的安全预警提供参数化指标,而且能够对结构长期的模态变化进行评估,从而对结构未来的刚度退化现象进行很好的预报。更多还原

【Abstract】 As the basic facility of offshore oil and gas resources development,offshore platform is the base of offshore production,operation and life.In order to ensure the safe service of offshore platform structure and avoid major malignant accidents,regular inspection and safety assessment of offshore platform structure must be carried out during its service period.So,the real-time modal feature extraction of offshore platform structure is an important monitoring method derived from this target.At present,the commonly used modal feature extraction methods include peak picking method,natural Excitation Technology,ITD(Ibrahim Time Domain)method,Stochastic Subspace Identification method and so on.Among them,the stochastic subspace identification(SSI)has attracted wide attention because of its advantages such as fewer calculation model parameters,no iteration,good convergence,fast operation and good robustness.Based on a jacket platform structure,this paper presents a method for real-time extraction of modal features based on covariance-driven stochastic subspace identification.Firstly,based on the finite element model of a large-scale jacket platform,the feasibility of the theory and algorithm of stochastic subspace identification is verified,and through the discussion and analysis of the calculation parameters involved in the method: system order,Hankel matrix dimension and calculation data length,the optimal parameter ratio is preliminarily selected.The next,the traditional stochastic subspace identification could not reflect the whole structure characteristics and is insensitive to the real structure under the limited measuring points.So a new method of Hankel element reconstruction is proposed.That is,the matrix elements are replaced by sub-matrix and the data arrangement of the sub-matrix is consistent with the spatial arrangement of the platform measuring points.The algorithm is highly correlated with the actual spatial structure,which improves the accuracy of modal identification and greatly improves the computational efficiency.And then,the traditional stochastic subspace identification has too many false modes when structures are large and complex.Based on cumulative characteristics of signals,false mode identification method and elimination criterion are established.In this way,the real modal of the structure can be identified effectively.not only the false modes can be effectively eliminated,but also the utilization rate of response data can be greatly increased.In addition,when the traditional random subspace method is used to calculate the modal information of real jacket platform structure,there are too many interference components(non-stationary signal sources)in the response signals and the time cost is too large.A new signal processing method are proposed to enhance the efficiency of data condensation calculation.Through the modal identification calculation of the response data after the signal smooth processing under a single rule,not only the modal information of the structure can be accurately identified,but also the calculation efficiency is greatly improved.Finally,By calculating and processing the response signals of the finite element model and field data of the target platform,it is proved that the method has good applicability and robustness.The improved stochastic subspace identification can not only provide parameterized indexes for structural model optimization,performance evaluation,damage prediction and structural safety early warning,but also evaluate the long-term modal change of the structure,so as to predict the future stiffness degradation of the structure.更多还原