Two-stage system identification approach for three-dimensional structural systems (original) (raw)
Related papers
IEEE Sensors Journal, 2000
Structural health monitoring (SHM) is a means of identifying damage from structural response to environmental loads. Real-time SHM is of particular use for rapid assessment of structural safety by owners and civil defense authorities. This paper presents an algorithm for real-time SHM during earthquake events using only acceleration measurements and infrequently measured displacement motivated by global positioning system. The algorithm identifies a nonlinear baseline model including hysteretic dynamics and permanent deformation using convex integral-based fitting methods and piecewise linear least squares fitting. The methodology identifies pre and postyield stiffness, elastic and plastic components of displacement, and final residual displacement. It thus identifies key measures of damage affecting the immediate safety or use of the structure and the long-term cost of repair and retrofit. The algorithm is tested with simulated response data using the El-Centro earthquake record and with measured response data. Both data sets are based on a four-story nonlinear steel frame structure using the El-Centro ground motion record. Overall, the algorithm is shown to provide accurate indications of the existence, location, and magnitude of structural damage for nonlinear shear-type buildings. Additionally, the identified permanent displacement is a particularly useful damage measure for the construction of probabilistic fragility functions.
Experimental studies on damage detection in frame structures using vibration measurements
2010
This paper presents an experimental study of frequency and time domain identification algorithms and discusses their effectiveness in structural health monitoring of frame structures using acceleration input and response data. Three algorithms were considered: 1) a frequency domain decomposition algorithm (FDD), 2) a time domain Observer Kalman IDentification algorithm (OKID), and 3) a subsequent physical parameter identification algorithm (MLK). Through experimental testing of a four-story steel frame model on a uniaxial shake table, the inherent complications of physical instrumentation and testing are explored. Primarily, this study aims to provide a dependable first-order and second-order identification of said test structure in a fully instrumented state. Once the characteristics (i.e. the stiffness matrix) for a benchmark structure have been determined, structural damage can be detected by a change in the identified structural stiffness matrix. This work also analyzes the stability of the identified structural stiffness matrix with respect to fluctuations of input excitation magnitude and frequency content in an experimental setting.
Structural damage identification using system dynamic properties
2005
A damage detection method is presented for the identification and quantification of damage that leads to a change in the structureÕs mass and/or stiffness properties. The proposed method requires the use of finite element to model the structure in its undamaged state as well as information on the dynamic properties such as frequencies and mode shapes of the structure in its damaged state. The technique is applicable to any structure that can be accurately modeled using the finite element method and whose frequencies and mode shapes can be reliably measured. A structure pseudo force vector derived from the residual force method is described to locate the damaged regions in the structure. A matrix condensation approach in conjunction with a proportional damage model is then employed to quantify the damage by calculating the change in stiffness and mass properties of the damaged elements in the structure. The validity of the method is demonstrated by applying it to three structures: a beam, a frame and a plate. It is shown that if the amount of damage is not excessively large, the proposed method can be used to detect damage in these structures even when the measured system dynamic properties are slightly erroneous.
Structural damage prognosis of three-dimensional large structural systems
Structure and Infrastructure Engineering, 2017
A novel procedure for the health assessment of large three-dimensional (3D) structures with several significant attractive features and improved implementation potential is proposed. Structures are represented by 3D finite elements and a substructure concept is used so that acceleration time histories can be measured only at small part(s) of the structure. Just by measuring relatively few noise-contaminated responses in the substructure, the health of the whole structure can be assessed by the system identification (SI) concept by tracking the stiffness parameter of all the elements using a significantly improved unscented Kalman filter (UKF) algorithm. Since measuring excitation time histories can be very problematic and expensive, the UKF algorithm is integrated with 3D iterative least-squares with unknown input algorithm. UKF fails to identify large structures due to convergence-related issues. The authors used short duration responses and multiple global iterations with weight factor and objective function instead of one long duration response generally used in UKF. For the preselected excitation, short duration eliminates multiple sources of excitation beyond the control of inspector. The weight factor helps accurately locate the defect spot. With informative examples, it is documented that the proposed method is superior to various other forms of Kalman filter-based algorithms.
Health Assessment of three dimensional Large Structural Systems – A novel Approach
2012
A finite-elements-based time-domain system identification technique is presented in this paper for health assessment of three dimensional structures and denoted as 3D GILS-EKF-UI. It is a modelbased procedure and represents structures by finite elements. The method integrates an iterative least-squares technique and the extended Kalman filter-based concept for detecting location(s) of defect(s) and their severity for the rapid assessment of structural health. It tracks the changes in stiffness parameters at the finite element level using dynamic response information. The procedure does not require information on dynamic excitation and uses noise-contaminated responses measured only at small part(s) of the structure. With the help of examples, it is demonstrated that the method is capable of accurately identifying defect-free and defective states of three dimensional structures. The method addresses several implementation issues for rapid, economical, and easier assessment of structu...
Structural Control and Health Monitoring, 2017
The unscented Kalman filter (UKF) is often used for nonlinear system identification in civil engineering; nevertheless, the application of the UKF to highly nonlinear structures could not provide accurate results. In this paper, an improvement of the UKF algorithm has been adopted. This methodology can consider state constraints, and it can estimate the measurement noise covariance matrix. The results obtained adopting a modified UKF have been compared to the ones obtained using the UKF for parameter estimation of a single degree of freedom nonlinear hysteretic system. The second part of this work shows results of an experimental activity on a base-isolated prototype structure. Both numerical and experimental results underline that the adopted algorithm produces better state estimation and parameter identification than the UKF, being capable of taking into account parameter boundaries. The adopted algorithm is more robust than the standard UKF in the case of measuring noise variation.
System Identification is an important concept in numerous engineering fields, such as those of structural, mechanical and aerospace engineering. It could be said that the purpose of System Identification is the determination of the dynamic characteristics of systems, which are, in addition to the frequencies, mode shapes, damping ratios and other modal characteristics, the matrices corresponding to mass, viscous damping, stiffness, Coulomb damping and Duffing stiffness. Without access to dynamic characteristics of structures, a well-defined diagnosis of the situation and of the rate of damage is not possible. If some of the nonlinear characteristics of structures, such as Coulomb damping and Duffing stiffness, are taken into account then identification errors can be reduced. In some cases, e.g. in the Damage Detection of structures, identification of the characteristic matrices of the system (mass, damping, stiffness, etc.) is as important as the modal characteristics (frequency, mo...
Localized identification of structures by Kalman filter
PROCEEDINGS-JAPAN SOCIETY OF …, 1993
A method to estimate the structural parameters of a small section of a structure was presented. A structure was decomposed into two substructures which were attached at a common boundary and three subsystems resulted which were the primary, boundary and secondary systems. The identification of the structural parameters was concentrated on the secondary system. Incorporating the state and observation equations of the secondary system in the extended Kalman filter, the stiffness and damping parameters of the secondary system can be estimated. To illustrate the proposed localized identification approach, a shear building was analyzed and the identification was concentrated on the first story.
Damage detection of a steel frame subjected to ground motion
This paper presents an experimental study of different algorithms for the health monitoring of frame structures subjected to base excitation (e.g. earthquake ground motion). These algorithms use only the acceleration time histories of the input and of the response output and are tested for the identification of the dynamic characteristics of the structure (natural frequencies and damping ratios) and for detecting and quantifying any possible structural damage that occurs in the frame. Three algorithms were considered: (1) a frequency domain decomposition algorithm, (2) a time domain Eigensystem Realization Algorithm together with Observer Kalman Identification algorithm, and (3) a subsequent physical parameter identification algorithm (MLK). Through extensive experimental testing of a four-story steel frame model on a uniaxial shake table, the performance of the various methods as well as the inherent complications of physical instrumentation and testing are explored.