悬臂梁损伤的支持向量机识别研究

针对结构损伤识别中,损伤与其影响因素之间的复杂非线性关系,提出了结构损伤识别的支持向量机方法。支持向量机是一种基于统计学理论的机器学习算法,本文以模态频率作为损伤标识量,通过支持向量机建立了损伤程度和频率之间的支持向量机模型,并以悬臂梁的损伤为例进行了计算分析,结果表明提出的方法是科学,可行的。     

Use of Ritz method for damage detection of reinforced and post-tensioned concrete beams

Damage-induced changes in modal characteristics can be detected using experimental modal analysis. In this article, based on changes in natural frequency, mode shapes, and damping ratios, a methodology for detecting damage location and severity is presented. The damage was induced by application of point load at half span location on the reinforced and post-tensioned concrete beams. The load was gradually increased to obtain different crack patterns to be used in simulation of damage scenarios. Experimental modal analysis was performed on the undamaged and damaged beams. The natural frequency and mode shapes were used to determine the location of damage. The approach is developed at an element level with a conventional finite element (FE) model by Ritz method, which is called Ritz damage detection method (RDDM). The mathematical model for both damped and undamped damaged structures have been established through the eigenvalue equations. The singular value decomposition (SVD) technique is used for determination of damage or sound index. These indexes are sensitive to the change of dynamic characteristics due to damages. This approach is applied to five simply supported post-tensioned concrete beams. The numerical results show that the exact location and severity of damage for different simulated damage scenarios could be efficiency found by the present methodology.     

Development of temperature-robust damage factor based on sensor fusion for a wind turbine structure

Wind power systems have gained much attention due to the relatively high reliability, maturity in technology and cost competitiveness compared to other renewable alternatives. Advances have been made to increase the power efficiency of the wind turbines while less attention has been focused on structural integrity assessment of the structural systems. Vibration-based damage detection has widely been researched to identify damages on a structure based on change in dynamic characteristics. Widely spread methods are natural frequency-based, mode shape-based, and curvature mode shape-based methods. The natural frequency-based methods are convenient but vulnerable to environmental temperature variation which degrades damage detection capability; mode shapes are less influenced by temperature variation and able to locate damage but requires extensive sensor instrumentation which is costly and vulnerable to signal noises. This study proposes novelty of damage factor based on sensor fusion to exclude effect of temperature variation. The combined use of an accelerometer and an inclinometer was considered and damage factor was defined as a change in relationship between those two measurements. The advantages of the proposed method are: 1) requirement of small number of sensor, 2) robustness to change in temperature and signal noise and 3) ability to roughly locate damage. Validation of the proposed method is carried out through numerical simulation on a simplified 5 MW wind turbine model.     

Wavelet-Based Structural Health Monitoring of Earthquake Excited Structures

Abstract:   The article presents a wavelet-based structural health monitoring technique for structures subjected to an earthquake excitation utilizing the instantaneous modal information. The instantaneous mode shape information is first extracted from the vibration response data collected during an earthquake event by using a wavelet packet sifting process. A confidence index (CI) is proposed to validate the results obtained. The identified normalized instantaneous mode shapes in conjunction with the corresponding CIs can be effectively used to monitor damage development in the structure. The effectiveness of the proposed approach is illustrated for two damage scenarios, sudden stiffness loss and progressive stiffness degradation, and different base excitations including three real earthquake signals and a random signal. Consistently good results were obtained in all cases. Issues related to robustness of the method in the presence of a measurement noise and sensitivity to damage severity are discussed.     

Damage location and assessment along structural elements using damage submatrices

A new method to locate and determine the magnitude of damage, expressed as the loss of stiffness, along structural elements of buildings, called Damage Submatrices Method is presented. The method uses information related to the stiffness of both states of the structure: undamaged and damaged, utilizing iterative eigenvalue computations. The novelty of the method relies on its capability to identify damage in specific zones of structural elements depending on how the undamaged model of the structure was discretized. Another advantage of the proposed method is its convergence. This paper also contributes a new approach to simulate noise perturbing stiffness matrices. This technique can be useful for realistic damage detection investigations. Also, the proposed technique used to expand condensed damaged stiffness matrices, to global coordinates, may be useful for system identification projects. Two examples taken from the literature are studied involving limited modal measurements and noise effects to calibrate the method and corroborate its capability for damage assessment. Application of the proposed method to these structures corroborates its efficiency for identifying stiffness degradation of structural elements.     

Optimal Sensor Placement for Mode Shapes Via Shannon's Sampling Theorem

Abstract: The feasibility of using Shannon's sampling theorem to reconstruct exact mode shapes of a structural system from a limited number of sampling points is investigated. Shannon's sampling theorem for the time domain is reviewed. The theorem is then extended to the spatial domain. Mode shapes are reconstructed from a limited amount of data, and the reconstructed mode shapes are compared with the exact mode shapes. On the basis of the results, simple rules are proposed for the placement of accelerometers in modal parameter extraction experiments. The feasibility of applying the rules and the extended Shannon's theorem to damage localization in a simple structure is demonstrated.     

Stationarity‐Based Approach for the Selection of Lag Length in Cointegration Analysis Used for Structural Damage Detection

Cointegration has been recently brought to structural health monitoring (SHM) as a new methodology for dealing with the problem of environmental and/or operational variability in monitored structures. However, it is well known that the choice of lag length in cointegration analysis has a strong influence on damage detection results. The article presents a new approach for optimal lag length selection in cointegration analysis used for structural damage detection. This new method is based on stationarity analysis of data representing undamaged condition. The proposed method is validated using Lamb wave data under the effects of temperature variations and vibroacoustic data obtained from nonlinear vibroacoustic modulation experiments with different low‐frequency vibration (or modal) excitations. The results demonstrate the effectiveness of the method for structural damage detection based on SHM data heavily affected by environmental or operational conditions.     

Static-based early-damage detection using symbolic data analysis and unsupervised learning methods

A large amount of researches and studies have been recently performed by applying statistical and machine learning techniques for vibration-based damage detection. However, the global character inherent to the limited number of modal properties issued from operational modal analysis may be not appropriate for early-damage, which has generally a local character. The present paper aims at detecting this type of damage by using static SHM data and by assuming that early-damage produces dead load redistribution. To achieve this objective a data driven strategy is proposed, consisting of the combination of advanced statistical and machine learning methods such as principal component analysis, symbolic data analysis and cluster analysis. From this analysis it was observed that, under the noise levels measured on site, the proposed strategy is able to automatically detect stiffness reduction in stay cables reaching at least 1%.     

Scalable Structural Modal Identification Using Dynamic Sensor Network Data with STRIDEX

This article uses the formulation of the structural identification using expectation maximization (STRIDE) algorithm for compatibility with the truncated physical model (TPM) to enable scalable, output‐only modal identification using dynamic sensor network (DSN) data. The DSN data class is an adaptable and efficient technique for storing measurements from a very large number of sensing nodes, which is the case in mobile sensor networks and BIGDATA problems. In this article, the STRIDEX output‐only identification algorithm is proposed for the stochastic TPM to estimate structural modal properties (frequencies, damping ratios, and mode shapes) directly from DSN data. The spatial information produced by this novel algorithm, called STRIDEX (“X” for extended), is scalable, as demonstrated in a strategy to construct high‐resolution mode shapes from a single DSN data set using a series of independent identification runs. The ability to extract detailed structural system information from DSN data in a computationally scalable framework is a step toward mobile infrastructure informatics in a large urban setting. The performance of the STRIDEX algorithm is demonstrated, using the simulated response of a 5,000 DOF structure, and experimentally, using measurements from two mobile sensor cars, which scanned about 8,000 points on a beam specimen in the laboratory. In the experimental results, a mobile sensor is shown to provide over 120 times more mode shape points than a fixed sensor.     

基于曲率模态的拱结构损伤识别方法研究

以两端铰支的圆弧形拱为研究对象,通过有限元数值模拟计算得到拱损伤前后的前四阶模态参数,然后运用中心差分近似求得拱的曲率模态并用于拱的损伤检测研究。结果表明：当布置有足够数量的振型测点时,拱损伤前后基于径向位移和转角位移的模态曲率差均可用于拱损伤的探测和定位,并大致判断其损伤程度。     

基于振型参数的装配式桥梁损伤识别研究

利用结构动力特性的振型参数对桥梁进行快速损伤诊断和定位,可以提高结构性能评价与损伤诊断的效率。本文以装配式预应力混凝土T梁为算例,通过定义位移振型和曲率振型的桥梁损伤识别指标进行损伤识别,计算结果表明采用位移振型和曲率振型的方法进行损伤识别和定位效果较好。     

Vibration based structural health monitoring: Wavelet packet transform based solution

One prominent problem for vibration-based structural health monitoring is to extract condition indices which are sensitive to damage and yet insensitive to measurement noise. In this paper, a condition index extraction method based on the wavelet packet transform (WPT) is proposed. This transform leads to the formulation of a novel condition index: wavelet packet signature (WPS). The sensitivity of the WPS to the change of structural parameters is derived and validated on a five-degrees-of-freedom spring-mass system. Results show that the WPS is significantly more sensitive to the stiffness change than the natural frequencies and the mode shapes. Its sensitivity is slightly better or comparable to that of the modal flexibility matrices depending on the location of damage. A variability analysis is also performed to study the effect of measurement noise on the proposed WPS. Results show that the WPS does not show any significant variation even under the presence of 10 dB noise. To illustrate the potential of the WPS, a damage indicator is formulated and used to monitor the health condition of the structural system. An experimental study on a three-storey frame shows that when incorporated with a statistical process control approach, the WPS-based damage indicator can distinctly identify the presence of damage in the system.     

Automatic pixel‐level multiple damage detection of concrete structure using fully convolutional network

Deep learning‐based structural damage detection methods overcome the limitation of inferior adaptability caused by extensively varying real‐world situations (e.g., lighting and shadow changes). However, most deep learning‐based methods detect structural damage at the image level and grid‐cell level. To provide pixel‐level detection of multiple damages, a Fully Convolutional Network (FCN)‐based multiple damages detection method for concrete structure is proposed. To realize this method, a database of 2,750 images (with 504 × 376 pixels) including crack, spalling, efflorescence, and hole images in concrete structure is built, and the four damages included in those images are labeled manually. Then, the architecture of the FCN is modified, trained, validated, and tested using this database. A strategy of model‐based transfer learning is used to initialize the parameters of the FCN during the training process. The results show 98.61% pixel accuracy (PA), 91.59% mean pixel accuracy (MPA), 84.53% mean intersection over union (MIoU), and 97.34% frequency weighted intersection over union (FWIoU). Subsequently, the robustness and adaptability of the trained FCN model is tested and the damage is extracted, where damage areas are provided according to a calibrated relation between the ratio (the pixel area and true area of the detected object) and the distance from the smartphone to the concrete surface using a laser range finder. A comparative study is conducted to examine the performance of the proposed FCN‐based approach using a SegNet‐based method. The results show that the proposed method substantiates quite better performance and can indeed detect multiple concrete damages at the pixel level in realistic situations.     

Modal Response‐Based Visual System Identification and Model Updating Methods for Building Structures

This article proposes a new system identification (SI) method using the modal responses obtained from the dynamic responses of a structure for estimating modal parameters. Since the proposed SI method visually extracts the mode shape of a structure through the plotting of modal responses based on measured data points, the complex calculation process for the correlation and the decomposition for vibration measurements required in SI methods can be avoided. Also, without dependence on configurations of SI methods inducing variations of modal parameters, mode shapes and modal damping ratios can be stably extracted through direct implementation of modal response. To verify the feasibility of the proposed method, the modal parameters of a shear frame were extracted from modal displacement data obtained from a vibration test, and the results were compared with those obtained from the existing frequency domain SI method. The proposed method introduces the maximum modal response ratio of each mode computed by modal displacement data, and from this, the contribution of each mode and each measured location to the overall structural response is indirectly evaluated. Moreover, this article proposes a model updating method establishing the error functions based on the differences between the analytical model and measurement for the natural frequencies and the modal responses reflecting both mode shape and modal contribution. The validity of the proposed method is verified through the response prediction and modal contributions of the models obtained from model updating based on dynamic displacement from a shaking table test for a shear‐type test frame.     

Modal Identification for High‐Rise Building Structures Using Orthogonality of Filtered Response Vectors

The modal parameters of civil structures (natural frequency, mode shape, and mode damping ratio) are used for structural health monitoring (SHM), damage detection, and updating the finite element model. Long‐term measurement has been necessary to conduct operational modal analysis (OMA) under various loading conditions, requiring hundreds of thousands of discrete data points for estimating the modal parameters. This article proposes an efficient output‐only OMA technique in the form of filtered response vector (frv)‐based modal identification, which does not need complex signal processing and matrix operations such as singular value decomposition (SVD) and lower upper (LU) factorization, thus overcoming the main drawback of the existing OMA technique. The developed OMA technique also simplifies parameters such as window or averaging, which should be designed for signal processing by the OMA operator, under well‐separated frequencies and loading conditions excited by white noise. Using a simulation model and a 4‐story steel frame specimen, the accuracy and applicability were verified by comparing the dynamic properties obtained by the proposed technique and traditional frequency‐domain decomposition (FDD). In addition, the applicability and efficiency of the method were verified by applying the developed OMA to measured data, obtained through a field test on a 55‐story, 214‐m‐tall high‐rise building.     

Dynamic reduction-based structural damage detection of transmission towers: Practical issues and experimental verification

Structural damage detection in large-scale three-dimensional spatial structures is a challenging problem. It is impractical to develop a general damage-detection method that is applicable to all types of structural systems and all kinds of damage. A practical and efficient structural damage detection method must consider the characteristics of the target structure and damage in the development stage. In 2009, Yin et al. [33] proposed a damage detection method for the health monitoring of transmission towers. The method was developed based on the dynamical finite element (FE) model reduction technique, which utilizes identified modal parameters, such as natural frequencies and mode shapes, with only a limited number of sensors. In Ref. [33], the proposed method was numerically verified by simulated noisy data from a three-dimensional transmission tower model for single and multiple damage cases. This paper discusses some practical issues related to the proposed method, such as sensor placement and computational efficiency. Rather than proposing a general sensor placement method, a set of preliminary sensor locations is determined based on engineering judgement. This set of sensor locations is then checked against the results of a sensitivity analysis to ensure that the measured data contain information for identifying all of the target damage scenarios. To reduce the required computational power, two simplified versions of the proposed method are presented. The proposed method is then verified with a scaled-down model of a transmission tower (2.4 m high) that was built at the Structural Vibration Laboratory (SVL) of the City University of Hong Kong. This paper reports the detailed experimental setup and the method of extracting the modal parameters from a series of free vibration tests with only a limited number of sensors. The verification results show that the proposed damage detection method identifies the damaged sub-structures in all of the experimental cases considered. It must be pointed out that the transmission tower structure, in its operating conditions, suffers from the effect of the forces transmitted from the cables it carries. The influence of this force on the damage identification result is great and can not be neglected in practice. In the present experimental case study, only a transmission tower-like structure without carrying the cables is investigated in laboratory conditions.     

An efficient two-stage approach for structural damage detection using meta-heuristic algorithms and group method of data handling surrogate model

In this study, the performance of an efficient two-stage methodology which is applied in a damage detection system using a surrogate model of the structure has been investigated. In the first stage, in order to locate the damage accurately, the performance of the modal strain energy based index for using different numbers of natural mode shapes has been evaluated using the confusion matrix. In the second stage, to estimate the damage extent, the sensitivity of most used modal properties due to damage, such as natural frequency and flexibility matrix is compared with the mean normalized modal strain energy (MNMSE) of suspected damaged elements. Moreover, a modal property change vector is evaluated using the group method of data handling (GMDH) network as a surrogate model during damage extent estimation by optimization algorithm; in this part of methodology, the performance of the three popular optimization algorithms including particle swarm optimization (PSO), bat algorithm (BA), and colliding bodies optimization (CBO) is examined and in this regard, root mean square deviation (RMSD) based on the modal property change vector has been proposed as an objective function. Furthermore, the effect of noise in the measurement of structural responses by the sensors has also been studied. Finally, in order to achieve the most generalized neural network as a surrogate model, GMDH performance is compared with a properly trained cascade feed-forward neural network (CFNN) with log-sigmoid hidden layer transfer function. The results indicate that the accuracy of damage extent estimation is acceptable in the case of integration of PSO and MNMSE. Moreover, the GMDH model is also more efficient and mimics the behavior of the structure slightly better than CFNN model.     

Vibration‐Based Damage Localization in Flexural Structures Using Normalized Modal Macrostrain Techniques from Limited Measurements

Abstract: This article presents damage locating indices based on normalized modal macrostrain (MMS) as improvement on the typical curvature‐dependent methods. Vulnerability to noise and the use of numerical differentiation procedures are the key factors for the poor performance of many curvature‐dependent methods using displacement mode shapes. Whereas dynamic distributed strain measurement data from long‐gauge FBG sensors have significantly improved the performance of many damage identification methods, the sensitivity to local damage diminishes as the gauge length increases. The proposed model‐free damage identification techniques based on normalized MMS vectors are successfully implemented to locate damage in beam‐like structures through numerical simulations and experimental verifications. The unique advantages of the techniques are their simplicity, robustness to noise, ability to precisely identify small damage extents, and localize single and multiple damage states using limited measurable modes from few sensors.     

Automatic detection method of cracks from concrete surface imagery using two‐step light gradient boosting machine

Automated crack detection based on image processing is widely used when inspecting concrete structures. The existing methods for crack detection are not yet accurate enough due to the difficulty and complexity of the problem; thus, more accurate and practical methods should be developed. This paper proposes an automated crack detection method based on image processing using the light gradient boosting machine (LightGBM), one of the supervised machine learning methods. In supervised machine learning, appropriate features should be identified to obtain accurate results. In crack detection, the pixel values of the target pixels and geometric features of the cracks that occur when they are connected linearly should be considered. This paper proposes a methodology for generating features based on pixel values and geometric shapes in two stages. The accuracy of the proposed methodology is investigated using photos of concrete structures with adverse conditions, such as shadows and dirt. The proposed methodology achieves an accuracy of 99.7%, sensitivity of 75.71%, specificity of 99.9%, precision of 68.2%, and an F‐measure of 0.6952. The experimental results demonstrate that the proposed method can detect cracks with higher performance than the pix2pix‐based approach. Furthermore, the training time is 7.7 times shorter than that of the XGBoost and 2.3 times shorter than that of the pix2pix. The experimental results demonstrate that the proposed method can detect cracks with high accuracy.     

Operational Modal Parameter Identification from Power Spectrum Density Transmissibility

Abstract: Operational modal analysis subjected to ambient or natural excitation under operational conditions has recently drawn great attention. In this article, the power spectrum density transmissibility (PSDT) is proposed to extract the operational modal parameters of a structure. It is proven that the PSDT is independent of the applied excitations and transferring outputs at the system poles. As a result, the modal frequencies and mode shapes can be extracted by combing the PSDTs with different transferring outputs instead of different load conditions where the outputs from only one load condition are needed. A five‐story shear building subjected to a set of uncorrelated forces at different floors is adopted to verify the property of PSDTs and illustrate the accuracy of the proposed method. Furthermore, a concrete‐filled steel tubular half‐through arch bridge tested in the field under operational conditions is used as a real case study. The identification results obtained from currently developed method have been compared with those extracted from peak‐picking method, stochastic subspace identification, and finite element analysis. It is demonstrated that the operational modal parameters identified by the current technique agree well with other independent methods. The real application to the field operational vibration measurements of a full‐sized bridge has shown that the proposed PSDTs are capable of identifying the operational modal parameters (natural frequencies and mode shapes) of a structure.