Damage Detection Methodology Research Articles

Early wheel flat detection: an automatic data-driven wavelet-based approach for railways

Wheel defects can induce damage to the railway tracks, increasing considerable maintenance costs for both railway administrations and rolling stock operators. This paper aims to develop an unsupervised early damage detection methodology, capable of automatically distinguishing a defective wheel from a healthy one, with respect to the small flat size. The proposed methodology is based on the acceleration and shear time histories evaluated on the rails for the passage of traffic loads, and involves the following steps: (i) data acquisition from sensors; (ii) feature extraction from acquired responses with continuous wavelet transform (CWT) model; (iii) feature normalisation to suppress environmental and operational variations; (iv) data fusion to merge the features from each sensor and enhance sensitivity to detect wheel defects; and (v) feature classification to classify the extracted features into two categories: a healthy wheel or a defective one. The shear and acceleration measurement points are strategically defined in order to examine the sensitivity of the proposed methodology, not only to the type of sensors, but also to the position where they are installed. It has been demonstrated that one sensor can detect a defective wheel automatically, allowing the development of an easy-to-implement low-cost monitoring system.

Deep learning for seismic structural monitoring by accounting for mechanics-based model uncertainty

This paper presents a hybrid deep learning methodology for seismic structural monitoring, damage detection, and localization of instrumented buildings. The proposed methodology develops mechanics-based structural models to generate sample response datasets by accounting for the uncertainty of model parameters that can highly affect the estimation of baseline model nonlinear responses. The uncertainty of model parameters is evaluated through the design of experiments methodology by employing the central composite design for sampling. The generated sample response dataset is utilized for training a hybrid data-driven model that combines a convolutional neural network and wavelet packet transform modules for feature extraction. The global story-level noise-contaminated response measurements are used as input for the data-driven model to perform damage detection and localization in a manner consistent with performance-based design criteria. The performance of the proposed methodology is studied in the context of numerical and experimental case studies developed based on the shake table testing of a concentrically braced frame subject to various input ground motion intensities at the E-Defense facility in Miki, Japan.

Damage Analysis and Prediction in Glass Fiber Reinforced Polyester Composite Using Acoustic Emission and Machine Learning

One of the most pervasive types of structural problems in aircraft industries is fatigue cracking that can potentially occur without anticipation with catastrophic failures and unexpected downtime. Acoustic emission (AE) is a passive structural health monitoring (SHM) technique, since it offers real time damage detection based on stress waves generated by cracking in the structure. Machine learning techniques have presented great success over the past few years with a large number of applications. This study assesses the progression of damage occurring on glass fiber reinforced polyester composite specimens using two approaches of machine learning, namely, Supervised and Unsupervised learning. A methodology for damage detection and characterization of composite is presented. The result shows that machine learning can predict damages in composite materials. All predictive models and their performance as well as AE parameters had a direct relationship with the applied stress values, suggesting that these correlations are reliable means of predicting fatigue life in a composite material.

A hybrid methodology for structural damage detection uniting FEM and 1D-CNNs: Demonstration on typical high-pile wharf

Vibration-based structural damage detection (SDD) has been a subject of intense research in structural health monitoring (SHM) for large civil engineering structures over the decades. The performance of the conventional SDD approaches predominantly relies on the rational choices of the damage feature and classifier. Hand-crafted features or fixed classifiers would not be the optimal choice for all structural damaged scenarios. This paper proposes a novel, quick and precise real-time SDD framework for high-pile wharf foundations using a combination of finite element modeling and 1D convolutional neural networks (CNNs). The distinct advantage of this method lies in extracting the damage-related features from the raw displacement response directly and automatically, and the computational complexity of the compact 1D CNNs is significantly lower because the data processing involves only simple 1D operations. The results show that the presented 1D CNNs have a superior ability to accurately identify the occurrence and location of damage in real time. In addition, the comprehensive performance of the CNNs trained by the displacement response dataset in component form is significantly better than that based on the dataset in absolute value form. The results also demonstrated that although the proposed CNNs are more sensitive to the longitudinal and lateral displacement responses of the high-pile wharf structure, the vertical component still has a positive effect on the improvement of the generalization and robustness of the CNNs.

A data-based structural health monitoring approach for damage detection in steel bridges using experimental data

There is a need for reliable structural health monitoring (SHM) systems that can detect local and global structural damage in existing steel bridges. In this paper, a data-based SHM approach for damage detection in steel bridges is presented. An extensive experimental study is performed to obtain data from a real bridge under different structural state conditions, where damage is introduced based on a comprehensive investigation of common types of steel bridge damage reported in the literature. An analysis approach that includes a setup with two sensor groups for capturing both the local and global responses of the bridge is considered. From this, an unsupervised machine learning algorithm is applied and compared with four supervised machine learning algorithms. An evaluation of the damage types that can best be detected is performed by utilizing the supervised machine learning algorithms. It is demonstrated that relevant structural damage in steel bridges can be found and that unsupervised machine learning can perform almost as well as supervised machine learning. As such, the results obtained from this study provide a major contribution towards establishing a methodology for damage detection that can be employed in SHM systems on existing steel bridges.

Nonparametric damage detection and localization model of framed civil structure based on local gravitation clustering analysis

Structural damage detection is still a stimulating problem due to the complicated non-linear behaviour of the structural system, incomplete sensed data, presence of noise in the data, and uncertainties in both experimental measurement and analytical model. This paper presents the application of a non-linear signal processing tool and artificial intelligence-based methodology for nonparametric damage detection to address the above stated issues. Local mean decomposition (LMD), as an adaptive signal processing technique, is exploited to extract multidimensional damage features over acquired non-linear non-stationary vibration signals. These features are classified into categories, which are then utilized to calculate a damage indicator. Classification of multidimensional feature space has been a challenging issue since its inception. To address this, local gravitation clustering (LGC), a self-evaluating, synergic clustering technique, is employed. The relevance and significance of the process corresponding to the problem have also been an important concern. The outcomes of the whole process prove its proficiency in damage identification. The efficiency of the process is then compared with existing clustering methods on several parameters. The proposed algorithm is also validated for operational and environmental conditions by considering finite cases analogues to physical ailments like temperature, ageing and live loads.

Deep learning-based Structural Health Monitoring for damage detection on a large space antenna

Due to the stringent requirements imposed by state-of-the-art technologies, most of modern spacecrafts are now equipped with very large substructures such as antennas, deployable booms and solar arrays. However, while the size of these elements increases, their mass is limited by the rocket maximum take-off weight and, therefore, they result to be lightweight and very flexible. A natural concern derived from this trend is that these structures are now more susceptible to structural damages during launch phase or operational life (impacts, transient thermal states and fatigue). Since the degradation of some structural elements would naturally result in a modification of the system dynamical behaviour, damage detection processes are usually performed by comparing the dynamical responses or the structural model matrices (i.e. stiffness, mass and damping) of the undamaged model with those of the damaged structure. However, for large structures, the presence of local damage does not generally induce substantial change in the global dynamic which makes the local failures difficult to detect. A methodology for structural damage detection based on data-driven techniques with Deep Neural Network (DNN) is hereby proposed for the study of a large in-orbit flexible system. The deep architecture evaluated in this work is composed by stacking several neural network layers with different functions. In order to generate data for training and testing the machine learning model, different damage scenarios are generated via Finite Element commercial code and, based on the extracted modal parameters, the fully coupled 3D equations for the flexible spacecraft are integrated to test typical profiles of attitude manoeuvres. The DNN model is trained using the sensor-measured time series responses, where each of the former has been associated with the label of the corresponding damage scenario.

Data Mining-based Structural Damage Identification of Composite Bridge using Support Vector Machine

A structural health monitoring system contains two components, i.e. a data collection approach comprising a network of sensors for recording the structural responses as well as an extraction methodology in order to achieve beneficial information on the structural health condition. In this regard, data mining which is one of the emerging computer-based technologies, can be employed for extraction of valuable information from obtained sensor databases. On the other hand, data inverse analysis scheme as a problem-based procedure has been developing rapidly. Therefore, the aforesaid scheme and data mining should be combined in order to satisfy increasing demand of data analysis, especially in complex systems such as bridges. Consequently, this study develops a damage detection methodology based on these strategies. To this end, an inverse analysis approach using data mining is applied for a composite bridge. To aid the aim, the support vector machine (SVM) algorithm is utilized to generate the patterns by means of vibration characteristics dataset. To compare the robustness and accuracy of the predicted outputs, four kernel functions, including linear, polynomial, sigmoid, and radial basis function (RBF) are applied to build the patterns. The results point out the feasibility of the proposed method for detecting damage in composite slab-on-girder bridges.

Advanced debonding detection technique for aerospace composite structures

PurposeThe purpose of this paper is to introduce a methodology aimed to detect debonding induced by low impacts energies in typical aeronautical structures. The methodology is based on high frequency sensors/actuators system simulation and the application of elliptical triangulation (ET) and probability ellipse (PE) methods as damage detector. Numerical and experimental results on small-scale stiffened panels made of carbon fiber-reinforced plastic material are discussed.Design/methodology/approachThe damage detection methodology is based on high frequency sensors/actuators piezoceramics system enabling the ET and the PE methods. The approach is based on ultrasonic guided waves propagation measurement and simulation within the structure and perturbations induced by debonding or impact damage that affect the signal characteristics.FindingsThe work is focused on debonding detection via test and simulations and calculation of damage indexes (DIs). The ET and PE methodologies have demonstrated the link between the DIs and debonding enabling the identification of position and growth of the damage.Originality/valueThe debonding between two structural elements caused in manufacturing or in-service is very difficult to detect, especially when the components are in low accessibility areas. This criticality, together with the uncertainty of long-term adhesive performance and the inability to continuously assess the debonding condition, induces the aircrafts’ manufacturers to pursuit ultraconservative design approach, with in turn an increment in final weight of these parts. The aim of this research’s activity is to demonstrate the effectiveness of the proposed methodology and the robustness of the structural health monitoring system to detect debonding in a typical aeronautical structural joint.

Damage Identification of Semi-Rigid Connection in Structures Based on Nonlinear Vibration Characteristics

A damage-detection methodology based on nonlinear vibration characteristics was presented in this paper to identify damage of semi-rigid connections due to looseness of high strength bolts. In various damage cases caused by different loose severity of bolts in different positions, a specimen with semi-rigid connections was test and acceleration signals of free vibration were recorded by sensors placed at its both ends. The recorded signals were then dealt with by wavelet de-noising and blind source separation to get the first vibrational components. By Hilbert transform of these components, the amplitude-frequency curve of each case was obtained. The results showed that vibration of the specimen with semi-rigid connections presented typical nonlinear characteristics. With the increase of connection damage, the nonlinear degree increased accordingly. When one end had larger connection damage than the other one, it exhibited greater nonlinear extent. The results prove that it is valuable to apply the suggested nonlinear vibration characteristics based procedure to identification of semi-rigid connection damage.

An artificial neural network methodology for damage detection: Demonstration on an operating wind turbine blade

This study presents a novel artificial neural network (ANN) based methodology within a vibration-based structural health monitoring framework for robust damage detection. The ANN-based methodology establishes the nonlinear relationships between selected damage sensitive features (DSF) influenced by environmental and operational variabilities (EOVs) and their corresponding novelty indices computed by the Mahalanobis distance (MD). The ANN regression model is trained and validated based on a reference state (i.e., a healthy structure). The trained model is used to predict the corresponding MD of new observations. The prediction error between the calculated and predicted MD is used as a new novelty index for damage detection. Firstly, an artificial 2D feature set is generated to illustrate how the limitations of solely using the MD-based novelty index can be overcome by the proposed ANN-based methodology. Secondly, the methodology is implemented in data obtained from an in-operation wind turbine with different artificially induced damage scenarios in one of its blades. Finally, the performance of the proposed methodology is evaluated by the metrics of accuracy, F1-score and Matthews correlation coefficient. The results demonstrate the advantages of the proposed methodology by improving damage detectability in all the different damage scenarios despite the influence of EOVs in both the simulated and real data.

A dynamic harmonic regression approach for bridge structural health monitoring

Structural damage in a bridge is defined as a significant deviation in the structural response from its standard operating conditions, not explainable by variations in external environmental and operational effects. However, environmental effects such as temperature fluctuations can cause significant seasonal variations in the structural response of a bridge and can mask its changes due to structural damage. This poses a challenge for structural health monitoring of bridges where reliable diagnosis of damage or deterioration is often related to isolation of the responses. To address it, a statistical damage-detection methodology is introduced where strain data are modelled using a dynamic harmonic regression time-series model. Prediction intervals of multi-step ahead forecasts from the dynamic harmonic regression model are then used as statistical control limits within which the observed phenomenon should fall under standard operating conditions. This single recursive structural health monitoring framework for automatic fitting and multi-step ahead forecasting of 1-min interval time-series strain data includes recorded temperature values and diurnal trends as regressors in the model to account for environmental variations. The potential of this method as a robust automatic structural health monitoring strategy is demonstrated on strain data sampled at 1-min interval from a full-scale damaged pre-stressed concrete bridge – before, during and after repair. The proposed method can capture both sudden and daily changes in structural response due to temperature effects, and a rolling multi-step ahead interval forecast was able to identify damage on back-cast data transitioning from a healthy state to a damaged state.

Negative selection algorithm based methodology for online structural health monitoring

Even small damages in engineered systems can endanger the users or lead to significant economic costs. The detection of damage existence is only the first task within the Damage Identification strategy, nonetheless, it is still very complex, especially for Civil Engineering products. The present study is devoted to the analysis of a global methodology for damage detection based on a recently developed version of the Negative Selection Algorithm (NSA). Given any new monitored data measured from the system, the methodology addresses the one-class classification problem by labelling it as normal or abnormal, thus potentially damaged. The strategy is independent of the engineered system under investigation and of the damage-sensitive features used to assess the state, as long as only two features per time are analysed by the classifier. The methodology includes a set of features meant to tackle typical shortcomings and open issues which might prevent an effective application of NSA to damage detection. Some of them are shared with other soft computing techniques for classifications. Finally, the methodology is applied to the data collected from a scaled masonry arch built and tested at the University of Minho (Portugal). The datasets are used to validate the algorithm features, to compare the NSA with other well-known classification algorithms and to simulate a real monitoring to assess the readiness of the algorithm response.

Time-domain Markov chain Monte Carlo–based Bayesian damage detection of ballasted tracks using nonlinear ballast stiffness model

This article reports the development of a methodology for detecting ballast damage under a sleeper based on measured sleeper vibration following the Bayesian statistical system identification framework. To ensure the methodology is applicable under large amplitude vibration of the sleeper (e.g. under trainload), the nonlinear stress–strain behavior of railway ballast is considered. This, on one hand, significantly reduces the problem of modeling error, but, on the other hand, increases the number of uncertain model parameters. The uncertainty associated with the identified model parameters of the rail–sleeper–ballast system may be very high. To overcome this difficulty, the Markov chain Monte Carlo–based Bayesian model updating is adopted in the proposed methodology for the approximation of the posterior probability density function of uncertain model parameters. Owing to the nonlinear behavior of the system, the model updating is performed in the time domain instead of the modal domain. The applicability of the proposed damage detection methodology was first verified numerically using simulated impact hammer test data in two damaged cases perturbed with Gaussian white noise. Second, impact hammer tests of in situ sleepers in the full-scale in-door ballasted track test panel were carried out to collect data for the experimental verification of the proposed methodology. Artificial ballast damage was simulated under the target concrete sleeper by replacing normal-sized ballast particles (∼60 mm) by small-sized ballast particles (∼15 mm). The proposed methodology successfully identified the location and severity of ballast damage under the sleeper. From the calculated posterior marginal probability density functions of model parameters, one can quantify the uncertainties associated with the damage detection results. The proposed methodology is an essential step in the development of a long-term railway track health monitoring system utilizing train-induced vibration.

Limited Sensor-Based Probabilistic Damage Detection Using Combined Normal–Lognormal Distributions

A methodology for probabilistic damage detection in Bayesian framework without any requirement of mode matching is presented with detailed formulations on finite element model updating using incomplete modal data measured using limited number of sensors. Multiple modal measurement/data sets from multiple different sensor set-ups can be used in the proposed methodology with further scope for using repeated measurements from any single sensor set-up. Combined normal–lognormal multivariate distribution is considered in the Bayesian framework. Strictly positive random parameters are assigned with lognormal distribution, while the remaining random parameters are assigned with normal distribution. In this work, the normal distribution is incorporated for the likelihood function, which consists of the eigen-system equation error and the error between the system mode shapes and experimental mode shapes. On the other hand, mass and stiffness parameters are assigned with the lognormal distribution. Detailed formulations for probabilistic identification of changes/damages are also developed. The proposed approach is validated using a three-dimensional building structure considering multiple simulated damage cases. Performance in updating and damage detection is evaluated based on multi-set-up and multi-dataset considerations. Besides, the proposed technique is compared with the similar Bayesian updating solely based on normal distribution and the Gibbs sampling.

Identifying damage on a bridge using rotation-based Bridge Weigh-In-Motion

Bridge Weigh-in-Motion (B-WIM) systems use the bridge response under a traversing vehicle to estimate its axle weights. The information obtained from B-WIM systems has been used for a wide range of applications such as pre-selection for weight enforcement, traffic management/planning and for bridge and pavement design. However, it is less often used for bridge condition assessment purposes which is the main focus of this study. This paper presents a bridge damage detection concept using information provided by B-WIM systems. However, conventional B-WIM systems use strain measurements which are not sensitive to local damage. In this paper the authors present a B-WIM formulation that uses rotation measurements obtained at the bridge supports. There is a linear relationship between support rotation and axle weight and, unlike strain, rotation is sensitive to damage anywhere in the bridge. Initially, the sensitivity of rotation to damage is investigated using a hypothetical simply supported bridge model. Having seen that rotation is damage-sensitive, the influence of bridge damage on weight predictions is analysed. It is shown that if damage occurs, a rotation-based B-WIM system will continuously overestimate the weight of traversing vehicles. Finally, the statistical repeatability of ambient traffic is studied using real traffic data obtained from a Weigh-in-Motion site in the U.S. under the Federal Highway Administration’s Long-Term Pavement Performance programme and a damage indicator is proposed as the change in the mean weights of ambient traffic data. To test the robustness of the proposed damage detection methodology numerical analysis are carried out on a simply supported bridge model and results are presented within the scope of this study.

Towards the generation of controlled one-inch impact damage in thick CFRP composites for SHM and NDE validation

In this paper, an attempt has been made to conduct controlled low velocity impact experiments in quasi-isotropic carbon fiber reinforced polymer (CFRP) composites of increasing thicknesses to generate barely visible impact damage (BVID) that creates 1″ impact damage diameter. Such impacted CFRP coupons with controlled damage diameter would serve in the future for validation of structural health monitoring (SHM) and nondestructive evaluation (NDE) damage detection methodologies. Various 2-mm, 4-mm and 6-mm thick quasi-isotropic CFRP composite plates having similar stacking sequence were manufactured in a compression molding machine. These plates were cut into numerous 6″ × 4″ coupons to be impacted in accordance with ASTM D7136 standard on a Dynatup impact testing machine. The goal of the experiments was to identify the combination of impactor mass, energy and momentum at which approximately 1″ impact damage diameter could be produced for test coupons of increasing thicknesses. Interesting observations were made with respect to the size and shape of the impact damage as the thickness of the composite coupons was increased from 2-mm to 6-mm. Modified experiments were also conducted on 6-mm coupons to achieve close to 1″ impact damage diameter. The greatest challenge in conducting these experiments was in the thick 4-mm and 6-mm coupons which did not display a predictable trend. Future work should focus on overcoming this challenge towards achieving a predictable impact damage size methodology.

Damage detection in girder bridges using modal curvatures gapped smoothing method and Convolutional Neural Network: Application to Bo Nghi bridge

This paper addresses a damage detection method based on changes in modal curvature combined with Convolutional Neural Network (CNN). The use of modal curvature for damage detection is very well known. Some methods require modal data of healthy structures as a reference, others do not. Gapped Smoothing Method (GSM) is a vibration-based damage detection methodology, which makes use of modal curvature for identifying the location of structural damage and does not require information of the intact structure. The Bo Nghi bridge is used as an illustrative example. This bridge consists of four T-shaped concrete simply supported girders. One single beam with the same length and cross-section of the bridge girder is modeled and used to extract numerical data to train the CNN. A CNN is trained by using images from the damage index of the GSM to classify the damage location in the numerical beam model. Finally, the finite element model of the bridge is built and used to model damage scenarios to test the trained CNN. The results indicate that the combination of GSM and CNN can be used for damage detection and localization.

Neural network-based damage identification in composite laminated plates using frequency shifts

Delamination is the principal mode of failure of laminated composites. It is caused by the rupture of the fiber–matrix interface and results in the separation of the layers. This failure is induced by interlaminate tension and shear that is developed due to a variety of factors such as fatigue. Composites are known for excellent structural performance, which can be significantly affected by delamination. Such damages are not always visible on the surface and can lead to sudden catastrophic failures. To ensure a structural performance and integrity, accurate methods to monitor damages are required. This work presents a methodology for damage detection and identification on laminated composite plates using artificial neural networks fed with modal data obtained by finite element analysis. The proposed neural network to quantify damage severity achieved up to 95% success rate. A comparative study was done to evaluate the effect of boundary conditions on damage location. With the comparative study results, another neural network was proposed to locate damage position, achieving excellent results by successfully locating or by significantly reducing the search area. Both proposed ANNs use only frequency variation values as inputs, an easily obtainable quantity that requires few equipment to be acquired. The obtained results from these numerical examples indicate that the proposed approach can detect true damage locations and estimate damage magnitudes with satisfactory accuracy for this particular geometry, even under high measurement noise.

High-throughput method for detection and quantification of lesions on leaf scale based on trypan blue staining and digital image analysis

BackgroundField-grown leafy vegetables can be damaged by biotic and abiotic factors, or mechanically damaged by farming practices. Available methods to evaluate leaf tissue damage mainly rely on colour differentiation between healthy and damaged tissues. Alternatively, sophisticated equipment such as microscopy and hyperspectral cameras can be employed. Depending on the causal factor, colour change in the wounded area is not always induced and, by the time symptoms become visible, a plant can already be severely affected. To accurately detect and quantify damage on leaf scale, including microlesions, reliable differentiation between healthy and damaged tissue is essential. We stained whole leaves with trypan blue dye, which traverses compromised cell membranes but is not absorbed in viable cells, followed by automated quantification of damage on leaf scale.ResultsWe present a robust, fast and sensitive method for leaf-scale visualisation, accurate automated extraction and measurement of damaged area on leaves of leafy vegetables. The image analysis pipeline we developed automatically identifies leaf area and individual stained (lesion) areas down to cell level. As proof of principle, we tested the methodology for damage detection and quantification on two field-grown leafy vegetable species, spinach and Swiss chard.ConclusionsOur novel lesion quantification method can be used for detection of large (macro) or single-cell (micro) lesions on leaf scale, enabling quantification of lesions at any stage and without requiring symptoms to be in the visible spectrum. Quantifying the wounded area on leaf scale is necessary for generating prediction models for economic losses and produce shelf-life. In addition, risk assessments are based on accurate prediction of the relationship between leaf damage and infection rates by opportunistic pathogens and our method helps determine the severity of leaf damage at fine resolution.