Eloi Figueiredo

Machine learning algorithms for damage detection under operational and environmental variability

The goal of this article is to detect structural damage in the presence of operational and environmental variations using vibration-based damage identification procedures. For this purpose, four machine learning algorithms are applied based on the auto-associative neural network, factor analysis, Mahalanobis distance, and singular value decomposition. A base-excited three-story frame structure was tested in laboratory environment to obtain time-series data from an array of accelerometers under several structural state conditions. Tests were performed with varying stiffness and mass conditions with the assumption that these sources of variability are representative of changing operational and environmental conditions. Damage is simulated through nonlinear effects introduced by a bumper mechanism that induces a repetitive, impact-type nonlinearity. This mechanism intends to simulate the cracks that open and close under dynamic loads or loose connections that rattle. The unique contribution of this study is a direct comparison of the four proposed machine learning algorithms that have been reported as reliable approaches to separate structural conditions with changes resulting from damage from changes caused by operational and environmental variations.

A mobile-agent-based wireless sensing network for structural monitoring applications

A new wireless sensing network paradigm is presented for structural monitoring applications. In this approach, both power and data interrogation commands are conveyed via a mobile agent that is sent to sensor nodes to perform intended interrogations, which can alleviate several limitations of the traditional sensing networks. Furthermore, the mobile agent provides computational power to make near real-time assessments on the structural conditions. This paper will discuss such prototype systems, which are used to interrogate impedance-based sensors for structural health monitoring applications. Our wireless sensor node is specifically designed to accept various energy sources, including wireless energy transmission, and to be wirelessly triggered on an as-needed basis by the mobile agent or other sensor nodes. The capabilities of this proposed sensing network paradigm are demonstrated in the laboratory and the field.

Structural health monitoring algorithm comparisons using standard data sets

The real-world structures are subjected to operational and environmental condition changes that impose difficulties in detecting and identifying structural damage. The aim of this report is to detect damage with the presence of such operational and environmental condition changes through the application of the Los Alamos National Laboratory’s statistical pattern recognition paradigm for structural health monitoring (SHM). The test structure is a laboratory three-story building, and the damage is simulated through nonlinear effects introduced by a bumper mechanism that simulates a repetitive impact-type nonlinearity. The report reviews and illustrates various statistical principles that have had wide application in many engineering fields. The intent is to provide the reader with an introduction to feature extraction and statistical modelling for feature classification in the context of SHM. In this process, the strengths and limitations of some actual statistical techniques used to detect damage in the structures are discussed. In the hierarchical structure of damage detection, this report is only concerned with the first step of the damage detection strategy, which is the evaluation of the existence of damage in the structure. The data from this study and a detailed description of the test structure are available for download at: http://institute.lanl.gov/ei/software-and-data/.

Influence of the Autoregressive Model Order on Damage Detection

Abstract: An important step for using time-series autoregressive (AR) models for structural health monitoring is the estimation of the appropriate model order. To obtain an optimal AR model order for such processes, this article presents and discusses four techniques based on Akaike information criterion, partial autocorrelation function, root mean squared error, and singular value decomposition. A unique contribution of this work is to provide a comparative study with three different AR models that is carried out to understand the influence of the model order on the damage detection process in the presence of simulated operational and environmental variability. A three-story base-excited frame structure was used as a test bed in a laboratory setting, and data sets were measured for several structural state conditions. Damage was introduced by a bumper mechanism that induces a repetitive impact-type nonlinearity. The operational and environmental effects were simulated by adding mass and by changing the stiffness properties of the columns. It was found that these four techniques do not converge to a unique solution, rather all require somewhat qualitative interpretation to define the optimal model order. The comparative study carried out on these data sets shows that the AR model order range defined by the four techniques provides robust damage detection in the presence of simulated operational and environmental variability.

A novel unsupervised approach based on a genetic algorithm for structural damage detection in bridges

This paper proposes a novel unsupervised and nonparametric genetic algorithm for decision boundary analysis (GADBA) to support the structural damage detection process, even in the presence of linear and nonlinear effects caused by operational and environmental variability. This approach is rooted in the search of an optimal number of clusters in the feature space, representing the main state conditions of a structural system, also known as the main structural components. This genetic-based clustering approach is supported by a novel concentric hypersphere algorithm to regularize the number of clusters and mitigate the cluster redundancy. The superiority of the GADBA is compared to state-of-the-art approaches based on the Gaussian mixture models and the Mahalanobis squared distance, on data sets from monitoring systems installed on two bridges: the Z-24 Bridge and the Tamar Bridge. The results demonstrate that the proposed approach is more efficient in the task of fitting the normal condition and its structural components. This technique also revealed to have better classification performance than the alternative ones in terms of false-positive and false-negative indications of damage, suggesting its applicability for real-world structural health monitoring applications.

A global expectation-maximization based on memetic swarm optimization for structural damage detection

During the service life of engineering structures, structural management systems attempt to manage all the information derived from regular inspections, evaluations and maintenance activities. However, the structural management systems still rely deeply on qualitative and visual inspections, which may impact the structural evaluation and, consequently, the maintenance decisions as well as the avoidance of collapses. Meanwhile, structural health monitoring arises as an effective discipline to aid the structural management, providing more reliable and quantitative information; herein, the machine learning algorithms have been implemented to expose structural anomalies from monitoring data. In particular, the Gaussian mixture models, supported by the expectation-maximization (EM) algorithm for parameter estimation, have been proposed to model the main clusters that correspond to the normal and stable state conditions of a structure when influenced by several sources of operational and environmental variations. Unfortunately, the optimal parameters determined by the EM algorithm are heavily dependent on the choice of the initial parameters. Therefore, this paper proposes a memetic algorithm based on particle swarm optimization (PSO) to improve the stability and reliability of the EM algorithm, a global EM (GEM-PSO), in searching for the optimal number of components (or data clusters) and their parameters, which enhances the damage classification performance. The superiority of the GEM-PSO approach over the state-of-the-art ones is attested on damage detection strategies implemented through the Mahalanobis and Euclidean distances, which permit one to track the outlier formation in relation to the main clusters, using real-world data sets from the Z-24 Bridge (Switzerland) and Tamar Bridge (United Kingdom).

Damage Detection and Quantification Using Transmissibility Coherence Analysis

A new transmissibility-based damage detection and quantification approach is proposed. Based on the operational modal analysis, the transmissibility is extracted from system responses and transmissibility coherence is defined and analyzed. Afterwards, a sensitive-damage indicator is defined in order to detect and identify the severity of damage and compared with an indicator developed by other authors. The proposed approach is validated on data from a physics-based numerical model as well as experimental data from a three-story aluminum frame structure. For both numerical simulation and experiment the results of the new indicator reveal a better performance than coherence measure proposed in Rizos et al., 2008, Rizos et al., 2002, Fassois and Sakellariou, 2007, especially when nonlinearity occurs, which might be further used in real engineering. The main contribution of this study is the construction of the relation between transmissibility coherence and frequency response function coherence and the construction of an effective indicator based on the transmissibility modal assurance criteria for damage (especially for minor nonlinearity) detection as well as quantification.

Single side damage simulations and detection in beam-like structures

Beam-like structures are the most common components in real engineering, while single side damage is often encountered. In this study, a numerical analysis of single side damage in a free-free beam is analysed with three different finite element models; namely solid, shell and beam models for demonstrating their performance in simulating real structures. Similar to experiment, damage is introduced into one side of the beam, and natural frequencies are extracted from the simulations and compared with experimental and analytical results. Mode shapes are also analysed with modal assurance criterion. The results from simulations reveal a good performance of the three models in extracting natural frequencies, and solid model performs better than shell while shell model performs better than beam model under intact state. For damaged states, the natural frequencies captured from solid model show more sensitivity to damage severity than shell model and shell model performs similar to the beam model in distinguishing damage. The main contribution of this paper is to perform a comparison between three finite element models and experimental data as well as analytical solutions. The finite element results show a relatively well performance.

Applicability of linear and nonlinear principal component analysis for damage detection

The goal of this work is to detect structural damage using vibration-based damage identification approaches even when the damage-sensitive features are camouflaged by the presence of operational and environmental conditions. For feature classification purposes, four machine learning algorithms are applied based on the principal component analysis (PCA), nonlinear PCA, kernel PCA and greedy kernel PCA. Time-series data from an array of accelerometers under several structural state conditions were obtained from a well-known base-excited three-story frame structure. The main contribution of this work is the applicability of those PCA-based algorithms, for damage detection, in the presence of operational and environmental effects. For these specific data sets, one can infer that the greedy kernel PCA algorithm is more appropriate when one wants to minimize false-positive indications of damage without increasing, significantly, the false-negative indications of damage.

Time series predictive models of piezoelectric active-sensing for SHM applications

In this paper, the use of time domain data from piezoelectric active-sensing techniques is investigated for structural health monitoring (SHM) applications. Piezoelectric transducers have been increasingly used in SHM because of their proven advantages. Especially, the use of known and repeatable inputs at high frequency ranges makes the development of SHM signal processing algorithm easier and more efficient. However, to date, most of these techniques have been based on frequency domain analyses, such as impedance-based or high-frequency response functions (FRF) -based SHM techniques. Even with Lamb wave propagations, most researchers adopt frequency domain or wavelets analysis for damage-sensitive feature extraction. This process usually requires excessive averaging to reduce measurement noise and more computational resources, which is not ideal from both memory and power consumption standpoints. Therefore in this study, we investigate the use of autoregressive models with exogenous inputs (ARX) with the measured time series data from piezoelectric active-sensors. The test structure considered in this study is a composite plate, where several damage conditions were manually imposed. The performance of this technique is compared to that of traditional autoregressive models, traditionally used in low-frequency passive sensing SHM applications, and that of FRF-based analysis, and its superior capability for SHM is demonstrated.