Ruigen Yao

Data-driven methods for threshold determination in time- series based damage detection

Structural vibration monitoring has received a lot of attention from the research community in the past few years. The objective is to create automatic structural assessment techniques that can be realized through programmed vibration analysis. Till now many vibration-based damage features have been proposed, yet to truly automate the damage identification process, reliable damage threshold construction techniques are also need. In this paper, two data-driven methods based on resampling and nearest neighbor rule are applied for threshold construction for damage features from autoregression (AR) analysis of vibration signals. Both threshold calculation techniques are rooted in empirical feature probability estimation. The proposed thresholds are then tested on features extracted acceleration measurements collected from a 5 DOF test specimen. The resampling method is applied to Mahalanobis distance of AR model coefficients, while the nearest neighbor rule is used on a combination of coefficient distance feature and the residual autocorrelation feature. Both methods perform well in this case study.

Iterative Spatial Compressive Sensing Strategy for Structural Damage Diagnosis as a BIG DATA Problem

Accurate structural damage identification calls for dense sensor networks, which are becoming more feasible as the price of electronic sensing systems reduces. To transmit and process data from all nodes of a dense network would be an onerous task which creates a BIG DATA problem; therefore scalable algorithms are needed so that decision on the current state of the structure can be made based on efficient data processing. In this paper, an iterative spatial compressive sensing scheme for damage existence identification and localization will be introduced. At each iteration, a subset of sensors is selected for data transmission and relevant information will be extracted at central station for damage existence identification/localization. This information will also provide useful guidance in future selection of sensing locations. The devised algorithm is applied to identify damage in a simulated gusset plate.

Data-Driven Structural Damage Identification Using DIT

Vibration-based damage detection research aims to develop efficient algorithms to identify structural damage from monitoring data. One of the main categories of such algorithms is data-driven techniques which extract features from measured signals, and identify the damage by evaluating the significance of potential changes in these features. This paper presents application of several data-driven damage identification methodologies on a multivariate simulated data set. First, general regression models are applied to data collected through clusters of sensors and damage sensitive features are extracted. For systems with linear topology, it is shown that substructural regression modeling can also be performed on time- and frequency-domain transforms of the measured signals to estimate local stiffness of the structure as damage features. Subsequently, change detection techniques are utilized to statistically determine the significance of changes in the extracted features in order to distinguish between assignable changes as a result of damage and common changes due to environmental factors. Finally, a toolsuite is developed to facilitate application of the developed algorithms and improve the damage identification performance through incorporation of various combinations of regression models, damage features and statistical tests.

Multisensor Aggregation Algorithms for Structural Damage Diagnosis Based on a Substructure Concept

One of the important goals of structural health monitoring is damage detection. Although many methods have been proposed to detect the existence of structural damage, relatively few studies are found on higher-level damage diagnosis such as identification of the location and extent of damage. In this paper, multiple substructural damage identification models based on regression between internal responses and boundary responses of individual beam elements in either plane or three-dimensional space are derived. Three damage indexes are defined from regression model characteristics, and two change-point analysis methods are adopted to capture changes in damage index sequences which are extracted from structural monitoring data sets from healthy and unknown states. Possible damage locations are identified as where the most significant changes in the damage indexes occur, and a voting scheme is used to synthesize the results from different algorithms. This damage detection approach is straightforward and efficient, with the regression coefficients directly related to the structural stiffness properties. The numerical and experimental application results show that the method successfully identifies and locates structural change in most of the cases.

Regression-Based Algorithms for Structural Damage Identification and Localization

Early damage detection and localization is very important for maintenance and retrofit of civil structures. In the past decades, a lot of research has been conducted on structural condition prognosis using vibration measurements, which can be very conveniently procured in large quantities at a moderate cost. Many of these approaches, however, concern only the identification of structural damage existence, and do not attempt higher level damage detection. In this paper, three regressionbased damage detection algorithms are be presented and applied for damage identification in a two-span steel girder in the lab. All of them can perform local damage detection and evaluation to a certain extent. They have different modeling complexities, and thus have different performance levels. Damage identification/localization/severity evaluation results obtained from these algorithms are compared and contrasted.

Structural Damage Detection Using Multivariate Time Series Analysis

Much research has been focused in the past few decades on data-driven structural health monitoring based on sensor measurements. Modal parameters from system identification are the most widely studied structural state indicators adopted for this purpose; however, recent research has showed that they are not sensitive enough to local damage. In an effort to seek more effective alternatives, univariate autoregressive (AR) modeling on structural response has been investigated in several publications, where model characteristics are used as damage indices. Although these methods are generally successful, they tend to generate false alarms when the environmental conditions are varying because responses from only one location/sensor are considered. To strike a balance between sensitivity and stability, in this paper autoregressive with exogenous input modeling on measurements from several adjacent sensing channels is presented and applied to detect damage in a space truss structure. The damage feature is extracted from the residuals obtained via fitting the baseline model to data from the current structure. Also, damage localization is attempted by examining the estimated mutual information statistic between data from adjacent sensing channels. The damage identification/localization results thus obtained are then compared to those from univariate AR modeling to evaluate their relative pros and cons.

Noise Sensitivity Evaluation of Autoregressive Features Extracted from Structure Vibration

In the past few decades many types of structural damage indices based on structural health monitoring signals have been proposed, requiring performance evaluation and comparison studies on these indices in a quantitative manner. One tool to help accomplish this objective is analytical sensitivity analysis, which has been successfully used to evaluate the influences of system operational parameters on observable characteristics in many fields of study. In this chapter, the sensitivity expressions of two damage features, namely the Mahalanobis distance of autoregressive coefficients and Cosh distance of autoregressive spectra, will be derived with respect to the measurement noise level. The effectiveness of the proposed methods is illustrated in a numerical case study on a 10 DOF system, where their results are compared with those from direct simulation and theoretical calculation.

Structural Damage Localization Using Sensor Cluster Based Regression Schemes

Automatic damage identification from sensor measurements has long been a topic of interest in the civil engineering research community. A number of methods, including classical system identification and time series analysis techniques, have been proposed to detect the existence of damage in structures. Not many of them, though, are reported efficient for higher-level damage detection which concerns damage localization and severity assessment. In this paper, regression-based damage localization schemes are proposed and applied to signals generated from a simulated two-bay steel frame. These regression algorithms operates on substructural beam models, and uses the acceleration/strain responses at beam ends as input and the acceleration from an intermediate node as output. From the regression coefficients and residuals three damage identification features are extracted, and two change point analysis techniques are adopted to evaluate if a change of statistical significance occurred in the extracted feature sequences. For the four damage scenarios simulated, the algorithms identified the damage existence and partially succeeded in locating the damage. More accurate inferences on damage location are drawn by combining the results from different algorithms using a weighted voting scheme.