Search

The parametric approach to eliminating the temperature-caused modal variability in vibration-based structural damage detection requires a correlation model between the modal properties and environmental temperatures. This paper examines the generalization capability of neural network models, established using long-term monitoring data, for correlation between the modal frequencies and environmental temperatures. A total of 770 h modal frequency and temperature data obtained from an instrumented bridge are available for this study, which are further divided into three sets: training data, validation data, and testing data. A two-layer back-propagation neural network (BPNN) is first trained using the training data by the conventional training algorithm, in which the number of hidden nodes is optimally determined using the validation data. Then two new BPNNs are configured with the same data by applying the early stopping technique and the Bayesian regularization technique, respectively. The reproduction and prediction capabilities of the two new BPNNs are examined in respect of the training data and the unseen testing data, and compared with the performance of the baseline BPNN model. This study indicates that both the early stopping and Bayesian regularization techniques can significantly ameliorate the generalization capability of BPNN-based correlation models, and the BPNN model formulated using the early stopping technique outperforms that using the Bayesian regularization technique in both reproduction and prediction capabilities. [ABSTRACT FROM AUTHOR]

While numerous studies have been published concerning the application of a variety of system identification techniques in conjunction with vibration measurements from civil infrastructure systems, there is a paucity of publications addressing the influence of algorithm-specific control parameters that impact the correct and efficient application of the selected identification scheme. Furthermore, as dense sensor arrays become widely accessible in civil infrastructure applications, voluminous amounts of multichannel data streams are becoming available for processing, thus imposing new demands on identification procedures regarding high-dimensionality (in both the spatial as well as the temporal domains) requirements that may render some methods inapplicable if careful attention is not paid to practical implementation issues. This paper provides a comprehensive study of three time-domain identification algorithms applied in conjunction with the Natural Excitation Technique in order to extract the modal parameters of a newly constructed long-span bridge that was monitored, in its virgin state, over a relatively long period of time with a state-of-the-art dense sensor array. The three methods used are: the eigensystem realization algorithm (ERA), the ERA with data correlations, and the least squares algorithm. One of the critical issues in the mentioned algorithms, is selection of the reference degree-of-freedom (DOF). Previous experiences have shown that one cannot rely on a single reference DOF for identification of all modes. Consequently, the aforementioned identification formulations were modified to include multiple reference DOF, simultaneously, or one at a time. An autonomous algorithm was presented to distinguish the genuine structural modes from spurious noise or computational modes. Based on some parameter studies, some useful guidelines for the selection of critical user-selectable parameters are presented. [ABSTRACT FROM AUTHOR]

Bridge engineers need to group bridges to determine adequate let projects based on their proximities, types of work, costs, and cost constraints for each clustered let project after bridges to be treated yearly are identified. This process is time-consuming and typically performed manually. First, this paper presents the formulation of a bridge clustering problem for determining let projects by considering the bridge proximity and type of work with preference membership functions to simulate the actual decision-making process. Second, a constrained fuzzy c-mean (FCM) clustering algorithm is presented to resolves this problem. A case study using the subset of bridges in the state of Georgia with the hypothetical treatments and costs was used to test the developed algorithm and to demonstrate its capability. The results show that the developed constrained FCM clustering algorithm can, in seconds, effectively determine adequate let projects by clustering bridges while meeting cost constraints. The presented formulation also allows incorporation of additional factors such as preference of clustering bridges with the same route number or same route type that are important to other state Departments of Transportation. Finally, conclusions about the benefits and characteristics of the developed algorithm are summarized, and recommendations for future research are discussed. [ABSTRACT FROM AUTHOR]

A nondestructive crack detection (NCD) scheme to locate a crack and estimate the crack size by using changes in the modal parameters of a class of bridges is presented in this paper. First, the theory underlying a nondestructive crack localization and crack-size estimation algorithm for beam-type structures is outlined. Next, a field experiment to verify the feasibility and applicability of the NCD scheme is detailed. In that experiment, predamage and postdamage modal data extracted from a full-scale plate–girder bridge are described. The modal data are fed into the NCD scheme that localizes crack and estimates crack size in the test structure. The result of the analysis indicates that the proposed NCD algorithm correctly locates the crack and accurately estimates the size of the crack. [ABSTRACT FROM AUTHOR]