A deep‐learning approach for health monitoring of a steel frame structure with bolted connections

Abstract

This study is motivated by the need to develop a data-driven deep-learning approach for vibration-based structural health monitoring of a steel frame structure with bolted connections. A convolutional-neural-network-based deep-learning architecture is designed and trained to extract discriminative features from the vibration-based time-frequency scalogram images and use those to distinguish the undamaged and damaged cases of the targeted frame structure. Different damage and undamaged classes corresponding to the loosening of bolts are categorized as fully loose, hand tight, and fully tight (undamaged) conditions. The average training and validation accuracy were found to be 100% and 98.1%, respectively. In order to check the performance and robustness of the technique, testing is carried out for an unseen dataset corresponding to the training classes as well as some additional cases close to the training classes. The proposed deep-learning approach can successfully classify the damage classes with high testing accuracy that demonstrates its efficacy as an automation tool for health monitoring of connections of plane frame structures.convolutional neural network, deep learning, loosening of bolts, steel frame structure, structural health monitoring | INTRODUCTIONDue to the higher material reliability property and speed of construction, structural steel is extensively used to construct bridges, buildings, industrial structures, and offshore structures. The members are generally connected by welds or bolts. These bolted joints act as semi-rigid connections, which absorb significant energy under the excitation of wind and/or earthquake forces. However, because of energy absorption, the bolts may get loosened, which changes the subsequent behavior of the structure.A review on vibration-based damage detection methods in civil engineering structures with the applications from traditional to machine learning (ML) to deep-learning techniques is presented in Avci et al. 1 Various traditional techniques for structural health monitoring (SHM) of different type of civil engineering structures are discussed in previous studies. [2][3][4][5][6][7] Some applications of traditional SHM techniques on frame structures are studied and reported in other studies. 3,5,8,9 Modern signal processing tools like wavelet transform are widely utilized on various responses of frame structures for identification and quantification of damage. [10][11][12][13] The model updating based damage detection techniques are found to be gained popularity in the area of SHM for the last two decades. Some studies on various framed structures

Vision-Based Detection of Bolt Loosening Using YOLOv5

“…Pham et al [ 45 ] used composite bolt images generated by graphical models as datasets trained by a neural network, which is helpful in reducing the time and cost of collecting high-quality training data. Pal et al [ 46 ] extracted identification features using convolutional neural network (CNN) from time-frequency scale images based on vibration to detect bolt loosening. The average accuracy of the method is respectively 100% and 98.1%.…”

Multiclass damage detection in concrete structures using a transfer learning‐based generative adversarial networks

“…Furthermore, the inherent nature of image‐based classification results in only surface‐level detections being considered, with no analysis of the internal features of the structure being conducted 21 . This limitation was addressed by a CNN architecture proposed by Pal et al, 15 which was able to classify the looseness of steel bolt connections from time‐frequency scalogram images generated from vibration signals. Lastly, few studies have developed CNN models which are capable of quantifying the physical parameters of the damage and the subsequent impact on the structure.…”

“…Presently, numerous studies have been conducted which implement various CNN architectures to classify postdisaster or end-of-life cycle damage for various structures. Broadly, CNNs can be divided into three broad categories based on the classification detail the architecture achieves for 2D images, including full-scale, 14,15 region-based, 16,17 and pixel-level 18,19 evaluations. Traditional CNNs are characterized by their ability to classify images at a full-scale level, which includes all features of the image, whose classification is often based on the image label (crack, undamaged, etc.).…”

Impedance‐based looseness detection of bolted joints using artificial neural network: An experimental study

“…Zhang et al 20 presented an analytical model to predict the EM impedance response of a cracked Timoshenko beam structure. Pal et al 21 presented a convolutional neural network (CNN)‐based deep learning architecture for vibration‐based SHM of bolted steel frame structures. Zhao et al 22 presented a multiharmonic electrical impedance tomography (MH‐EIT)‐based imaging methodology for NDT of incipient fatigue cracks and delamination type damages.…”