• Smart Structures and Systems
• /
• 제29권1호
• /
• pp.77-91
• /
• 2022

Various monitoring systems have been implemented in civil infrastructure to ensure structural safety and integrity. In long-term monitoring, these systems generate a large amount of data, where anomalies are not unusual and can pose unique challenges for structural health monitoring applications, such as system identification and damage detection. Therefore, developing efficient techniques is quite essential to recognize the anomalies in monitoring data. In this study, several machine learning techniques are explored and implemented to detect and classify various types of data anomalies. A field dataset, which consists of one month long acceleration data obtained from a long-span cable-stayed bridge in China, is employed to examine the machine learning techniques for automated data anomaly detection. These techniques include the statistic-based pattern recognition network, spectrogram-based convolutional neural network, image-based time history convolutional neural network, image-based time-frequency hybrid convolution neural network (GoogLeNet), and proposed ensemble neural network model. The ensemble model deliberately combines different machine learning models to enhance anomaly classification performance. The results show that all these techniques can successfully detect and classify six types of data anomalies (i.e., missing, minor, outlier, square, trend, drift). Moreover, both image-based time history convolutional neural network and GoogLeNet are further investigated for the capability of autonomous online anomaly classification and found to effectively classify anomalies with decent performance. As seen in comparison with accuracy, the proposed ensemble neural network model outperforms the other three machine learning techniques. This study also evaluates the proposed ensemble neural network model to a blind test dataset. As found in the results, this ensemble model is effective for data anomaly detection and applicable for the signal characteristics changing over time.

• 비파괴검사학회지
• /
• 제36권3호
• /
• pp.211-216
• /
• 2016

에폭시 수지는 여러 산업분야에서 다양한 구조물의 접합과 제조에 사용된다. 구조물의 안전성과 접합재료의 최적 성능 확보를 위해서는 에폭시 수지의 경화 과정 모니터링 기반의 공정 제어가 요구된다. 광섬유 센서는 실과 같은 형태적 특징으로 인해 에폭시 수지의 경화 모니터링에 적합한 센서이다. 본 연구에서는 광섬유 브래그 격자 센서(fiber Bragg grating, FBG)를 이용하여 에폭시 수지의 경화 모니터링 연구를 수행하였다. 실제 실험을 통해, FBG를 기반으로 에폭시 수지의 경화과정에서 에폭시의 부피 수축에 의해 센서에 가해지는 변형률을 측정하고 온도의 변화에 의한 신호 변화를 보정하여 경화과정에서 발생하는 변형률의 정확한 모니터링이 가능함을 확인하였다. 추가적으로, 두 가지 에폭시 수지의 경화도 과정을 비교 분석하여 에폭시 종류에 따른 경화과정의 차이를 확인하였다. 결론적으로 FBG 센서가 에폭시 수지의 경화도 모니터링에 유용하다는 점을 확인하였다.

• Smart Structures and Systems
• /
• 제9권5호
• /
• pp.461-472
• /
• 2012

This study proposes an optics-based active sensing system for continuous monitoring of underground pipelines in nuclear power plants (NPPs). The proposed system generates and measures guided waves using a single laser source and optical cables. First, a tunable laser is used as a common power source for guided wave generation and sensing. This source laser beam is transmitted through an optical fiber, and the fiber is split into two. One of them is used to actuate macro fiber composite (MFC) transducers for guided wave generation, and the other optical fiber is used with fiber Bragg grating (FBG) sensors to measure guided wave responses. The MFC transducers placed along a circumferential direction of a pipe at one end generate longitudinal and flexural modes, and the corresponding responses are measured using FBG sensors instrumented in the same configuration at the other end. The generated guided waves interact with a defect, and this interaction causes changes in response signals. Then, a damage-sensitive feature is extracted from the response signals using the axi-symmetry nature of the measured pitch-catch signals. The feasibility of the proposed system has been examined through a laboratory experiment.

• Smart Structures and Systems
• /
• 제29권1호
• /
• pp.207-220
• /
• 2022

Structural health monitoring (SHM) plays a vital role in the maintenance and operation of constructions. In recent years, autonomous inspection has received considerable attention because conventional monitoring methods are inefficient and expensive to some extent. To develop autonomous inspection, a potential approach of crack identification is needed to locate defects. Therefore, this study exploits two deep learning-based segmentation models, DeepLabv3+ and Mask R-CNN, for crack segmentation because these two segmentation models can outperform other similar models on public datasets. Additionally, impacts of label quality on model performance are explored to obtain an empirical guideline on the preparation of image datasets. The influence of image cropping and label refining are also investigated, and different strategies are applied to the dataset, resulting in six alternated datasets. By conducting experiments with these datasets, the highest mean Intersection-over-Union (mIoU), 75%, is achieved by Mask R-CNN. The rise in the percentage of annotations by image cropping improves model performance while the label refining has opposite effects on the two models. As the label refining results in fewer error annotations of cracks, this modification enhances the performance of DeepLabv3+. Instead, the performance of Mask R-CNN decreases because fragmented annotations may mistake an instance as multiple instances. To sum up, both DeepLabv3+ and Mask R-CNN are capable of crack identification, and an empirical guideline on the data preparation is presented to strengthen identification successfulness via image cropping and label refining.

• 비파괴검사학회지
• /
• 제34권5호
• /
• pp.362-368
• /
• 2014

광섬유센서는 전자파 무간섭, 부식 방지, 다중화 등의 장점들을 갖고 있어 다양한 상태 감시 시스템을 위한 연구에 많이 활용되고 있다. 본 논문에서는 광섬유센서 기반의 인체 근육 상태 감시 시스템을 제안한다. 상용화되어 있는 인체 상태 감시 센서는 전자기 기반의 센서가 대부분이다. 이는 전자기 간섭 및 왜곡의 우려가 있어, 이를 보완하고 장치의 간소화 및 사용자 편의성을 위해 광섬유 브래그 격자센서를 사용하였다. 근육 상태의 지표가 되는 근육 수축 및 이완을 측정하기 위해 원주방향으로의 운동 감시가 가능한 밴드형태의 광섬유 브래그 격자센서 모듈을 제작하였다. 그리고 광섬유 브래그 격자센서 모듈의 적용성 평가를 위해 단축 인장시험을 수행하였다. 실험 결과 인장 크기에 따른 브래그 파장 변화가 상호 연관성을 보였으며, 이를 통해 브래그 격자센서 기반의 근육 상태 감시 시스템 개발의 가능성을 확인하였다.

• 대한기계학회논문집A
• /
• 제41권7호
• /
• pp.613-620
• /
• 2017

금속이 코팅된 FBG(fiber Bragg grating) 센서는 구조물이 과거에 겪은 최대 변형률을 기억하는 기억효과(memory effect)를 가진다. 본 연구에서는 무전해 도금법과 전해 도금법을 이용하여 약 $43{\mu}m$의 두께를 가지는 니켈(nickel)이 코팅된 FBG 센서를 제작하였다. 니켈 코팅된 FBG 센서의 잔류 변형률 생성 성능, 즉, 기억효과를 검증하기 위해 반복하중 실험(잔류 변형률 생성실험)을 수행하였다. 인가한 최대 변형률의 크기가 증가함에 따라 잔류 변형률이 증가함을 확인함으로써 기억효과를 검증하였다. 본 연구에서 수행한 니켈이 코팅된 FBG 센서의 제작 기법과 센서에 대한 반복하중 실험결과는 향후 광섬유 센서를 이용한 구조물 건전성 감시(SHM, structural health monitoring)기법 개발에 기본 데이터로서 활용될 것이다.

• 비파괴검사학회지
• /
• 제36권2호
• /
• pp.93-101
• /
• 2016

This paper presents the results obtained using time-series-based methods for structural damage assessment. The methods are applied to a wind turbine blade structure subjected to fatigue loads. A 9 m CX-100 (carbon experimental 100 kW) blade is harmonically excited at its first natural frequency to introduce a failure mode. Consequently, a through-thickness fatigue crack is visually identified at 8.5 million cycles. The time domain data from the piezoelectric active-sensing techniques are measured during the fatigue loadings and used to detect incipient damage. The damage-sensitive features, such as the first four moments and a normality indicator, are extracted from the time domain data. Time series autoregressive models with exogenous inputs are also implemented. These features could efficiently detect a fatigue crack and are less sensitive to operational variations than the other methods.

• Smart Structures and Systems
• /
• 제26권6호
• /
• pp.735-751
• /
• 2020

Identification of structure parameters is crucial in Structural Health Monitoring (SHM) context for activities such as model validation, damage assessment and signal processing of structure response. In this paper, guided waves generated by piezoelectric transducers are used for in-situ and non-destructive structural parameter estimation based on Bayesian approach. As Bayesian approach needs iterative process, which is computationally expensive, this paper proposes a method in which an analytical model is selected and developed in order to decrease computational time and complexity of modeling. An experimental set-up is implemented to estimate three target elastic and geometrical parameters: Young's modulus, Poisson ratio and thickness of aluminum and steel plates. Experimental and simulated data are combined in a Bayesian framework for parameter identification. A significant accuracy is achieved regarding estimation of target parameters with maximum error of 8, 11 and 17 percent respectively. Moreover, the limitation of analytical model concerning boundary reflections is addressed and managed experimentally. Pulse excitation is selected as it can excite the structure in a wide frequency range contrary to conventional tone burst excitation. The results show that the proposed non-destructive method can be used in service for estimation of material and geometrical properties of structure in industrial applications.

• Smart Structures and Systems
• /
• 제29권1호
• /
• pp.153-166
• /
• 2022

Identifying fine cracks in steel bridge facilities is a challenging task of structural health monitoring (SHM). This study proposed an end-to-end crack image segmentation framework based on a one-step Convolutional Neural Network (CNN) for pixel-level object recognition with high accuracy. To particularly address the challenges arising from small object detection in complex background, efforts were made in loss function selection aiming at sample imbalance and module modification in order to improve the generalization ability on complicated images. Specifically, loss functions were compared among alternatives including the Binary Cross Entropy (BCE), Focal, Tversky and Dice loss, with the last three specialized for biased sample distribution. Structural modifications with dilated convolution, Spatial Pyramid Pooling (SPP) and Feature Pyramid Network (FPN) were also performed to form a new backbone termed CrackDet. Models of various loss functions and feature extraction modules were trained on crack images and tested on full-scale images collected on steel box girders. The CNN model incorporated the classic U-Net as its backbone, and Dice loss as its loss function achieved the highest mean Intersection-over-Union (mIoU) of 0.7571 on full-scale pictures. In contrast, the best performance on cropped crack images was achieved by integrating CrackDet with Dice loss at a mIoU of 0.7670.

• Structural Engineering and Mechanics
• /
• 제88권1호
• /
• pp.83-94
• /
• 2023

An accurate and highly efficient inverse element labelled iPCB is developed based on the inverse finite element method (iFEM) for real-time shape estimation of plane-curved structures (such as arch bridges) utilizing onboard strain data. This inverse problem, named shape sensing, is vital for the design of smart structures and structural health monitoring (SHM) procedures. The iPCB formulation is defined based on a least-squares variational principle that employs curved Timoshenko beam theory as its baseline. The accurate strain-displacement relationship considering tension-bending coupling is used to establish theoretical and measured section strains. The displacement fields of the isoparametric element iPCB are interpolated utilizing nonuniform rational B-spline (NURBS) basis functions, enabling exact geometric modelling even with a very coarse mesh density. The present formulation is completely free from membrane and shear locking. Numerical validation examples for different curved structures subjected to different loading conditions have been performed and have demonstrated the excellent prediction capability of iPCBs. The present formulation has also been shown to be practical and robust since relatively accurate predictions can be obtained even omitting the shear deformation contributions and considering polluted strain measures. The current element offers a promising tool for real-time shape estimation of plane-curved structures.