• 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.

• 한국식물병리학회:학술대회논문집
• /
• 한국식물병리학회 2003년도 정기총회 및 추계학술발표회
• /
• pp.66.2-66
• /
• 2003

A gain-of-function mutant, SHM-11 obtained through gamma-ray mutagenesis, is resistant to rice blast caused by Magnaporthe grisea while wild type Sanghaehyanghyella is highly susceptible to the same disease. The resistance in the mutant was not race-specific when we tested with four races (KJ-201, KI-1113a, KI-313, KI-409) of M. grisea. To identify genes involved disease resistance in the gain-of-function mutant, genes specifically expressed in the mutant were selected by suppression subtractive hybridization using cDNAS of blast-inoculated mutant and wild type as a tester and a driver, respectively, Random 200 clones from the subtracted library were selected and analyzed by DNA sequencing. The sequenced genes represented three major groups related with disease resistance； genes encoding PR proteins, genes probably for phytoalexin biosynthesis, and genes involved in disease resistance signal transduction. A gene encoding a putative receptor-like protein kinase was identified as highly expressed only in the gain-of-function mutant after blast infection. The role of the putative receptor-like protein kinase gene during blast resistance will be further studied.

• Smart Structures and Systems
• /
• 제31권4호
• /
• pp.351-363
• /
• 2023

Visual structural inspections are an inseparable part of post-earthquake damage assessments. With unmanned aerial vehicles (UAVs) establishing a new frontier in visual inspections, there are major computational challenges in processing the collected massive amounts of high-resolution visual data. We propose twin deep learning models that can provide accurate high-resolution structural components and damage segmentation masks efficiently. The traditional approach to cope with high memory computational demands is to either uniformly downsample the raw images at the price of losing fine local details or cropping smaller parts of the images leading to a loss of global contextual information. Therefore, our twin models comprising Trainable Resizing for high-resolution Segmentation Network (TRS-Net) and DmgFormer approaches the global and local semantics from different perspectives. TRS-Net is a compound, high-resolution segmentation architecture equipped with learnable downsampler and upsampler modules to minimize information loss for optimal performance and efficiency. DmgFormer utilizes a transformer backbone and a convolutional decoder head with skip connections on a grid of crops aiming for high precision learning without downsizing. An augmented inference technique is used to boost performance further and reduce the possible loss of context due to grid cropping. Comprehensive experiments have been performed on the 3D physics-based graphics models (PBGMs) synthetic environments in the QuakeCity dataset. The proposed framework is evaluated using several metrics on three segmentation tasks: component type, component damage state, and global damage (crack, rebar, spalling). The models were developed as part of the 2^nd International Competition for Structural Health Monitoring.

• Smart Structures and Systems
• /
• 제31권4호
• /
• pp.393-407
• /
• 2023

To assess structural condition in a non-destructive manner, computer vision-based structural health monitoring (SHM) has become a focus. Compared to traditional contact-type sensors, the advantages of computer vision-based measurement systems include lower installation costs and broader measurement areas. In this study, we propose a novel computer vision-based vibration measurement and coarse-to-fine damage assessment method for truss bridges. First, a deep learning model FairMOT is introduced to track the regions of interest (ROIs) that include joints to enhance the automation performance compared with traditional target tracking algorithms. To calculate the displacement of the tracked ROIs accurately, a normalized cross-correlation method is adopted to fine-tune the offset, while the Harris corner matching is utilized to correct the vibration displacement errors caused by the non-parallel between the truss plane and the image plane. Then, based on the advantages of the stochastic damage locating vector (SDLV) and Bayesian inference-based stochastic model updating (BI-SMU), they are combined to achieve the coarse-to-fine localization of the truss bridge's damaged elements. Finally, the severity quantification of the damaged components is performed by the BI-SMU. The experiment results show that the proposed method can accurately recognize the vibration displacement and evaluate the structural damage.

• Smart Structures and Systems
• /
• 제31권4호
• /
• pp.365-381
• /
• 2023

The existing vision-based techniques for inspection and condition assessment of civil infrastructure are mostly manual and consequently time-consuming, expensive, subjective, and risky. As a viable alternative, researchers in the past resorted to deep learning-based autonomous damage detection algorithms for expedited post-disaster reconnaissance of structures. Although a number of automatic damage detection algorithms have been proposed, the scarcity of labeled training data remains a major concern. To address this issue, this study proposed a semi-supervised learning (SSL) framework based on consistency regularization and cross-supervision. Image data from post-earthquake reconnaissance, that contains cracks, spalling, and exposed rebars are used to evaluate the proposed solution. Experiments are carried out under different data partition protocols, and it is shown that the proposed SSL method can make use of unlabeled images to enhance the segmentation performance when limited amount of ground truth labels are provided. This study also proposes DeepLab-AASPP and modified versions of U-Net++ based on channel-wise attention mechanism to better segment the components and damage areas from images of reinforced concrete buildings. The channel-wise attention mechanism can effectively improve the performance of the network by dynamically scaling the feature maps so that the networks can focus on more informative feature maps in the concatenation layer. The proposed DeepLab-AASPP achieves the best performance on component segmentation and damage state segmentation tasks with mIoU scores of 0.9850 and 0.7032, respectively. For crack, spalling, and rebar segmentation tasks, modified U-Net++ obtains the best performance with Igou scores (excluding the background pixels) of 0.5449, 0.9375, and 0.5018, respectively. The proposed architectures win the second place in IC-SHM2021 competition in all five tasks of Project 2.

• Computers and Concrete
• /
• 제31권5호
• /
• pp.457-468
• /
• 2023

Buildings, bridges, and dams are examples of civil infrastructure that play an important role in public life. These structures are prone to structural variations over time as a result of external forces that might disrupt the operation of the structures, cause structural integrity issues, and raise safety concerns for the occupants. Therefore, monitoring the state of a structure, also known as structural health monitoring (SHM), is essential. Owing to the emergence of the fourth industrial revolution, next-generation sensors, such as wireless sensors, UAVs, and video cameras, have recently been utilized to improve the quality and efficiency of building forensics. This study presents a method that uses a target-based system to estimate the dynamic displacement and its corresponding dynamic properties of structures using UAV-based video. A laboratory experiment was performed to verify the tracking technique using a shaking table to excite an SDOF specimen and comparing the results between a laser distance sensor, accelerometer, and fixed camera. Then a field test was conducted to validate the proposed framework. One target marker is placed on the specimen, and another marker is attached to the ground, which serves as a stationary reference to account for the undesired UAV movement. The results from the UAV and stationary camera displayed a root mean square (RMS) error of 2.02% for the displacement, and after post-processing the displacement data using an OMA method, the identified natural frequency and damping ratio showed significant accuracy and similarities. The findings illustrate the capabilities and reliabilities of the methodology using UAV to evaluate the dynamic properties of structures.

• 한국구조물진단유지관리공학회 논문집
• /
• 제24권2호
• /
• pp.23-32
• /
• 2020

압축센싱 기술인 CAFB는 대상 구조물의 원시신호를 목적된 주파수 범위의 신호로 압축하여 획득하도록 개발되었다[27]. 이때 압축센싱을 위해 CAFB는 대상 구조물의 목적된 주파수 범위에 따라 다양한 기준신호로 최적화 될 수 있다. 또한, 최적화된 CAFB는 지진과 같은 돌발/위험상황에서도 대상 구조물의 유효한 구조응답을 효율적으로 압축할 수 있어야 한다. 본 논문에서는 상대적으로 유연한 구조물의 효율적인 구조 건전도 모니터링을 위하여 목적된 주파수 범위를 10Hz 미만으로 설정하고, 이를 위한 CAFB의 최적화 방법과 지진상황에서 CAFB의 지진응답성능을실험적으로 평가하였다. 이를 위해 본 논문에서는, 먼저 Kobe 지진파형을 이용하여 CAFB를 최적화하였고, 이를 자체 개발한 무선 IDAQ 시스템에 임베디드 하였다. 그리고, Kobe 지진파형을 이용하여 2경간 교량에 대한 지진응답실험을 수행하였다. 마지막으로 CAFB가 내장된 IDAQ 시스템을 이용하여 실시간으로 2경간 교량의 지진응답을 무선으로 획득하고, 획득된 압축신호는 원시신호와 상호 비교하였다. 실험의 결과로부터 압축신호는 원시신호와 대비하여 우수한 응답성능과 데이터 압축효과를 보였고, 또한 CAFB는 지진상황에서도 구조물의 유효한 구조응답을 효과적으로 압축센싱할 수 있었다. 최종적으로 본 논문에서는 목적된 주파수 범위(10Hz 미만)에 적합하도록 CAFB의 최적화 방법을 제시하였고, CAFB는 지진상황의 계측-모니터링을 위해 경제적이고 효율적인 데이터 압축센싱 기술임을 증명하였다.

• 비파괴검사학회지
• /
• 제36권5호
• /
• pp.345-352
• /
• 2016

높은 고도에서 운행되는 항공기는 -$50^{\circ}C$이하의 극저온 피로환경에 노출된다. 이때 반복하중을 통해 발생되는 크랙과 같은 미세결함은 항공기 구조물의 물성변화를 야기하고 구조물 파단과 같은 심각한 구조적 결함을 야기한다. 따라서 효율적인 구조물의 유지보수 및 수명 예측을 위해 구조물의 지속적인 상태진단이 필요하다. 본 연구에서는 실제 항공기 운행조건과 유사한 극저온 피로환경에서 항공기 날개의 구조 건전성 모니터링을 수행하였다. 초기 결함 탐지를 위해 사각배열 압전구동기 및 센서를 구조물 하단에 부착한 뒤, 유도초음파 기반 능동센싱 기법을 통해 손상에 의한 산란 및 반사파를 측정하였다. 이후 통계학적 모델 분석과 위상배열기법을 통해 손상 발생 시점을 파악 및 손상 위치 탐지를 실시하였다. 또한, 극저온 환경에서의 센서의 생존성 파악과 구조 건전성 모니터링 결과의 신뢰성 향상을 위해 센서자가진단을 실시하였다. 실험 결과, 제안된 기법을 통해 극한환경에서 운행되는 구조물의 초기 손상 탐지 및 손상 위치 탐지가 높은 정확도로 가능함을 확인하였다.

• Molecules and Cells
• /
• 제40권9호
• /
• pp.655-666
• /
• 2017

We constructed a large $na{\ddot{i}}ve$ human Fab library ($3{\times}10^{10}$ colonies) from the lymphocytes of 809 human donors, assessed available diversities of the heavy-chain variable (VH) and ${\kappa}$ light-chain variable (VK) domain repertoires, and validated the library by selecting Fabs against 10 therapeutically relevant antigens by phage display. We obtained a database of unique 7,373 VH and 41,804 VK sequences by 454 pyrosequencing, and analyzed the repertoires. The distribution of VH and VK subfamilies and germline genes in our antibody repertoires slightly differed from those in earlier published natural antibody libraries. The frequency of somatic hypermutations (SHMs) in heavy-chain complementarity determining region (HCDR)1 and HCDR2 are higher compared with the natural IgM repertoire. Analysis of position-specific SHMs in CDRs indicates that asparagine, threonine, arginine, aspartate and phenylalanine are the most frequent non-germline residues on the antibody-antigen interface and are converted mostly from the germline residues, which are highly represented in germline SHM hotspots. The amino acid composition and length-dependent changes in amino acid frequencies of HCDR3 are similar to those in previous reports, except that frequencies of aspartate and phenylalanine are a little higher in our repertoire. Taken together, the results show that this antibody library shares common features of natural antibody repertoires and also has unique features. The antibody library will be useful in the generation of human antibodies against diverse antigens, and the information about the diversity of natural antibody repertoires will be valuable in the future design of synthetic human antibody libraries with high functional diversity.

• 한국전산구조공학회논문집
• /
• 제22권5호
• /
• pp.445-452
• /
• 2009

본 논문에서는 유한요소모델개선기법을 이용하여 골조구조물의 손상탐지를 실시하였다. 유한요소모델개선기법은 계측된 동특성을 모사하는 구조해석모델을 구하는 방법으로서 손상탐지 및 구조건전도감시를 위해 효과적으로 이용될 수 있다. 유한요소모델개선기법에는 다양한 종류의 동특성데이터가 사용될 수 있는데, 본 연구에서는 고유진동수와 모드형상, 주파수응답함수를 사용한 모델 개선식을 유도한 후, 고유진동수와 모드형상, 고유진동수와 주파수응답함수식을 조합한 경우에 대해 실험실 규모의 구조물의 손상위치 및 손상정도를 추정하였다. 구조물은 4층 철골조 구조물로서 진동대를 이용하여 원구조물에 백색잡음 가진실험을 실시한 후 손상의 모사를 위해 1층 부분의 보 부재를 다양한 단면의 부재로 교체하고 동일한 실험을 반복하였다. 보 부재 교체 전 후에 계측된 데이터와 두 종류의 모델개선기법을 각각 적용하여 손상탐지를 실시한 후, 손상위치 및 손상정도에 대한 정확도를 분석하였다. 분석결과 고유진동수+모드형상을 사용한 경우보다 고유진동수+주파수응답함수를 사용한 경우 더욱 정확히 손상을 탐지할 수 있었다.