• KSCE Journal of Civil and Environmental Engineering Research
• /
• v.42 no.1
• /
• pp.107-116
• /
• 2022

Recently, research has been actively conducted on the technology of inspection facilities through image-based analysis assessment of human-inaccessible facilities. This research was conducted to study the conditions of deep learning-based imaging data on bridges and to develop an evaluation prototype program for bridges. To develop a deep learning-based bridge damage detection prototype, the Semantic Segmentation model, which enables damage detection and quantification among deep learning models, applied Mask-RCNN and constructed learning data 5,140 (including open-data) and labeling suitable for damage types. As a result of performance modeling verification, precision and reproduction rate analysis of concrete cracks, stripping/slapping, rebar exposure and paint stripping showed that the precision was 95.2 %, and the recall was 93.8 %. A 2^nd performance verification was performed on onsite data of crack concrete using damage rate of bridge members.

• KSCE Journal of Civil and Environmental Engineering Research
• /
• v.43 no.4
• /
• pp.511-523
• /
• 2023

Research on the integration of unmanned aerial vehicles and deep learning for reinforced concrete damage detection is actively underway. Convolutional neural networks have a high impact on the performance of image classification, detection, and segmentation as backbones. The MobileNet, a pre-trained convolutional neural network, is efficient as a backbone for an unmanned aerial vehicle-based damage detection model because it can achieve sufficient accuracy with low computational complexity. Analyzing vanilla convolutional neural networks and MobileNet under various conditions, MobileNet was evaluated to have a verification accuracy 6.0~9.0% higher than vanilla convolutional neural networks with 15.9~22.9% lower computational complexity. MobileNetV2, MobileNetV3Large and MobileNetV3Small showed almost identical maximum verification accuracy, and the optimal conditions for MobileNet's reinforced concrete damage image feature extraction were analyzed to be the optimizer RMSprop, no dropout, and average pooling. The maximum validation accuracy of 75.49% for 7 types of damage detection based on MobilenetV2 derived in this study can be improved by image accumulation and continuous learning.

• Transactions of the Korean Society of Mechanical Engineers A
• /
• v.37 no.7
• /
• pp.891-897
• /
• 2013

In this study, the compressive deformation behaviors of aluminum alloy under high strain rates were investigated by means of a SHPB test. An acoustic emission (AE) technique was also employed to monitor the signals detected from the deformation during the entire impact by using an AE sensor connected to the specimen with a waveguide in real time. AE signals were analyzed in terms of AE amplitude, AE energy and peak frequency. The impacted specimen surface and side area were observed after the test to identify the particular features in the AE signal corresponding to the specific types of damage mechanisms. As the strain increased, the AE amplitude and AE energy increased whereas the AE peak frequency decreased. It was elucidated that each AE signal was closely associated with the specific damage mechanism in the material.

• Composites Research
• /
• v.13 no.1
• /
• pp.89-97
• /
• 2000

Fatigue damage detection and vibration sensing for a laminated composites and impact location detection for a steel beam have been carried out using optical fiber sensor. Intensity based optical fiber sensor is constructed by placing two cleaved fiber end in a hollow glass tube, and multiple reflection within the cavity is considered. Fatigue signals are measured by embedded optical fiber, surface mounted optical fiber sensor and strain gage simultaneously. For vibration sensing, optical fiber sensor is mounted on the carbon fiber composite beam and its response to free vibration and forced vibration is investigated. In impact location detection, two optical fiber sensors are used and the information obtained from two sensors is arrival time delay of vibration caused by impact. Impact location can be calculated from this time delay. The obtained results show that the intensity based optical fiber sensor provide reliable data during long-term fatigue loading, unlike strain gage which deteriorate during the early part of the fatigue test. Optical fiber sensor signals coincide with gap sensor in vibration sensing. The precise locations of impact can be detected within 4.1% error limit.

• Journal of the Korean Society of Propulsion Engineers
• /
• v.5 no.4
• /
• pp.20-30
• /
• 2001

This study represents the Nondestructive Test about impact damage of composite materials made by different lay-up patterns and degrees. For this study, they were examined by the drop test on composite materials of two type lap-up patterns with fabric and unidirectional prepreg and examined nondestructive test of those. Nondestructive methods were X-ray test with $ZnI_2$ penetrant and Ultrasonic C-scan. The defect detectability of X-ray and Ultrasonic test was compared according to defect species. And the amounts of damage on impacted zone wert compared according to impact energy on two type test specimens. At results, Ultrasonic test was more effective to detect delamination and Penetrant X-ray test was more effective to detect matrix crack and fiber fracture. There were some differences in defect shapes and grades according to lay-up patterns and degrees, and the trend appeared that matrix crack, delamination, fiber fracture occured and increasing defects sizes according to increasing impact energy.

• Journal of the Korea institute for structural maintenance and inspection
• /
• v.13 no.5 s.57
• /
• pp.169-179
• /
• 2009

The purpose of this paper is to find an efficient heat source to amplify the surface temperature of damaged concrete structures for infrared thermography. we compare two different heat sources of far-infrared lamp and halogen lamp each other for their efficiency. The two heat sources were applied to the concrete specimens. Two different concrete specimens were used: one was the concrete containing internal void and the other was wrapped with partially unbonded fiber reinforced polymer sheet. it was found that the far-infrared lamp was more efficient than the halogen lamp. In addition, we propose a new algorithm to make the damage zone displayed clear in the image obtained from the thermographic operation. The algorithm is a combination of Gauss filtering process and the Prewitt mask operation.

• Research in Plant Disease
• /
• v.23 no.3
• /
• pp.278-282
• /
• 2017

Maize is one of the most cultivated cereals as a staple food in the world. The harvested maize is mainly stored after drying, but its quality and nutrition could be debased by fungal spoilage and mycotoxin contamination. In this study, we surveyed mycotoxin contamination fungal infection of maize kernels that were stored for almost one year after harvest in 2015. The amount of deoxynivalenol and zearalenone detected were higher than the other mycotoxin, such as aflatoxin, ochratoxin, fumonisin and T-2 toxin. In particular, level of deoxynivalenol was detected as $1200{\pm}610{\mu}g/kg$ in small size kernels, which was four to six times higher than the large and the medium size kernels. Moreover, the amount of deoxynivalenol, zearalenone, and fumonisin were increased with discolored kernels. 10 species including Fusarium spp., Aspergillus spp. and Penicillium spp. were isolated from the maize kernels. F. graminearum was predominant in the discolored kernels with detection rates of 60% (red) and 40% (brown). Our study shows that the mycotoxin contents of stored maize can be increased by discolored maize kernels mixed. Therefore elimination of the contaminated maize kernels will help prevent fungal infection and mycotoxin contamination in stored maize.

• Korean Journal of Optics and Photonics
• /
• v.31 no.5
• /
• pp.205-212
• /
• 2020

In the performance of a wind turbine system, the blades play a vital role. However, they are susceptible to damage arising from complex and irregular loading (which may even cause catastrophic collapse), and they are expensive to maintain. Therefore, it is very important both to find defects after blade manufacturing is completed and to find damage after the blade is used for a certain period of time. This study provides a new perspective for the detection of internal defects in glass-fiber- and carbon-fiber-reinforced panels, which are used as the main materials in wind turbine blades. A gap or fracture between fiber-reinforced materials, which may occur during blade manufacturing or operation, is simulated by drilling a hole 5 mm in diameter in the middle layer of the laminated material. Then, a digital-image-correlation (DIC) method is used to detect internal defects in the blade. Tensile load is applied to the fabricated specimen using a tensile tester, and the generated changes are recorded and analyzed with the DIC system. In the glass-fiber-reinforced laminated specimen, internal defects were detected from a strain value of 5% until the end of the experiment, while in the case of the carbon-fiber-reinforced laminated specimen, internal defects were detected from 1% onward. It was proved using the DIC system that the defect was detected as a certain level of strain difference developed around the internal defects, according to the material properties.

• The Journal of the Acoustical Society of Korea
• /
• v.39 no.5
• /
• pp.468-475
• /
• 2020

A good acoustic word embedding that can well express the characteristics of word plays an important role in wake-up word detection (WWD). However, the representation ability of acoustic word embedding may be weakened due to various types of environmental noise occurred in the place where WWD works, causing performance degradation. In this paper, we proposed triplet loss based Domain Adversarial Training (tDAT) mitigating environmental factors that can affect acoustic word embedding. Through experiments in noisy environments, we verified that the proposed method effectively improves the conventional DAT approach, and checked its scalability by combining with other method proposed for robust WWD.

• Proceedings of the Acoustical Society of Korea Conference
• /
• spring
• /
• pp.91-94
• /
• 2002

음성인식 시스템의 실용화를 위해서 우선적으로 해결되어야 될 문제중 하나로 잡음환경하에서의 끝점검출을 들 수 있다. 잡음이 존재하지 않는 환경에서는 기존의 에너지 파라미터만으로도 어느정도 신뢰성있는 끝점 구간을 검출할 수 있으나 도심 소음과 같은 실제 잡음환경하에서는 대부분 좋지 않은 결과를 보인다. 본 논문에서는 도심환경의 배경잡음을 제거하는 방법으로 입력되는 음성에 대하여 주변소음에 의해 손상된 음성스펙트럼의 크기 성분만을 제거하는 전처리 기법인 Bark scale에 기반한 스펙트럼 차감법을 사용하고, 인간의 청각특성을 고려하여 음성의 주파수 대역을 3개의 대역으로 분리한 후, 대역별로 세밀한 에너지 문턱치값을 설정하여 음성의 끝점을 탐색하는 방법을 제안한다. 제안한 방법의 유효성을 확인하기 위해 실제 사무실 및 지하철역 등의 잡음환경하에서 녹음된 데이터베이스를 이용하여 끝점검출을 수행한 결과 기존의 에너지와 영교차율을 이용한 방법에 비해 평균 $46\%$의 오차율 감소와 대역에너지만을 사용한 경우에 비해 평균 $17\%$의 오차율 감소를 나타내어 제안한 방법의 유효성을 확인할 수 있었다.