• Smart Media Journal
• /
• v.12 no.9
• /
• pp.45-59
• /
• 2023

Weakly supervised object localization (WSOL) is a task of localizing an object in an image using only image-level labels. Previous studies have followed the conventional class activation mapping (CAM) pipeline. However, we reveal the current CAM approach suffers from problems which cause original CAM could not capture the complete defects features. This work utilizes a convolutional neural network (CNN) pretrained on image-level labels to generate class activation maps in a multi-scale manner to highlight discriminative regions. Additionally, a vision transformer (ViT) pretrained was treated to produce multi-head attention maps as an auxiliary detector. By integrating the CNN-based CAMs and attention maps, our approach localizes defective regions without requiring bounding box or pixel-level supervision during training. We evaluate our approach on a dataset of apple images with only image-level labels of defect categories. Experiments demonstrate our proposed method aligns with several Object Detection models performance, hold a promise for improving localization.

• Proceedings of the Safety Management and Science Conference
• /
• 2012.04a
• /
• pp.233-243
• /
• 2012

HBD(Hot Box Detector) is a device to monitor temperature rises to inappropriate lubricant use or mechanical defects. If a train operates without recognizing such an effect, it might result in bearing overheating due to defects and cause a dangerous situation that it could derail a train owing to the damage of axles. Now for the Gyeongbu HSL at 300km/h, the laws related to monitoring overheated axle bearings are notified in the Railway Safety Law and the Railway Construction Law. But in case of the conventional speed-up lines that a train operates at 180 to 230 km/h, the revised bill of relevant standards is ongoing. Therefore in this paper we present references and reviews investigated in order to use the optimal HBD in the conventional speed-up lines.

• Proceedings of the KSR Conference
• /
• 2011.10a
• /
• pp.1439-1443
• /
• 2011

Bimodal Tram is the new conceptual and environmental-friendly public transportation which adopted series hybrid system. The generator driven by CNG engine supplies the electric power to Battery and traction motor. The generator installed on the vehicle will experience the mechanical vibration and electrical transient variation. Those may cause some defects on the generator which will be the hazardous effects to the vehicle. This paper has investigated the possibility to find out some diagnostic features for the defects of generator through the voltage and current generated from it. Those were analyzed in both time and frequency regions. For the next, more works will be needed to complete the purpose of this paper.

• Proceedings of the KSR Conference
• /
• 2005.11a
• /
• pp.318-325
• /
• 2005

In present, KTX's wheel is worn inevitably according to the interaction with rail in service. It was analyzed by wearing type, damage type and wheel reprofiling cycle on the running surface. As a result, damage on the running surface is main cause to reprofile the wheel. Wearing type of wheel is normal and it hasn't reprofiled according to normal wheel wear. As a follow-up result of the whee] defects on the running surface, if the defects size is well managed by periodical inspection, it would be efficient to increase the wheel life.

• Journal of Korea Foundry Society
• /
• v.29 no.3
• /
• pp.138-143
• /
• 2009

Vacuum suction is applied to the mold during pouring in the inclined gravity die casting to remove defects such as misrun and gas porosity in the brake master cylinder. Casting defects are observed after solidification and their cause is analyzed by using the calculated results with commercial solidification and flow analysis code(ZCAST). The optimum vacuum suction is -2 cmHg, and when the start time of vacuum suction is 3 seconds after pouring, better filled result is obtained by holding it for 15 seconds. Reproducibility test under the optimum conditions attained from the above pouring tests is performed, and it is confirmed that these pouring conditions can be applied to the mass production immediately.

• Transactions of the Korean Society for Noise and Vibration Engineering
• /
• v.14 no.12
• /
• pp.1287-1294
• /
• 2004

Hub bearings not only sustain the body of a cat, but permit wheels to rotate freely. Excessive radial or axial load and many other reasons can cause defects to be created and grown in each component. Therefore, nitration and noise from unwanted defects in outer-race, inner-race or ball elements of a Hub bearing are what we want to detect as early as possible. How early we can detect the faults has to do with how the detection algorithm finds the fault information from measured signal. Fortunately, the bearing signal has Periodic impulse train. This information allows us to find the faults regardless how much noise contaminates the signal. This paper shows the basic signal processing idea and experimental results that demonstrate how good the method is.

• Proceedings of the Korean Society for Noise and Vibration Engineering Conference
• /
• 2004.05a
• /
• pp.593-596
• /
• 2004

Hub bearings not only sustain the body of a car, but permit wheels to rotate freely. Excessive radial or axial load and many other reasons can cause defects to be created and grown in each component. Therefore, vibration and noise from unwanted defects in outer-race, inner-race or ball elements of a Hub bearing are what we want to detect as early as possible. How early we can detect the faults has to do with how the detection algorithm finds the fault information from measured signal. Fortunately, the bearing signal has periodic impulse train. This information allows us to find the faults regardless how much noise contaminates the signal. This paper shows the basic signal processing idea and experimental results that demonstrate how good the method is.

• The KIPS Transactions:PartD
• /
• v.12D no.7 s.103
• /
• pp.965-978
• /
• 2005

There are many defects that cause the schedule and qualify problems during software development. This paper designs the opportunity tree framer work that removes and manages the schedule and quality problems as well. For the similar projects, we can estimate defects and prepare to solve them by using domain expert knowledge and the opportunity tree framework which can greatly improve the software process. This research provides solution of schedule defect problem and detection of defect and its causes that happen on software development.

• Journal of the Korea Institute of Building Construction
• /
• v.21 no.6
• /
• pp.653-664
• /
• 2021

As the construction standards for energy-saving eco-friendly housing have recently been strengthened, the proportion of window work has increased with the demand for high-efficiency housing. Windows have high frequency of use, and there is the potential for many defects to occur depending on the characteristics of construction. According to a government agency's survey of defects in public rental apartment housing, defects in the windows work accounted for the highest portion of complaints received. Accordingly, related previous studies were considered, and it was found that the existing studies in Korea lacked research that reflected the construction characteristics of window work and the importance of maintenance. In addition, existing overseas studies considered both the constructor and the resident's position, considering the cost aspect together, and showed a trend of structuring the relationship between defects and causes. Therefore, this study will analyze the causes of defects that can occur in the construction phase of the windows work, reflect the construction characteristics, and derive major factors that consider the importance of maintenance based on the possibility of recurrence after repairing defects. Ultimately, this research will contribute to preventing defects in the construction phase and reducing maintenance costs by presenting a highly effective defect management plan through selecting the major factors for each defect type that can be intuitively judged by analyzing the causal relationship between defect types and causes.

• Korean Journal of Construction Engineering and Management
• /
• v.25 no.1
• /
• pp.42-49
• /
• 2024

In recent years, the construction industry has diligently focused on improving the quality and safety of buildings through smart technologies. However, there is a growing trend of leakage defects, especially in educational facilities, due to aging. The objective of this study is to analyze the causes of these defects in educational environments using the Fault Tree Analysis (FTA) technique and propose preventive measures based on the findings. The FTA technique is explained through a review of domestic literature, and data from the Educational Support Center from 2019 to 2021 are examined to identify major defects. The construction of the Fault Tree (FT) for leakage defects resulted in the identification of 12 basic events. Subsequently, a comprehensive understanding of the causes of leakage is achieved through FTA analysis, leading to the identification of the primary causes of defects. Leakage defects accounted for 46.8% of all reported issues in educational facilities, with roof (ceiling) leaks being the most common problem. FTA analysis revealed that poor substrate treatment was the main cause of roof (ceiling) leaks, which could be attributed to cracks in the waterproof layer, joint cracks, and microvoids in the waterproof layer. The primary achievement of this research is to provide essential data for preventing leakage defects in educational facilities and developing preventive measures through the FTA technique. These results are expected to significantly enhance the management of educational facilities and the prevention of leakage issues.