• 산업경영시스템학회지
• /
• 제46권4호
• /
• pp.312-320
• /
• 2023

Recently, there has been an increasing attempt to replace defect detection inspections in the manufacturing industry using deep learning techniques. However, obtaining substantial high-quality labeled data to enhance the performance of deep learning models entails economic and temporal constraints. As a solution for this problem, semi-supervised learning, using a limited amount of labeled data, has been gaining traction. This study assesses the effectiveness of semi-supervised learning in the defect detection process of manufacturing using the MixMatch algorithm. The MixMatch algorithm incorporates three dominant paradigms in the semi-supervised field: Consistency regularization, Entropy minimization, and Generic regularization. The performance of semi-supervised learning based on the MixMatch algorithm was compared with that of supervised learning using defect image data from the metal casting process. For the experiments, the ratio of labeled data was adjusted to 5%, 10%, 25%, and 50% of the total data. At a labeled data ratio of 5%, semi-supervised learning achieved a classification accuracy of 90.19%, outperforming supervised learning by approximately 22%p. At a 10% ratio, it surpassed supervised learning by around 8%p, achieving a 92.89% accuracy. These results demonstrate that semi-supervised learning can achieve significant outcomes even with a very limited amount of labeled data, suggesting its invaluable application in real-world research and industrial settings where labeled data is limited.

• 구강회복응용과학지
• /
• 제32권1호
• /
• pp.60-69
• /
• 2016

목적: 본 연구에서는 실제 환자의 석고모형의 콘빔CT (Cone Beam Computed Tomography)이미지로 디지털 모델을 제작하고 이 디지털 모델을 동일한 석고모형을 3차원 광학 스캐너로 스캔 하여 얻은 디지털 모델과 비교하였다. 연구 재료 및 방법: 총 11쌍의 석고모형에 대하여 실험을 진행 하였다. 콘빔CT를 이용하여 CT 영상을 촬영하여 디지털 모델을 제작 하였고 3차원 광학 스캐너를 사용해 대조군이 되는 디지털 모델을 제작하였다. 이를 이용해 각 석고모형에 대하여 콘빔CT와 3차원 광학 스캐너를 이용하여 만든 디지털 모델을 한 쌍으로 묶어 상, 하악 11개의 비교 쌍을 구성하고 각 쌍에 대하여 차이점을 분석하였다. 결과: 대조군과 비교 시 콘빔CT 영상으로부터 구성된 디지털 모델이 대조군 보다 과다 추정된 부분인 양의 오차의 평균은 0.059 - 0.117 mm, 과소 추정된 부분인 음의 오차의 평균은 0.066 - 0.146 mm의 범위 내에 존재했다. 또한 유의수준 0.05에서 양의 오차의 평균은 $70-100{\mu}m$, 음의 오차의 평균은 $100-120{\mu}m$ 내에 존재 함을 확인하였다. 결론: 석고모형의 콘빔CT 영상으로부터 구성된 디지털 모델은 최종 수복물 제작에는 부적합하나 임시수복물 제작 및 교정 진단 과정에 활용될 수 있는 가능성이 있다.

• 대한치과보철학회지
• /
• 제40권1호
• /
• pp.30-41
• /
• 2002

This study was to evaluate the fabrication method and marginal adaptation of the conical inner crown fabricated with CAD/CAM. The informations on abutment teeth were transferred to a computer with a micro contact digitizer, which had a $50{\mu}m$ accuracy on the master die. A conical inner crown was designed on a computer and a real crown was machined based on this design using CAM. The marginal fit of a computer-machined conical inner crown was assessed using electron microscopy Measurement of the marginal gap between the conical inner crown and the abutment was performed on four different locations (mesial, distal, buccal, and lingual surfaces) of the finish line. The evaluation was based on 10 test specimens. The results were as follow. 1 The mean marginal gap between the conical inner crown and abutment tooth was $83.2{\pm}43{\mu}m$, 28.9% of the specimen showed marginal gap over $100{\mu}m$. 2. The fabrication method using CATRS and CAM provided clinically acceptable marginal fitness compared to conventional casting method (P<0.05).

• Steel and Composite Structures
• /
• 제1권3호
• /
• pp.295-312
• /
• 2001

The purpose of this paper is to propose a new type of steel reinforced concrete (SRC) beam-column joints and to examine the structural performance of the proposed joints, which simplify the construction procedure of steel fabrication, welding works, concrete casting and joint strengthening. In the proposed beam-column joints, the steel element of columns forms continuously built-in crossing of H-sections (${\Box}$), with adjacent flanges of column being connected by horizontal stiffeners in a joint at the level of the beam flanges. In addition, simplified lateral reinforcement (${\Box}$) is adopted in a joint to confine the longitudinal reinforcing bars in columns. Experimental and analytical studies have been carried out to estimate the structural performance of the proposed joints. Twelve cruciform specimens and seven SRC beam-column subassemblage specimens were prepared and tested. The following can be concluded from this study: (1) SRC subassemblages with the proposed beam-column joints show adequate seismic performances which are superior to the demand of the current code; (2) The yield and ultimate strength capacities of the beam-to-column connections can be estimated by analysis based on the yield line theory; (3) The skeleton curves and the ultimate shear capacities of the beam-column joint panel are predicted with a fair degree of accuracy by considering a simple stress transfer mechanism.

• 한국소성가공학회:학술대회논문집
• /
• 한국소성가공학회 2006년도 제5회 박판성형 SYMPOSIUM
• /
• pp.99-102
• /
• 2006

In the present study, the blow forming characteristics of AZ31 sheet was investigated to test the feasibility of the practical application of wrought Mg alloys. Mg alloys have drawn a huge attention in the field of transportation and consumer electronics industries since it is the lightest alloy which could be industrially applicable. Most Mg alloy components have been fabricated by casting method. However, there have been a lot of research activities on the wrought alloys and their plastic forming process recently. Shallow cups for the small electronics cases have been stamped with warm die system. However, some technical issues will challenge Mg forming when large parts are considered with warm die system over $200^{\circ}C$. Most of all, thermal expansion of die system will deteriorate a die accuracy. On the other hand, blow forming does not have a problem with inaccuracy with die system. In this study, tensile tests were followed by blow forming at various temperature and pressure. AZ31 sheet showed a superplastic deformation behavior with extensive grain boundary sliding at the temperature above $300^{\circ}C$. However, the deformation behavior was likely to differ depending on stress condition.

• 한국정밀공학회지
• /
• 제19권9호
• /
• pp.166-171
• /
• 2002

The objective of this paper is to investigate the effect of clearance and the configuration of die system on burr formation by FEM analysis and experimental tests. Compared with casting, forging and machining, shearing has been known, especially in heavy or mass-production industries, as a very economical and fast way to obtain the desired shape Recently, the shearing process becomes widely used in the small and light electronic component manufacturing industries. When shearing a part of sheet metal, the burr formed on the cutting edge is usually unavoidable. The burr would not only degrade the precision of products but also causes additional cost for the deburring process. In this paper, the influence of shearing parameters such as clearance and configurations of the lower pad (ejector) on burr formation is investigated by using the experimental and numerical approach. From the experimental results, it has been shown that the more narrow clearance gives the smaller burr height and the higher shearing forces. The removal of lower holder also makes the sheared surface integrity and the dimensional accuracy become worse. The FEM results (using DEFORM-2D) show good agreement with the experimental results.

• International Journal of Precision Engineering and Manufacturing
• /
• 제8권3호
• /
• pp.14-19
• /
• 2007

Rapid prototyping (RP) technology can fabricate any 3D physical model regardless of geometric complexity using the layered manufacturing (LM) process. Stereolithography (SL) is the best-known example of RP technology. In general, the surface quality of a raw SL-generated part is unsatisfactory for industrial purposes due to the step artefact created by the LM process. Despite of the increased number of applications for SL parts, this side effect limits their uses. In order to improve their surface quality, additional post-machining finishing, such as traditional grinding, is required, but post-machining is time consuming and can reduce the geometric accuracy of a part. Therefore, this study proposes a post-machining technology combining coating and grinding processes to improve the surface quality of SL parts. Paraffin wax and pulp are used as the coating and grinding materials. By grinding the coating wax only up to the boundary of the part, the surface smoothness can be improved without damaging the surface. Finally, moulding and casting experiments were performed to confirm the suitability of the SL parts finished using the proposed process with rapid tooling (RT) techniques.

• 대한치과보철학회지
• /
• 제40권6호
• /
• pp.572-576
• /
• 2002

This investigation was designed to estimate the biaxial flexure strength and fracture toughness of lithium disilicate glass-ceramics of IPS Empress 2 system pressed with as-received ingots and their sprue buttons. Two groups of the lithium disilicate glass-ceramics were prepared as follows: group 1 is ingot-pressed group; group 2 is sprue button-pressed group. A ball-on-three-ball test was used to determine biaxial flexure strength (BFS) of disks in wet environment. Scanning electron microscopy(SEM) analysis was conducted to observe the microstructure of the ceramics. Unpaired t-test showed that there were no differences in the mean biaxial Hem strength (BFS) and KIC values between group 1 and 2 (p > 0.05). Two groups showed similar values in the KIC and the strength at 5% failure probability. The SEM micrographs of the IPS Empress 2 glass-ceramic showed a closely packed, multi-directionally interlocking pattern of numerous lithium disilicate crystals protruding from the glass matrix. The lithium orthophosphate crystals could not be observed on the fracture surface etched. There was no a marked difference of the microstructure between group 1 and 2. Although there were no tests including color stability, casting accuracy, etc., the results of this study implied that we could reuse the sprue button of the pressed lithium disilicate glass-ceramic of IPS Empress 2 system.

• Structural Engineering and Mechanics
• /
• 제83권4호
• /
• pp.475-493
• /
• 2022

The results of research focusing on the experimental and numerical performance of ferrocement RC beams with openings reinforced with welded steel mesh, expanded steel mesh, fiber glass mesh, and polyethylene mesh independently are presented in this article. Casting and testing of fourteen reinforced concrete beams with dimensions of 200×100×2000 mm under concentric compression loadings were part of the research program. The type of reinforcing materials, the volume fraction of reinforcement, the number of mesh layers, and the number of stirrups are the major parameters that change. The main goal is to understand the impact of using new appealing materials in reinforcing RC beams with openings. Using ANSYS-16.0 Software, nonlinear finite element analysis (NLFEA) was used to demonstrate the behavior of composite RC beams with openings. A parametric study is also conducted to discuss the variables that can have the greatest impact on the mechanical behavior of the proposed model, such as the number of openings. The obtained experimental and numerical results demonstrated the FE simulations' acceptable accuracy in estimating experimental values. Furthermore, demonstrating that the strength gained of specimens reinforced with fiber glass meshes was reduced by approximately 38% when compared to specimens reinforced with expanded or welded steel meshes is significant. In addition, when compared to welded steel meshes, using expanded steel meshes in reinforcing RC beams with openings results in a 16 percent increase in strength. In general, when ferrocement beams with openings are tested under concentric loadings, they show higher-level ultimate loads and energy-absorbing capacity than traditional RC beams.

• 한국산업보건학회지
• /
• 제16권4호
• /
• pp.346-355
• /
• 2006

The purpose of this study was conducted to ascertain the difference between impinger and passive sampling methods in the process of sampling and analyzing on airborne formaldehyde. Formaldehyde generating workplaces included in this study comprised four types of manufacturing industry such as two workplaces of products of wood, cork, straw and plaiting materials manufacturing industries, one casting metal manufacturing industry, and one parts and accessories for motor vehicles and it′s engines manufacturing industry. Workplaces contained in this study were located in some manufacturing area of Busan industrial complex and this study was carried out during a period from January 2003 to December 2004. Analytical accuracy, precision and detection limit of two methods was compared. Exposure level of its airborne concentration was evaluated in formaldehyde generating workplaces those were classified by types of industry, working process, and time. The results were as follows ; 1. A rate of recovery was 107.1% in impinger method and 101.8% in passive method, and precision was 7.79% in impinger method and 4.40% in passive method. There was no statistical significance in analytical accuracy and precision between two methods. A limitation of detection was 0.011 ppm in passive method which was lower than that of impinger method (0.020 ppm) by 1.8 times. 2. Airborne formaldehyde concentration of impinger method was different from passive method. It′s concentration by passive method was higher by 5.1 times than that by impinger method in the parts and accessories for motor vehicles and it′s engines manufacturing industry (P<0.05). Only in molding process among several types of processes, formaldehyde concentration in passive method was higher by 5.1 times than that in impinger method (P<0.05). Furthermore, formaldehyde concentration in passive method was higher by 1.7 times than that in impinger method (P<0.05) in the first half of year 2003. 3. The geometric mean of formaldehyde concentration in impinger method was lower than that in passive method, but there was no statistical significance of formaldehyde concentration by the difference of sampling method. In conclusion, it is difficult to conclude which is better between the two sampling methods because of no statistical significance for the difference of concentration. Because of lacks of certified passive sampling and analytical method, at present situation, studies on verification of accuracy and precision, obstructive reaction against validity on its exposure assessment, and research to develop domestically manufactured passive sampler in terms of cost-effectiveness should be continuously carried out.