Search

The semiconductor industry, driven by technological advancements, is continuously undergoing process micronization. This micronization has led to an increased complexity in the wafer fabrication process and equipment. Inevitably, this change leads to a rise in defect rates. Furthermore, the traditional manual analysis methods for these defects possess several limitations. Notably, the defect analysis process is time-consuming and heavily relies on the expertise of the workers, making it challenging to achieve consistent analysis results. In modern semiconductor manufacturing sites, a vast amount of sensor data is being generated in real-time from various equipment. These sensor data contain information on the wafer’s condition, the progress of the process, and the operating conditions of the equipment. The purpose of this study is to examine the use of this sensor data in creating a deep learning model that leverages information generated within multi-stage manufacturing processes. The proposed method is designed to automatically classify the defect status of wafers and interpret the causes of these defects. Using the sensor data from two process equipment in semiconductor manufacturing, we achieve better classification performance than other existing methods. Through the interpretation of the classification results, we can identify the sensors and equipment that significantly impact the classification results. The proposed methodology is expected to significantly reduce the effort and reliance on engineers for defect analysis, thus greatly improving production efficiency.

Abstract Bersiporocin, a potent and selective prolyl‐tRNA synthetase inhibitor, is expected to show a synergistic effect with pirfenidone or nintedanib in patients with idiopathic pulmonary fibrosis. To validate the combination therapy of bersiporocin with pirfenidone or nintedanib, a randomized, open‐label, two‐part, one‐sequence, three‐period, three‐treatment study was designed to evaluate the effect of drug–drug interactions (DDI) regarding their pharmacokinetics, safety, and tolerability in healthy participants. In addition, the pharmacokinetic profiles of the newly formulated, enteric‐coated bersiporocin tablet were evaluated after single and multiple administrations. The potential effects of cytochrome P450 2D6 (CYP2D6) genotyping on bersiporocin pharmacokinetics and DDI were also explored. In Part 1, participants were sequentially administered a single dose of pirfenidone 600 mg, a single dose of bersiporocin 150 mg followed by multiple doses, and bersiporocin in combination with pirfenidone. In Part 2, participants were sequentially administered a single dose of nintedanib 150 mg, multiple doses of bersiporocin 150 mg, and bersiporocin in combination with nintedanib. Forty‐six participants completed the study. There was no significant pharmacokinetic DDI between bersiporocin, and pirfenidone or nintedanib. All adverse events (AEs) were mild to moderate and did not include serious AEs, suggesting bersiporocin alone or in combination therapy were well‐tolerated. The newly formulated bersiporocin 150 mg tablet showed a moderate accumulation index. There was no significant difference in the pharmacokinetic profiles after administration of bersiporocin alone or in combination therapy between CYP2D6 phenotypes. In conclusion, there are no significant DDI regarding the pharmacokinetics, safety, and tolerability of bersiporocin administration with pirfenidone or nintedanib.

Background: While efforts have been made to differentiate fall risk in older adults using wearable devices and clinical methodologies, technologies are still infancy. We applied a decision tree (DT) algorithm using inertial measurement unit (IMU) sensor data and clinical measurements to generate high performance classification models of fall risk of older adults. Objects: This study aims to develop a classification model of fall risk using IMU data and clinical measurements in older adults. Methods: Twenty-six older adults were assessed and categorized into high and low fall risk groups. IMU sensor data were obtained while walking from each group, and features were extracted to be used for a DT algorithm with the Gini index (DT1) and the Entropy index (DT2), which generated classification models to differentiate high and low fall risk groups. Model’s performance was compared and presented with accuracy, sensitivity, and specificity. Results: Accuracy, sensitivity and specificity were 77.8%, 80.0%, and 66.7%, respectively, for DT1; and 72.2%, 91.7%, and 33.3%, respectively, for DT2. Conclusion: Our results suggest that the fall risk classification using IMU sensor data obtained during gait has potentials to be developed for practical use. Different machine learning techniques involving larger data set should be warranted for future research and development.

Background: Like many other countries, falls and related injuries in older adults are great concerns in South Korea. In particular, falls are common in medical institutions, often causing the increase of the length of hospitalization. Objects: The purpose of this review was to help understand and address falls in hospitalized individuals in South Korea. Methods: The review was conducted on literature published in Korean from 2010 to 2022, searched in the Korea Citation Index and PubMed. Keywords used for the search were as follows: falls, fall risk, fall risk assessment, hospital, inpatient, intervention, Korea, and prevention. Results: A total of 54 articles were found and reviewed. The most common place of fall accidents was the inpatient room, where there were many cases of falls while walking. Loss of balance was the most common cause of falls, and many falls occurred in patients admitted to the internal medicine. Furthermore, a risk of falling increased with the type of medications taken. In terms of tools to assess patients’ fall risk, the Morse Fall Scale (MFS) was commonly used. Patient-specific fall prevention activities were common to address falls, and they decreased the frequency of falls and the fear of falling. Factors influencing the effectiveness of the fall prevention activities included attitudes toward falls, education, environmental factors, patient safety culture, and self-efficacy in preventing falls. Conclusion: Our results should help understand and address falls and injuries in medical institutions.

Background: Despite fall prevention strategies suggested by researchers, falls are still a major health concern in older adults. Understanding factors that differentiate successful versus unsuccessful balance recovery may help improve the prevention strategies. Objects: The purpose of this review was to identify biomechanical factors that differentiate successful versus unsuccessful balance recovery in the event of a fall. Methods: The literature was searched through Google Scholar and PubMed. The following keywords were used: ‘falls,‘ ‘protective response,‘ ‘protective strategy,’ ‘automated postural response,’ ‘slips,’ ‘trips,’ ‘stepping strategy,‘ ‘muscle activity,’ ‘balance recovery,‘ ‘successful balance recovery,‘ and ‘failed balance recovery.’ Results: A total of 64 articles were found and reviewed. Most of studies included in this review suggested that kinematics during a fall was important to recover balance successfully. To be successful, appropriate movements were required, which governed by several things depending on the direction and characteristics of the fall. Studies also suggested that lower limb muscle activity and joint moments were important for successful balance recovery. Other factors associated with successful balance recovery included fall direction, age, appropriate protective strategy, overall health, comorbidity, gait speed, sex and anticipation of the fall. Conclusion: This review discusses biomechanical factors related to successful versus unsuccessful balance recovery to help understand falls. Our review should help guide future research, or improve prevention strategies in the area of fall and injuries in older adults.

Image-mixing augmentations (e.g., Mixup and CutMix), which typically involve mix ing two images, have become the de-facto training techniques for image classification. Despite their huge success in image classification, the number of images to be mix ed has not been elucidated in the literature: only the naive K-image expansion has been shown to lead to performance degradation. This study derives a new K-image mixing augmentation based on the stick-breaking process under Dirichlet prior distribution. We demonstrate the superiority of our K-image expansion augmentation over conventional two-image mixing augmentation methods through extensive experiments and analyses: 1) more robust and generalized classifiers; 2) a more desirable loss landscape shape; 3) better adversarial robustness. Moreover, we show that our probabilistic model can measure the sample-wise uncertainty and boost the efficiency for network architecture search by achieving a 7-fold reduction in the search time. Code will be available at https://github.com/yjyoo3312/DCutMix-PyTorch.git.

Few-shot learning aims to develop well-trained models by using only a few annotated samples. However, the performance of few-shot learning deteriorates if inappropriate support samples are selected. This happens because of its large dependency on only a few support samples, which limits the stable upgrade of performance even when additional training samples are provided. To overcome this issue, we introduce an active instance selection mechanism where effective support instances are chosen from the unlabeled pool to improve the performance of few-shot classification while maintaining the same annotation cost. Our mechanism comprises four processing steps: feature pre-processing, feature distribution fitting, instance scoring, and instance selection. First, the feature pre-processing normalizes the scattered features to obtain a well-defined distribution stably. Then, the feature distribution fitting finds the distributional relationship between the support images and the unlabeled query samples. Finally, in instance scoring and instance selection, we select the instances based on the instance-wise score estimated by using the fitted distributions. To verify the validity of the proposed mechanism, we employ various algorithms for the feature distribution fitting and the instance scoring. We perform several experiments by integrating the proposed mechanism into a variety of few-shot classification methods. Furthermore, we validate the effectiveness of our mechanism’s components through extensive ablation studies. The experimental results verify that the proposed algorithm presents stable performance improvement with some additional labeling cost for the few-shot classification task. We also present the optimal methods to select effective samples to be labeled for the target few-shot classification algorithms.

Deep learning-based crowd density estimation can greatly improve the accuracy of crowd counting. Though a Bayesian loss method resolves the two problems of the need of a hand-crafted ground truth (GT) density and noisy annotations, counting accurately in high-congested scenes remains a challenging issue. In a crowd scene, people’s appearances change according to the scale of each individual (i.e., the person-scale). Also, the lower the sparsity of a local region (i.e., the crowd-sparsity), the more difficult it is to estimate the crowd density. In this paper, we propose a novel congestion-aware Bayesian loss method that considers the person-scale and crowd-sparsity. We estimate the person-scale based on scene geometry, and we then estimate the crowd-sparsity using the estimated person-scale. The estimated person-scale and crowd-sparsity are utilized in the novel congestion-aware Bayesian loss method to improve the supervising representation of the point annotations. We verified the effectiveness of each proposed component through several ablation experiments, and in the various experiments on public datasets, our proposed method achieved state-of-the-art performance.

A new linear regression form is derived for a flux observer and a position observer is designed. In general, the observability of the permanent-magnet synchronous motor is lost at zero speed. In this work, the proposed regressor vector contains current derivative terms in both directions (dq-axis), and it gives the chance for the model-based flux observer to operate at zero speed. When an excitation signal is injected into d and q axes with the proposed flux observer, it helps to satisfy the persistent excitation condition in the low-speed range. Therefore, the sensorless performance of the model-based is improved greatly, even at zero speed. However, it appears with a disturbance term, which depends on the derivative of the d-axis current. Thus, the disturbance does not vanish when an excitation signal is injected. In this work, the disturbance term is also taken care of in constructing an observer. It results in an observer which allows signal injection. Thus, high frequency signal can be injected in the low speed region and turned off when it is unnecessary as the speed increases. This model-based approach utilizes the signal injection directly without recurring to a separate high frequency model. In other words, it provides a seamless transition without switching to the other algorithm. The validity is demonstrated by simulation and experimental results under various load conditions near zero speed.

To log in to a mobile social network service (SNS) server, users must enter their ID and password to get through the authentication process. At that time, if the user sets up the automatic login option on the app, a sort of security token is created on the server based on the user’s ID and password. This security token is called a credential. Because such credentials are convenient for users, they are utilized by most mobile SNS apps. However, the current state of credential management for the majority of Android SNS apps is very weak. This paper demonstrates the possibility of a credential cloning attack. Such attacks occur when an attacker extracts the credential from the victim’s smart device and inserts it into their own smart device. Then, without knowing the victim’s ID and password, the attacker can access the victim’s account. This type of attack gives access to various pieces of personal information without authorization. Thus, in this paper, we analyze the vulnerabilities of the main Android-based SNS apps to credential cloning attacks, and examine the potential leakage of personal information that may result. We then introduce effective countermeasures to resolve these problems.

Two different kinds of nuclear power plants produce a substantial amount of spent fuel annually in Korea. According to the current projection, it is expected that around 60,000 MtU of spent fuel will be produced from 36 PWR and APR reactors and 4 CANDU reactors by the end of 2089. In 2006, KAERI proposed a conceptual design of a geological disposal system (called KRS, Korean Reference disposal System for spent fuel) for PWR and CANDU spent fuel, as a product of a 4-year research project from 2003 to 2006. The major result of the research was that it was feasible to construct a direct disposal system for 20,000 MtU of PWR spent fuels and 16,000 MtU of CANDU spent fuel in the Korean peninsula. Recently, KAERI and MEST launched a project to develop an advanced fuel cycle based on the pyroprocessing of PWR spent fuel to reduce the amount of HLW and reuse the valuable fissile material in PWR spent fuel. Thus, KAERI has developed a geological disposal system for high-level waste from the pyroprocessing of PWR spent fuel since 2007. However, since no decision was made for the CANDU spent fuel, KAERI improved the disposal density of KRS by introducing several improved concepts for the disposal canister. In this paper, the geological disposal systems developed so far are briefly outlined. The amount and characteristics of spent fuel and HLW, 4 kinds of disposal canisters, the characteristics of a buffer with domestic Ca-bentonite, and the results of a thermal design of deposition holes and disposal tunnels are described. The different disposal systems are compared in terms of their disposal density.

Among the several options to solve PWR spent fuel accumulation problem in Korea, the dry storage method could be the most realistic and applicable solution in the near future. As the basic objectives of dry storage are to prevent a gross rupture of spent fuel during operation and to keep its retrievability until transportation, at the same time the importance of a spent fuel integrity evaluation that can estimate its condition at the final stage of dry storage is very high. According to the national need and technology progress, two representative nations of spent fuel dry storage, the USA and Japan, have established different system temperature criteria, which is the only controllable factor in a dry storage system. However, there are no technical criteria for this evaluation in Korea yet, it is necessary to review the previously well-organized methodologies of advanced countries and to set up our own domestic evaluation direction due to the nation's need for dry storage. To satisfy this necessity, building a domestic spent fuel test database should be the first step. Based on those data, it is highly recommended to compare domestic data range with foreign results, to build our own criteria, and to expand on evaluation work into recently issued integrity problems by using a comprehensive integrity evaluation code.

A sensorless position scheme was developed for wound synchronous machines. The demodulation process is fundamentally the same as the conventional signal-injection method. The scheme is different from techniques for permanent-magnet synchronous machines, in that it injects a carrier signal into the field (rotor) winding. The relationship between the high-frequency current responses and the angle estimation error was derived with cross-coupling inductances. Furthermore, we develop a compensation method for the cross-coupling effect, and present several advantages of the proposed method in comparison with signal injection into the stator winding. This method is very robust against magnetic saturation because it does not depend on the saliency of the rotor. Furthermore, the proposed method does not need to check the polarity at a standstill. Experiments were performed to demonstrate the improvement in the compensation of cross-coupling, and the robustness against magnetic saturation with full-load operation.

The aim of this study is to investigate the factors affecting rehabilitation to society of Emergency Department (ED) based post-suicidal care program recipients. This prospective observational study was conducted at the ED of a general hospital in Seoul between March 2018 and October 2020. Patients who agreed to ED based post-suicidal care were followed up after 2 and 4 months and classified into a rehabilitated group and a non-rehabilitated group respectively. We analyzed the patients' demographic data and suicide-related data. We used the Chi-square test or the Fisher's exact test, the student's t-test and multivariate logistic regression analysis. A total of 86 patients were included in this study. Factors that disturb rehabilitation of suicidal attempters were the presence of a housemate, a history of psychiatric admission, an acute disease or disability on 1st follow-up and the presence of a housemate, educational level and 1st non-rehabilitation on 2nd follow-up on univariate analysis. Multivariate logistic regression analysis showed that the absence of a housemate and 1st rehabilitation were associated with rehabilitation. This study shows that fast rehabilitation is important for rehabilitation of suicidal attempters and the presence of housemate can be a risk factor for non-rehabilitation. [ABSTRACT FROM AUTHOR]