• Journal of the Korea Institute of Information and Communication Engineering
• /
• v.26 no.10
• /
• pp.1490-1499
• /
• 2022

Arrhythmia is a condition in which the heart has an irregular rhythm or abnormal heart rate, early diagnosis and management is very important because it can cause stroke, cardiac arrest, or even death. In this paper, we propose arrhythmia classification using hybrid combination model of CNN-BLSTM. For this purpose, the QRS features are detected from noise removed signal through pre-processing and a single bit segment was extracted. In this case, the GAN oversampling technique is applied to solve the data imbalance problem. It consisted of CNN layers to extract the patterns of the arrhythmia precisely, used them as the input of the BLSTM. The weights were learned through deep learning and the learning model was evaluated by the validation data. To evaluate the performance of the proposed method, classification accuracy, precision, recall, and F1-score were compared by using the MIT-BIH arrhythmia database. The achieved scores indicate 99.30%, 98.70%, 97.50%, 98.06% in terms of the accuracy, precision, recall, F1 score, respectively.

• Smart Structures and Systems
• /
• v.29 no.6
• /
• pp.757-766
• /
• 2022

To train deep learning algorithms, a sufficient number of data are required. However, in most engineering systems, the acquisition of fault data is difficult or sometimes not feasible, while normal data are secured. The dearth of data is one of the major challenges to developing deep learning models, and fault diagnosis in particular cannot be made in the absence of fault data. With this context, this paper proposes an anomaly detection methodology for rotating machines using only normal data with self-labeling. Since only normal data are used for anomaly detection, a self-labeling method is used to generate a new labeled dataset. The overall procedure includes the following three steps: (1) transformation of normal data to self-labeled data based on a pretext task, (2) training the convolutional neural networks (CNN), and (3) anomaly detection using defined anomaly score based on the softmax output of the trained CNN. The softmax value of the abnormal sample shows different behavior from the normal softmax values. To verify the proposed method, four case studies were conducted, on the Case Western Reserve University (CWRU) bearing dataset, IEEE PHM 2012 data challenge dataset, PHMAP 2021 data challenge dataset, and laboratory bearing testbed; and the results were compared to those of existing machine learning and deep learning methods. The results showed that the proposed algorithm could detect faults in the bearing testbed and compressor with over 99.7% accuracy. In particular, it was possible to detect not only bearing faults but also structural faults such as unbalance and belt looseness with very high accuracy. Compared with the existing GAN, the autoencoder-based anomaly detection algorithm, the proposed method showed high anomaly detection performance.

• Journal of the Society of Naval Architects of Korea
• /
• v.60 no.2
• /
• pp.95-109
• /
• 2023

In the 4th industrial revolution, changes in the technological paradigm have had a direct impact on the maintenance system of ships. The 2-stroke low speed engine system integrates with the core equipment required for propulsive power. The Condition Based Management (CBM) is defined as a technology that predictive maintenance methods in existing calender-based or running time based maintenance systems by monitoring the condition of machinery and diagnosis/prognosis failures. In this study, we have established a framework for CBM technology development on our own, and are engaged in engineering-based failure analysis, data development and management, data feature analysis and pre-processing, and verified the reliability of failure mode DB using LSTM algorithms. We developed various simulated failure mode scenarios for 2-stroke low speed engine and researched to produce data on onshore basis test_beds. The analysis and pre-processing of normal and abnormal status data acquired through failure mode simulation experiment used various Exploratory Data Analysis (EDA) techniques to feature extract not only data on the performance and efficiency of 2-stroke low speed engine but also key feature data using multivariate statistical analysis. In addition, by developing an LSTM classification algorithm, we tried to verify the reliability of various failure mode data with time-series characteristics.

• Physical Therapy Korea
• /
• v.30 no.2
• /
• pp.136-143
• /
• 2023

Background: The International Classification of Functioning, Disability and Health (ICF) model, created by the World Health Organization, provides a theoretical framework that can be applied in the diagnosis and treatment of various disorders. Objects: Our research purposed to ascertain the relationship between structure/function, activity, and participation domain variables of the ICF and pain, pain-associated disability, activities of daily living (ADL), and quality of life in patients with chronic low back pain (LBP). Methods: Two-hundred patients with chronic LBP (mean age: 35.5 ± 8.8 years, females, n = 40) were recruited from hospital and community settings. We evaluated the body structure/function domain variable using the Numeric Pain Rating Scale (NPRS) and Roland-Morris disability (RMD) questionnaire. To evaluate the activity domain variable, we used the Oswestry Disability Index (ODI) and Quebec Back Pain Disability Scale (QBDS). For clinical outcome measures, we used Short-form 12 (SF-12). Pearson's correlation coefficient was used to ascertain the relationships among the variables (p < 0.05). All the participants with LBP received 30 minutes of conventional physical therapy 3 days/week for 4 weeks. Results: There were significant correlations between the body structure/function domain (NPRS and RMD questionnaire), activity domain (ODI and QBDS), and participation domain variables (SF-12), rending from pre-intervention (r = -0.723 to 0.783) and postintervention (r = -0.742 to 0.757, p < 0.05). Conclusion: The identification of a significant difference between these domain variables point to important relationships between pain, disability, performance of ADL, and quality in participants with LBP.

• The Journal of the Korea institute of electronic communication sciences
• /
• v.18 no.5
• /
• pp.999-1008
• /
• 2023

Early prediction of chronic diseases such as diabetes is an important issue, and improving the accuracy of diabetes prediction is especially important. Various machine learning and deep learning-based methodologies are being introduced for diabetes prediction, but these technologies require large amounts of data for better performance than other methodologies, and the learning cost is high due to complex data models. In this study, we aim to verify the claim that DNN using the pima dataset and k-fold cross-validation reduces the efficiency of diabetes diagnosis models. Machine learning classification methods such as decision trees, SVM, random forests, logistic regression, KNN, and various ensemble techniques were used to determine which algorithm produces the best prediction results. After training and testing all classification models, the proposed system provided the best results on XGBoost classifier with ADASYN method, with accuracy of 81%, F1 coefficient of 0.81, and AUC of 0.84. Additionally, a domain adaptation method was implemented to demonstrate the versatility of the proposed system. An explainable AI approach using the LIME and SHAP frameworks was implemented to understand how the model predicts the final outcome.

• Journal of the Korean Geotechnical Society
• /
• v.40 no.3
• /
• pp.33-39
• /
• 2024

Inspecting cracks to determine a structure's condition is crucial for accurate safety diagnosis. However, visual crack inspection methods can be subjective and are dependent on field conditions, thereby resulting in low reliability. To address this issue, this study automates the detection of concrete cracks in image data using ResNet, FPN, and the Mask R-CNN components as the backbone, neck, and head of a convolutional neural network. The performance of the proposed model is analyzed using the intersection over the union (IoU). The experimental dataset contained 1,203 images divided into training (70%), validation (20%), and testing (10%) sets. The model achieved an IoU value of 95.83% for testing, and there were no cases where the crack was not detected. These findings demonstrate that the proposed model realized highly accurate detection of concrete cracks in image data.

• The Transactions of the Korean Institute of Power Electronics
• /
• v.11 no.5
• /
• pp.480-488
• /
• 2006

Up to recently the harmonic generation has deteriorated the quality of electricity and affected the performance on the electrical installation including OA, FA, IT devices and so on. Some studies of harmonic affects in diagnosis and the cause of accident has not done by the experimental data of harmonic source but merely by presumption according to qualitative analysis. So, in order to research the harmonic affect on the electrical installation according to quantitative analysis and gather reliable data over and over again, it is necessary to develop an AC power source which is capable of generating some harmonics. In this paper, we described about realization of AC power source which can produce and compose harmonics for the analysis of accident due to harmonics.

• Journal of Sensor Science and Technology
• /
• v.27 no.4
• /
• pp.254-258
• /
• 2018

A Wi-Fi signal network (WSN) system is introduced in this paper. This system consists of several data-transmitting sensor modules and a data-receiving server. Each sensor module and the server contain a unique intranet IP address. A piezoelectric accelerometer with a bandwidth of 12 kHz, a 24-bit analog-digital converter with a sampling rate of 15.625 kS/s, a 32-bit microprocessor unit, and a 1-Mbps Wi-Fi module are used in the data-transmitting sensor module. A 300-Mbps router and a PC are used in the server. The system is verified using an accelerometer calibrator. The voltage output from the sensor is converted into 24-bit digital data and transmitted via the Wi-Fi module. These data are received by a Wi-Fi router connected to a PC. The input frequencies of the accelerometer calibrator (320 Hz, 640 Hz, and 1280 Hz) are used in the data transfer verification. The received data are compared to the data retrieved directly from the analog-to-digital converter used in the sensor module. The comparison shows that the developed system represents the original data considerably well. Theoretically, the system can acquire vibration signals from 600 sensor modules at an accelerometer bandwidth of 15.625 kHz. However, delay exists owing to software processes, multiplexing between sensor modules, and the use of non-real time operating system. Hence, it is recommended that this system may be used to acquire vibration signals with up to 10 kHz, which is approximately 70% of the theoretical maximum speed of the system. The system can be upgraded using parts with higher performance

• Knowledge Management Research
• /
• v.16 no.1
• /
• pp.209-224
• /
• 2015

Previous studies provided that limiting the number of services provided in hospital had influences in decreasing cost in delivering medical services. Hospitals could have positive effects on their profit by concentrating small number of services which they have comparative advantages. This study purposed to analyze the relationship between the concentration status of hospitals and medical charge for inpatients. National Inpatient sample data provided by the Health Insurance Review and Assessment Service (HIRA) for three years, 2009 to 2011 was used to compute the three concentration indices (Information Theory Index (ITI), Internal Herfindahl Index (IHI), and number of distinct Diagnosis-Related Groups (DRGs) treated) and total medical charge per inpatient case in each year. It was also used to select the control variables such as bed size, number of doctors per 100 beds, and locations. The ordinary least square regression models were developed and tested for hospital and general hospitals separately. The results showed that the total medical charge per inpatient case was significantly differed depending on the concentration indices, and there were positive relationships in ITI and IHI. The number of distinct DRGs had different directions in regression coefficients depending on the locations and hospital types. Hospitals had larger absolute standardized regression coefficients compare to those of general hospitals. However, their effects could be varied by the hospital types, number of doctors, and locations. It seems that hospitals have more influences on medical charges by concentrating their services than general hospitals. Study results provide knowledges to hospital administrators that concentration strategy can positive influences on the performance of small size hospitals.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.15 no.2
• /
• pp.940-951
• /
• 2014

This study was to investigate of nursing activities performed by nursing staffs in long-term care hospitals. A survey of was conducted with 131 nurses who were working in 15 geriatric hospitals using a structural questionnaire. Data were collected from March 7 to June 30, 2013 and analyzed with SPSS 17.0, using descriptive statistics and t-test, ANOVA. This survey confirmed that the 88 activities except to 'making nursing diagnosis', 'setting nursing goals', and 'writting patient evaluation checklist' had been practiced in a wide range of nurses with nurse assistants. The degree of performance by nurse assistant was significantly different among nurses group according to the total number of patient and number of nurse(p<.001). Therefore, it is necessary to establish the legal and institutional regulation and development of algorithm for delegation, classify the impossible nursing task and work that can be delegated long-term care hospital.