• 대한임베디드공학회논문지
• /
• 제19권3호
• /
• pp.151-158
• /
• 2024

Although on-device artificial intelligence (AI) has gained attention to diagnosing machine faults in real time, most previous studies did not consider the model retraining and redeployment processes that must be performed in real-world industrial environments. Our study addresses this challenge by proposing an on-device AI-based real-time machine fault diagnosis system that utilizes continual learning. Our proposed system includes a lightweight convolutional neural network (CNN) model, a continual learning algorithm, and a real-time monitoring service. First, we developed a lightweight 1D CNN model to reduce the cost of model deployment and enable real-time inference on the target edge device with limited computing resources. We then compared the performance of five continual learning algorithms with three public bearing fault datasets and selected the most effective algorithm for our system. Finally, we implemented a real-time monitoring service using an open-source data visualization framework. In the performance comparison results between continual learning algorithms, we found that the replay-based algorithms outperformed the regularization-based algorithms, and the experience replay (ER) algorithm had the best diagnostic accuracy. We further tuned the number and length of data samples used for a memory buffer of the ER algorithm to maximize its performance. We confirmed that the performance of the ER algorithm becomes higher when a longer data length is used. Consequently, the proposed system showed an accuracy of 98.7%, while only 16.5% of the previous data was stored in memory buffer. Our lightweight CNN model was also able to diagnose a fault type of one data sample within 3.76 ms on the Raspberry Pi 4B device.

• 한국정밀공학회지
• /
• 제13권6호
• /
• pp.168-176
• /
• 1996

A diagnosis system which can monitor tool breakage and chipping in real time was developed using a DSP(Digital Signal Processor) board in face milling operation. AR modelling and band energy method were used to extract the feature of tool states from cutting force signals. Artificial neural network embedded on DSP board discriminates different patterns from features got after signal processing. The features extracted from AR modelling are more accurate for the malfunction of a process than those from band energy method, even though the computing speed of the former is slow. From the processed features, we can construct the real time diagnosis system which monitors malfunction by using a DSP board having a parallel processing capability.

• 한국임상수의학회지
• /
• 제37권1호
• /
• pp.23-27
• /
• 2020

Between May and November 2018, babesiosis was examined in 162 bloods samples obtained to an animal hospital in Jeju island for anemia and medical examination. Sixty-two of 162 (38.3%) were positive by PCR. The ultra fast real-time PCR test with blood directly analyzed without DNA extraction showed the same results. Accurate diagnosis, treatment and prognosis of babesiosis should be combined with clinical symptoms, blood tests, the babesia antibody test, and the PCR antigen test. Ultra fast real-time PCR, with these tests, is expected to be a point-of-care testing (POCT) for easy, fast and accurate diagnosis of babesiosis in the veterinary clinic.

• Tuberculosis and Respiratory Diseases
• /
• 제78권1호
• /
• pp.1-7
• /
• 2015

Background: The incidence of tuberculosis (TB) in Korea is relatively high compared to the other Organisation for Economic Co-operation and Development (OECD) countries, with a prevalence of 71 per 100,000 in 2012, although the incidence is declining. Real-time polymerase chain reaction (PCR) has been introduced for the rapid diagnosis of TB. Recently, its advantage lies in higher sensitivity and specificity for the diagnosis of TB. This study evaluated the clinical accuracy of real-time PCR using respiratory specimens in a clinical setting. Methods: Real-time PCR assays using sputum specimens and/or bronchoscopic aspirates from 2,877 subjects were reviewed retrospectively; 2,859 subjects were enrolled. The diagnosis of TB was determined by positive microbiology, pathological findings of TB in the lung and pleura, or clinical suspicion of active TB following anti-TB medication for more than 6 months with a favorable response. Results: Sensitivity, specificity, and accuracy were 44%, 99%, and 86% from sputum, and 65%, 97%, and 87% from bronchoscopic aspirates, respectively. For overall respiratory specimens, sensitivity was 59%, specificity was 98%, and accuracy increased to 89%. Conclusion: Positivity in real-time PCR using any respiratory specimens suggests the possibility of active TB in clinically suspected cases, guiding to start anti-TB medication. Real-time PCR from selective bronchoscopic aspirates enhances the diagnostic yield much more when added to sputum examination.

• 한국항공우주학회지
• /
• 제49권4호
• /
• pp.311-320
• /
• 2021

무인기용 터보제트엔진의 운전 중 발생하는 고장을 실시간으로 진단하기 위한 방안 및 성능 열화와 관련된 건정성 추정에 관해 연구하였다. 이를 위해서, 동적 열역학 가스경로해석을 통한 비선형 동특성 방정식으로부터 실시간 선형모델을 도출하였고, 연출된 운전상황과 고장 발생을 실시간으로 진단하기 위해 칼만필터와 가설 검증에 기초한 확률적 판단 기법을 적용하였다. 이 결과, 분명한 고장 검출과 분리 성능을 보임으로써 그 효용성을 확인하였다. 측정변수를 통한 건전성 추정과 관련하여, 실제 엔진 구성품의 성능 열화 추이를 모사하였고, 적응형 칼만필터를 적용하여 추정 기법의 타당성을 입증함으로써, 상태 기반 고장 진단 및 정비 기법에 효과적으로 사용될 수 있음을 보였다.

• 대한전기학회:학술대회논문집
• /
• 대한전기학회 1988년도 전기.전자공학 학술대회 논문집
• /
• pp.731-733
• /
• 1988

This paper describes a design of hardware system for real time automatic diagnosis of ECG arrhythmia based on pipeline processor consisting of the three microcomputer. ECG data is acquisited by 12 bit A/D converter with hardware QRS triggered detector. Four diagnostic parameters - heart rate, morphology, axis, and ST segment - are used for the classification and the diagnosis of arrhythmia. The functions of the main CPU were distributed and processed with three microcomputers. There-fore the effective data process and the real time process using microcomputer can be obtained. The interconnection structure consisting of two common memory units is designed to decrease the delay time caused by data transfer between processors and by which the delay time can be taken 1 % of one clock period.

• 한국정밀공학회:학술대회논문집
• /
• 한국정밀공학회 1994년도 추계학술대회 논문집
• /
• pp.670-675
• /
• 1994

Yhis paper describes a fault diagnosis simulation of the Real-Time Multiple Fault Dignosis System (RTMFDS) for forcasting faults in a system and deciding current machine state from signal information. Comparing with other diagnosis system for single fault,the system developed deals with multiple fault diagnosis,comprising two main parts. One is a remotesignal generating and transimission terminal and the other is a host system for fault diagnosis. Signal generator generate the random fault signal and the image information, and send this information to host. Host consists of various modules and agents such as Signal Processing Module(SPM) for sinal preprocessing, Performence Monotoring Module(PMM) for subsystem performance monitoring, Trigger Module(TM) for multi-triggering subsystem fault diagnosis, Subsystem Fault Diagnosis Agent(SFDA) for receiving trigger signal, formulating subsystem fault D\ulcornerB and initiating diagnosis, Fault Diagnosis Module(FDM) for simulating component fault with Hierarchical Artificial Neural Network (HANN), numerical models and Hofield network,Result Agent(RA) for receiving simulation result and sending to Treatment solver and Graphic Agent(GA). Each agent represents a separate process in UNIX operating system, information exchange and cooperation between agents was doen by IPC(Inter Process Communication : message queue, semaphore, signal, pipe). Numerical models are used to deseribe structure, function and behavior of total system, subsystems and their components. Hierarchical data structure for diagnosing the fault system is implemented by HANN. Signal generation and transmittion was performed on PC. As a host, SUN workstation with X-Windows(Motif)is used for graphic representation.

• Journal of Advanced Marine Engineering and Technology
• /
• 제38권10호
• /
• pp.1275-1280
• /
• 2014

In this study, the performance of a real-time micro telescopic monitoring system is evaluated, in which an artificial neural network is adopted for the diagnoses of vibratory bodies, such as solid piping system or machinery. The structural vibration was measured by a non-contact remote sensing method, in which images of a high-speed high-definition camera were used. The structural vibration data that can be obtained by the PIV (particle image velocimetry) technique were used for training the neural network. The structures of the neural network are dynamically changed and their performances are evaluated for the constructed diagnosis system. Optimized structures of the neural network are proposed for real-time diagnosis for the piping system. It was experimentally verified that the performances of the neural network used for real-time monitoring are influenced by the types of the vibration data, such as minimum, maximum and average values of the vibration data. It concludes that the time-mean values are most appropriate for monitoring the piping system.

• 한국정밀공학회지
• /
• 제26권11호
• /
• pp.47-54
• /
• 2009

Spin coaters are the essential instruments in micro-fabrication processes, which apply uniform thin films to flat substrates. In this research, a spin coater diagnosis system is developed to detect the abnormal operation of TFT-LCD process in real time. To facilitate the real-time data acquisition and analysis, the circular-buffered continuous data transfer and the short-time Fourier transform are applied to the fault diagnosis system. To determine whether the system condition is normal or not, a steady-state detection algorithm and a frequency spectrum comparison algorithm using confidence interval are newly devised. Since abnormal condition of a spin coater is rarely encountered, algorithm is tested on a CD-ROM drive and the developed program is verified by a function generator. Actual threshold values for the fault detection are tuned in a spin coater in process.

• Parasites, Hosts and Diseases
• /
• 제48권2호
• /
• pp.99-103
• /
• 2010

Malaria is endemic in the Cukurova region while it is sporadic in other regions of Turkey. Therefore, the laboratory and clinical diagnosis of malaria is important for the treatment of malaria. In this study, 92 blood samples that were taken from the suspected malaria patients for routine diagnosis in a period of 10 years between 1999 and 2009 were analyzed. All of these blood samples were examined by microscopic examinations using Giemsa-stained thick blood films, nested PCR, and real-time PCR. The sensitivity-specificity and positive-negative predictive values for these diagnostic tests were then calculated. It was found that the positive predictive values of microscopic examination of thick blood films, nested PCR, and real-time PCR were 47.8%, 56.5%, and 60.9% for malaria, respectively. The real-time PCR was found to have a specificity of 75% and sensitivity of 100%, while specificity and sensitivity of nested PCR was found 81.2% and 97.7% according to the microscopic examination of thick blood films, respectively.