• Journal of Navigation and Port Research
• /
• v.46 no.4
• /
• pp.367-374
• /
• 2022

The purpose of this study was to propose a deep learning algorithm that applies to the fault diagnosis of fuel pumps and purifiers of autonomous ships. A deep learning algorithm reflecting the time dependence of the measured signal was configured, and the failure pattern was trained using the vibration signal, measured in the equipment's regular operation and failure state. Considering the sequential time-dependence of deterioration implied in the vibration signal, this study adopts Conv1D with sliding window computation for fault detection. The time dependence was also reflected, by transferring the measured signal from two-dimensional to three-dimensional. Additionally, the optimal values of the hyper-parameters of the Conv1D model were determined, using the grid search technique. Finally, the results show that the proposed data preprocessing method as well as the Conv1D model, can reflect the sequential dependency between the fault and its effect on the measured signal, and appropriately perform anomaly as well as failure detection, of the equipment chosen for application.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.19 no.1
• /
• pp.75-84
• /
• 2018

Increasing the operation rate of components and stabilizing the operation through timely management of the core parts are crucial for improving the efficiency of the railroad maintenance industry. The demand for diagnosis technology to assess the condition of rolling stock components, which employs history management and automated big data analysis, has increased to satisfy both aspects of increasing reliability and reducing the maintenance cost of the core components to cope with the trend of rapid maintenance. This study developed a big data platform-based system to manage the rolling stock component condition to acquire, process, and analyze the big data generated at onboard and wayside devices of railroad cars in real time. The system can monitor the conditions of the railroad car component and system resources in real time. The study also proposed a machine learning technique that enabled the distributed and parallel processing of the acquired big data and automatic component fault diagnosis. The test, which used the virtual instance generation system of the Amazon Web Service, proved that the algorithm applying the distributed and parallel technology decreased the runtime and confirmed the fault diagnosis model utilizing the random forest machine learning for predicting the condition of the bearing and wheel parts with 83% accuracy.

• The Transactions of the Korean Institute of Electrical Engineers A
• /
• v.48 no.10
• /
• pp.1272-1278
• /
• 1999

A fault is considered as a variation of physical parameters; therefore the design of fault detection and identification(FDI) can be reduced to the parameter identification of a non linear system and to the association of the set of the estimated parameters with the mode of faults. Neuro-Fuzzy Inference System which contains multiple linear models as consequent part is used to model nonlinear systems. Generally, the linear parameters in neuro-fuzzy inference system can be effectively utilized to fault diagnosis. In this paper, we proposes an FDI system for nonlinear systems using neuro-fuzzy inference system. The proposed diagnostic system consists of two neuro-fuzzy inference systems which operate in two different modes (parallel and series-parallel mode). It generates the parameter residuals associated with each modes of faults which can be further processed by additional RBF (Radial Basis Function) network to identify the faults. The proposed FDI scheme has been tested by simulation on two-tank system.

• Journal of Magnetics
• /
• v.18 no.2
• /
• pp.135-141
• /
• 2013

This paper proposes an efficient fault detection method for onboard degaussing coils which are installed to minimize underwater magnetic fields due to the ferromagnetic hull. To achieve this, the method basically uses field signals measured at specific magnetic treatment facilities instead of time-consuming numerical field solutions in a three-dimensional analysis space. In addition, an analytical design sensitivity formula and the linear property of degaussing coil fields is being exploited for detecting fault coil positions and assessing individual degaussing coil currents. Such peculiar features make it possible to yield fast and accurate results on the fault detection of degaussing coils. For foreseeable fault conditions, the proposed method is tested with a model ship equipped with 20 degaussing coils.

• Journal of Institute of Control, Robotics and Systems
• /
• v.17 no.11
• /
• pp.1117-1124
• /
• 2011

An integrated fault detection and isolation method is proposed in this paper. The main objective of this paper is development fault detection, isolation and diagnosis algorithm based on the DKF (Decentralized Kalman Filter) and the bank of IMM (Interacting Multiple Model) filters using penalty scalar for both partial and total faults and the outlier detection algorithm for preventing false alarm also included. The proposed FDI (Fault Detection and Isolation) scheme is developed in four phases. In the first phase, the outlier detection filter is designed to prevent false alarm as a pre-filter. In the second phases, two local filters and master filter are designed to detect sensor faults. In the third phases, the proposed FDI scheme checks sensor residual to isolate sensor faults and 11 EKFs actuator fault models are designed to detect wherever actuator faults occur. In the last phases, four filters are designed to identify the fault type which is either the total fault or partial fault. The developed scheme can deal with not only sensor and actuator faults, but also preventing false alarm. An important feature of the proposed FDI scheme can decreases fault isolation time and figure out not only fault detection and isolation but also fault type identification. To verify the proposed FDI algorithm performance, the Simulator is also developed under the Matlab/Simulink environment.

• Proceedings of the KIEE Conference
• /
• 1997.07b
• /
• pp.456-458
• /
• 1997

본 논문에서는 플랜트의 계측신호들에 나타나는 현상적인 특징파형들을 감지하여 플랜트의 고장을 진단하는 기법을 제시하고, 플랜트의 고장시 나타나는 여러 특징 파형들의 전형인 oscillatory 파형을 감지하는 알고리즘을 prediction filtering 기법을 적용하여 얻는 과정을 기술하였다. 또한 이의 구체적 적용 예로 500MW급 화력발전소의 보일러 제어 시스템을 대상으로 여러 고장상황에 대한 고장감지 알고리즘을 설계하였고 제시한 기법의 유용성을 보이기 위한 시뮬레이션 적용 결과를 보였다.

• Korean Chemical Engineering Research
• /
• v.58 no.3
• /
• pp.408-415
• /
• 2020

In this study, the AE-SOM method, which combines auto-encoder and self-organizing map, is used to detect and diagnose faults in EVA production process. Then, the fault propagation pathways are identified using Granger causality test. One year and seven months of operation data were obtained to detect faults of the process, and the process variables of the autoclave reactor are mainly analyzed. In the data pretreatment process, the data are standardized and 200 samples of each grade are randomly chosen to obtain a fault detection model. After that, the best matching unit (BMU) of each grade is confirmed by applying AE-SOM. The faults are determined based on each BMU. When a fault is found, the most causative variable of the fault is identified by using a contribution plot, and the fault propagation pathway is identified by Granger causality test. The prognostic of the two shutdowns is detected, and the fault propagation pathway caused by the faulty variable was analyzed.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.20 no.9
• /
• pp.551-557
• /
• 2019

Diagnosis of abnormal faults is essential for producing high quality products. The role of real-time diagnosis is quite increasing in the batch processes of producing high value-added products such as semiconductors, pharmaceuticals, and so forth. In this study, we evaluate the effect of variable selection and future-value estimation techniques on the performance of the diagnosis system, which is based on nonlinear classification and measurement data. The diagnostic performance can be improved by selecting only the variables that are important and have high contribution for diagnosis. Thus, the diagnostic performance of several variable selection techniques is compared and evaluated. In addition, missing data of a new batch, called future observations, should be estimated because the full data of a new batch is not available before the end of the cycle. In this work the use of different estimation techniques is analyzed. A case study on the polyvinyl chloride batch process was carried out so that optimal variable selection and estimation methods were obtained: maximum 21.9% and 13.3% improvement by variable selection and maximum 25.8% and 15.2% improvement by estimation methods.

• The Journal of the Convergence on Culture Technology
• /
• v.9 no.3
• /
• pp.845-850
• /
• 2023

In this paper, we propose a method for diagnosing ball bearing vibration using transfer learning. STFT, which can analyze vibration signals in time-frequency, was used as input to CNN to diagnose failures. In order to rapidly learn CNN-based deep artificial neural networks and improve diagnostic performance, we proposed a transfer learning-based deep learning learning technique. For transfer learning, the feature extractor and classifier were selectively learned using a VGG-based image classification model, the data set for learning was publicly available ball bearing vibration data provided by Case Western Reserve University, and performance was evaluated by comparing the proposed method with the existing CNN model. Experimental results not only prove that transfer learning is useful for condition diagnosis in ball bearing vibration data, but also allow other industries to use transfer learning to improve condition diagnosis.

• International Journal of Safety
• /
• v.11 no.1
• /
• pp.26-32
• /
• 2012

This paper introduces a feasibility evaluation method for prognosis systems based on an empirical model in nuclear power plants. By exploiting the dynamical signature characterized by abnormal phenomena, the prognosis technique can be applied to detect the plant abnormal states prior to an unexpected plant trip. Early $operator^{\circ}{\emptyset}s$ awareness can extend available time for operation action; therefore, unexpected plant trip and time-consuming maintenance can be reduced. For the practical application in nuclear power plant, it is important not only to enhance the advantages of prognosis systems, but also to quantify the negative impact in prognosis, e.g., uncertainty. In order to apply these prognosis systems to real nuclear power plants, it is necessary to conduct a feasibility evaluation; the evaluation consists of 4 steps (: the development of an evaluation method, the development of selection criteria for the abnormal state, acquisition and signal processing, and an evaluation experiment). In this paper, we introduce the feasibility evaluation method and propose further study points for applying prognosis systems from KHNP's experiences in testing some prognosis technologies available in the market.