• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.21 no.11
• /
• pp.53-60
• /
• 2020

The bogies of railway vehicles are one of the most critical components for service. Fatigue defects in the bogie can be initiated for various reasons, such as material imperfection, welding defects, and unpredictable and excessive overloads during operation. To prevent the derailment of a railway vehicle, it is necessary to evaluate and detect the defect of a connection weldment between the car body and bogie accurately. The safety of the bogie weldment was checked using an ultrasonic test, and it is necessary to determine the occurrence of defects using a learning method. Recently, studies on deep learning have been performed to identify defects with a high recognition rate with respect to a fine and similar defect. In this paper, the databases of weldment specimens with artificial defects were constructed to detect the defect of a bogie weldment. The ultrasonic inspection using the wedge angle was performed to understand the detection ability of fatigue cracks. In addition, the convolutional neural network was applied to minimize human error during the inspection. The results showed that the defects of connection weldment between the car body and bogie could be classified with more than 99.98% accuracy using CNN, and the effectiveness can be verified in the case of an inspection.

• Proceedings of the Korean Society of Machine Tool Engineers Conference
• /
• 2000.04a
• /
• pp.585-590
• /
• 2000

Recently, It is gradually raised necessity that thickness of thin film is measured accuracy and managed in industrial circles and medical world. Ultrasonic Signal processing method is likely to become a very powerful method for NDE method of detection of microdefects and thickness measurement of thin film below the limit of Ultrasonic distance resolution in the opaque materials, provides useful information that cannot be obtained by a conventional measuring system. In the present research, considering a thin film below the limit of ultrasonic distance resolution sandwiched between three substances as acoustical analysis model, demonstrated the usefulness of ultrasonic Signal processing technique using information of ultrasonic frequency for NDE of measurements of thin film thickness, sound velocity, and step height, regardless of interference phenomenon. Numeral information was deduced and quantified effective information from the image. Also, pattern recognition of a defected input image was performed by neural network algorithm. Input pattern of various numeral was composed combinationally, and then, it was studied by neural network. Furthermore, possibility of pattern recognition was confirmed on artifical defected input data formed by simulation. Finally, application on unknown input pattern was also examined.

• Journal of the Semiconductor & Display Technology
• /
• v.19 no.3
• /
• pp.12-18
• /
• 2020

In semiconductor manufacturing, defect detection is critical to maintain high yield. Currently, computer vision systems used in semiconductor photo lithography still have adopt to digital image processing algorithm, which often occur inspection faults due to sensitivity to external environment. Thus, we intend to handle this problem by means of using Mask R-CNN instead of digital image processing algorithm. Additionally, Mask R-CNN can be trained with image dataset pre-processed by means of the specific designed digital image filter to extract the enhanced feature map of Convolutional Neural Network (CNN). Our approach converged advantage of digital image processing and instance segmentation with deep learning yields more efficient semiconductor photo lithography inspection system than conventional system.

• Proceedings of the KIEE Conference
• /
• 1996.07b
• /
• pp.1240-1242
• /
• 1996

In a still mill, the effective surface defect inspection algorithm is necessary. For this purpose, this paper proposed the preprocessing algorithm for surface defect inspection of cold mill strip. This consists of live steps. They are edge detection, binarizing, noise deletion, combining of fragmented defect and selecting the largest defect. Especially, binarizing is a critical problem. Bemuse the performance of the preprocessing is largely depend on the binarized image. So, we develope the adaptive thresholding method, which is multilevel thresholding. The thresholding value is varied according to the mean graylevel value of each test image. To investigate the performance of the proposed algorithm, we classified the detected defect using neural network. The test image is 20 defect images captured at German Sick Co. This algorithm is proved to have good property in cold mill strip surface inspection.

• Journal of Information Technology Services
• /
• v.10 no.3
• /
• pp.113-126
• /
• 2011

The smelting and the continuous casting of steel are important processes that determine the quality of steel products. Especially most of quality defects occur during solidification of the steel continuous casting process. Although quality control techniques such as six sigma, SQC, and TQM can be applied to the continuous casting process for improving quality of steel products, these techniques don't provide real-time analysis to identify the causes of defect occurrence. To solve problems, we have developed a detection model using decision tree which identified abnormal transactions to have a coarse grain structure. And we have compared the proposed model with models using neural network and logistic regression. Experiments on steel data showed that the performance of the proposed model was higher than those of neural network model and logistic regression model. Thus, we expect that the suggested model will be helpful to control the quality of steel products in real-time in the continuous casting process.

• Journal of Internet Computing and Services
• /
• v.21 no.6
• /
• pp.23-31
• /
• 2020

This study was carried out to generate various images of railroad surfaces with random defects as training data to be better at the detection of defects. Defects on the surface of railroads are caused by various factors such as friction between track binding devices and adjacent tracks and can cause accidents such as broken rails, so railroad maintenance for defects is necessary. Therefore, various researches on defect detection and inspection using image processing or machine learning on railway surface images have been conducted to automate railroad inspection and to reduce railroad maintenance costs. In general, the performance of the image processing analysis method and machine learning technology is affected by the quantity and quality of data. For this reason, some researches require specific devices or vehicles to acquire images of the track surface at regular intervals to obtain a database of various railway surface images. On the contrary, in this study, in order to reduce and improve the operating cost of image acquisition, we constructed the 'Defective Railroad Surface Regeneration Model' by applying the methods presented in the related studies of the Generative Adversarial Network (GAN). Thus, we aimed to detect defects on railroad surface even without a dedicated database. This constructed model is designed to learn to generate the railroad surface combining the different railroad surface textures and the original surface, considering the ground truth of the railroad defects. The generated images of the railroad surface were used as training data in defect detection network, which is based on Fully Convolutional Network (FCN). To validate its performance, we clustered and divided the railroad data into three subsets, one subset as original railroad texture images and the remaining two subsets as another railroad surface texture images. In the first experiment, we used only original texture images for training sets in the defect detection model. And in the second experiment, we trained the generated images that were generated by combining the original images with a few railroad textures of the other images. Each defect detection model was evaluated in terms of 'intersection of union(IoU)' and F1-score measures with ground truths. As a result, the scores increased by about 10~15% when the generated images were used, compared to the case that only the original images were used. This proves that it is possible to detect defects by using the existing data and a few different texture images, even for the railroad surface images in which dedicated training database is not constructed.

• The Journal of the Institute of Internet, Broadcasting and Communication
• /
• v.11 no.4
• /
• pp.99-104
• /
• 2011

In this paper the approach of neural network was proposed which detects a variety of defects in the molded parts. In an attempt to improve the response of the system, It is designed to minimize the use of memory via LookUp table in software. The goal of these methods was to extract the features of samples in learning of neural networks, overcoming the algorithms of defects detection and classification. Through the learning of 500 sample patterns of molded parts, defects of 3% molded parts was detected and classified as the incorrect diameter parts. We expect that proposed approach is an effective alternative to save test time and cost for defect detection of a fine pattern within the molded parts.

• Proceedings of the Korean Institute of Intelligent Systems Conference
• /
• 2003.09a
• /
• pp.539-542
• /
• 2003

A motor is the workhorse of our industry. The issues of preventive and condition-based maintenance, online monitoring, system fault detection, diagnosis, and prognosis are of increasing importance. Different internal motor faults (e.g., inter-turn short circuits, broken bearings, broken rotor bars) along with external motor faults (e.g., phase failure, mechanical overload, blocked rotor) are expected to happen sooner or later. This paper introduces the fault detection technique of induction motors based upon the stator current. The fault motors have rotor bar broken or rotor unbalance defect, respectively. The stator currents are measured by the current meters and stored by the time domain. The time domain is not suitable to represent the current signals, so the frequency domain is applied to display the signals. The Fourier Transformer is used for the conversion of the signal. After the conversion of the signals, the features of the signals have to be extracted by the signal processing methods like a wavelet analysis, a spectrum analysis, etc. The discovered features are entered to the pattern classification model such as a neural network model, a polynomial neural network, a fuzzy inference model, etc. This paper describes the fault detection results that use wavelet decomposition. The wavelet analysis is very useful method for the time and frequency domain each. Also it is powerful method to detect the features in the signals.

• Smart Structures and Systems
• /
• v.31 no.4
• /
• pp.383-392
• /
• 2023

Post-earthquake building condition assessment is crucial for subsequent rescue and remediation and can be automated by emerging computer vision and deep learning technologies. This study is based on an endeavour for the 2nd International Competition of Structural Health Monitoring (IC-SHM 2021). The task package includes five image segmentation objectives - defects (crack/spall/rebar exposure), structural component, and damage state. The structural component and damage state tasks are identified as the priority that can form actionable decisions. A multi-task Convolutional Neural Network (CNN) is proposed to conduct the two major tasks simultaneously. The rest 3 sub-tasks (spall/crack/rebar exposure) were incorporated as auxiliary tasks. By synchronously learning defect information (spall/crack/rebar exposure), the multi-task CNN model outperforms the counterpart single-task models in recognizing structural components and estimating damage states. Particularly, the pixel-level damage state estimation witnesses a mIoU (mean intersection over union) improvement from 0.5855 to 0.6374. For the defect detection tasks, rebar exposure is omitted due to the extremely biased sample distribution. The segmentations of crack and spall are automated by single-task U-Net but with extra efforts to resample the provided data. The segmentation of small objects (spall and crack) benefits from the resampling method, with a substantial IoU increment of nearly 10%.

• The Transactions of The Korean Institute of Electrical Engineers
• /
• v.67 no.5
• /
• pp.669-674
• /
• 2018

Data augmentation is an efficient way to reduce overfitting on models and to improve a performance supplementing extra data for training. It is more important in deep learning based industrial machine vision. Because deep learning requires huge scale of learning data to learn a model, but acquisition of data can be limited in most of industrial applications. A very generic method for augmenting image data is to perform geometric transformations, such as cropping, rotating, translating and adjusting brightness of the image. The effectiveness of data augmentation in image classification has been reported, but it is rare in defect inspections. We explore and compare various basic augmenting operations for the metal surface defects. The experiments were executed for various types of defects and different CNN networks and analysed for performance improvements by the data augmentations.