• Journal of Convergence for Information Technology
• /
• v.12 no.1
• /
• pp.31-38
• /
• 2022

This paper applies an expert independent unsupervised neural network learning-based multivariate time series data analysis model, MSCRED(Multi-Scale Convolutional Recurrent Encoder-Decoder), and to overcome the limitation, because the MCRED is based on Auto-encoder model, that train data must not to be contaminated, by using learning data sampling technique, called Subset Sampling Validation. By using the vibration data of power plant equipment that has been labeled, the classification performance of MSCRED is evaluated with the Anomaly Score in many cases, 1) the abnormal data is mixed with the training data 2) when the abnormal data is removed from the training data in case 1. Through this, this paper presents an expert-independent anomaly diagnosis framework that is strong against error data, and presents a concise and accurate solution in various fields of multivariate time series data.

• Transactions of the Korean Society for Noise and Vibration Engineering
• /
• v.16 no.6 s.111
• /
• pp.565-573
• /
• 2006

Acoustic analysis of a moving sound source required that the measured sound signals be do-Dopplerized and restored as of the original emission signals. The purpose of this research is development of beamforming technique can be applied to the rotor noise source identification. For the do-Dopplerization and reconstruction of emitted sound wave, Forward Propagation Method is applied to the time domain beamforming technique. And validation test were performed using rotating sound source constructed by bended pipe and horn driver. In the validation test using sinusoidal sound wave, sufficient performance of signal processing can be seen, and the effect of measuring duration for accuracy was compared. In the prop-rotor measurements, the acoustic source locations were successfully verified in varying positions for different frequencies and collective pitch angle, in hover condition.

• KIPS Transactions on Software and Data Engineering
• /
• v.4 no.6
• /
• pp.247-252
• /
• 2015

As software has gotten bigger in magnitude and the complexity of software has been increased, the maintenance has gained in-creasing attention for its significant impact on the cost. Identifiers have an impact on more than 90 percent of the readability which accounts for a majority portion of the maintenance activities. For this reason, the existing works focus on domain-specific features based on identifiers. However, their approaches have a limitation when either a class name does not reflect the intention of its context or a class naming is incorrect. Therefore, this paper suggests a series of class name validation process by extracting properties of classes, building learning model by applying a decision tree technique of machine learning, and generating a validation report containing the list of recommendable postfixes of classes to be validated. To evaluate this, four open source projects are selected and indicators such as precision, recall, and ROC curve present the value of this work when it comes to five specific postfixes including functional information on class names.

• Journal of the Korean Society of Marine Environment & Safety
• /
• v.26 no.3
• /
• pp.254-261
• /
• 2020

Welding is the most convenient method for fabricating steel materials to build ships and of shore structures. However, welding using high heat processes inevitably produces welding displacements on welded structures. To mitigate these, heavy industries introduce various welding techniques such as back-step welding and skip-step welding. These techniques effect on the change of the distribution of high heat on welded structures, leading to a reduction of welding displacements. In the present study, various cases using different and newly introduced welding techniques are numerically simulated to ascertain the most efficient technique to minimize welding displacements. A numerical simulation using a finite element method based on the inherent strain, interface element and multi-point constraint function is introduced herein. Based on several simulation results, the optimal welding technique for minimizing welding displacements to build a general ship grillage structure is finally proposed.

• Journal of Astronomy and Space Sciences
• /
• v.28 no.4
• /
• pp.273-284
• /
• 2011

A method of stellar source selection for validating the quality of image is investigated for a low Earth orbit optical remote sensing satellite. Image performance of the optical payload needs to be validated after its launch into orbit. The stellar sources are ideal source points that can be used to validate the quality of optical images. For the image validation, stellar sources should be the brightest as possible in the charge-coupled device dynamic range. The time delayed and integration technique, which is used to observe the ground, is also performed to observe the selected stars. The relations between the incident radiance at aperture and V magnitude of a star are established using Gunn & Stryker's star catalogue of spectrum. Applying this result, an appropriate image performance index is determined, and suitable stars and areas of the sky scene are selected for the optical payload on a remote sensing satellite to observe. The result of this research can be utilized to validate the quality of optical payload of a satellite in orbit.

• Journal of Intelligence and Information Systems
• /
• v.5 no.1
• /
• pp.95-101
• /
• 1999

In this paper, a comparison between different methods to combine predictions from neural networks will be given. These methods are bagging, bumping, and balancing. Those are based on the analysis of the ensemble generalization error into an ambiguity term and a term incorporating generalization performances of individual networks. Neural Networks and AI machine learning models are prone to overfitting. A strategy to prevent a neural network from overfitting, is to stop training in early stage of the learning process. The complete data set is spilt up into a training set and a validation set. Training is stopped when the error on the validation set starts increasing. The stability of the networks is highly dependent on the division in training and validation set, and also on the random initial weights and the chosen minimization procedure. This causes early stopped networks to be rather unstable: a small change in the data or different initial conditions can produce large changes in the prediction. Therefore, it is advisable to apply the same procedure several times starting from different initial weights. This technique is often referred to as training ensembles of neural networks. In this paper, we presented a comparison of three statistical methods to prevent overfitting of neural network.

• 한국지구물리탐사학회:학술대회논문집
• /
• 2003.11a
• /
• pp.352-361
• /
• 2003

We have evaluated the extent of saltwater intrusion from electrical resistivity distribution in a coastal aquifer system in the southeastern part of Busan, Korea. This aquifer system is divided into four layers according to the hydrogeologic characteristics and the horizontal extent of intruded saltwater is determined at each layer through the geostatistical interpretation of electrical resistivity data. In order to define the statistical structure of electrical resistivity data, variogram analysis is carried out to obtain best generalized covariance models. IRF-k (intrinsic random function of order k) kriging is performed with covariance models to produce the plane of spatial mean resistivities. The kriged estimates are evaluated by cross validation to show a good agreement with the true values and the statistics of cross validation represented low errors for the estimates. In the resistivity contour maps more than 5 m below the surface, we can see a dominant direction of saltwater intrusion beginning from the east side. The area of saltwater intrusion increases with depth. The northeast side has low resistivities less than 5 ohm-m due to the presence of saline water in the depth range of 20 m through 70 m. These results show that the application of geostatistical technique to electrical resistivity data is useful for assessing saltwater intrusion in a coastal aquifer system.

• Proceedings of the Korean Institute of Information and Commucation Sciences Conference
• /
• 2012.10a
• /
• pp.495-497
• /
• 2012

Development of application risk management for medical device software test. First, Through questionnaires, Medical device manufacturers, Analysis of software validation and risk management status. Second, Analyzed by comparing the difference between black box testing and white box testing. Third, After analyzing the potential for software analysis tools using code derived factors were quantified, Finally, Medical device risk management process so that it can be applied to build the framework by FMEA(Failure Mode and Effect Analysis) technique. Through this Difficult to build software validation and risk management processes for manufacturers to take advantage of support in medical device GMP(Good Manufacture Practice).

• International Journal of Fuzzy Logic and Intelligent Systems
• /
• v.14 no.4
• /
• pp.332-339
• /
• 2014

Face classification has wide applications in security and surveillance. However, this technique presents various challenges caused by pose, illumination, and expression changes. Face recognition with long-distance images involves additional challenges, owing to focusing problems and motion blurring. Multiple frames under varying spatial or temporal settings can acquire additional information, which can be used to achieve improved classification performance. This study investigates the effectiveness of multi-frame decision-level fusion with photon-counting linear discriminant analysis. Multiple frames generate multiple scores for each class. The fusion process comprises three stages: score normalization, score validation, and score combination. Candidate scores are selected during the score validation process, after the scores are normalized. The score validation process removes bad scores that can degrade the final output. The selected candidate scores are combined using one of the following fusion rules: maximum, averaging, and majority voting. Degraded facial images are employed to demonstrate the robustness of multi-frame decision-level fusion in harsh environments. Out-of-focus and motion blurring point-spread functions are applied to the test images, to simulate long-distance acquisition. Experimental results with three facial data sets indicate the efficiency of the proposed decision-level fusion scheme.

• Nuclear Engineering and Technology
• /
• v.49 no.6
• /
• pp.1358-1367
• /
• 2017

The data integration with modeled predictions (DIMP) model is a promising inverse radiation transport method for solving the special nuclear material (SNM) holdup problem. Unlike previous methods, DIMP is a completely passive nondestructive assay technique that requires no initial assumptions regarding the source distribution or active measurement time. DIMP predicts the most probable source location and distribution through Bayesian inference and quasi-Newtonian optimization of predicted detector responses (using the adjoint transport solution) with measured responses. DIMP performs well with forward hemispherical collimation and unshielded measurements, but several considerations are required when using narrow-view collimated detectors. DIMP converged well to the correct source distribution as the number of synthetic responses increased. DIMP also performed well for the first experimental validation exercise after applying a collimation factor, and sufficiently reducing the source search volume's extent to prevent the optimizer from getting stuck in local minima. DIMP's simple point detector response function (DRF) is being improved to address coplanar false positive/negative responses, and an angular DRF is being considered for integration with the next version of DIMP to account for highly collimated responses. Overall, DIMP shows promise for solving the SNM holdup inverse problem, especially once an improved optimization algorithm is implemented.