• Journal of the Korean Geotechnical Society
• /
• v.21 no.5
• /
• pp.51-58
• /
• 2005

In this study, the large scale model tests were executed to estimate the Load Sharing Ratio (LSR) of raft in a piled footing under various conditions. The conditions such as the subsoil type, pile length, pile spacing, away type and pile installation method etc. were varied in the pile loading tests about the free-standing group piles and a piled footing. As a result of this study, it was found that there was no difference in the load-settlement curves, resulting from the pile installation method and subsoil type. The piles supported most of the external load until a yielding load of the piled footing, but the raft supported a considerable load after a yielding load. As the relative density of sands increased, the LSR decreased. As the pile spacing was wider and the pile length increased, there was a tendancy for the LSR to increase. But it was also found that the LSR was not affected by the pile installation method and the subsoil type.

• International Journal of Naval Architecture and Ocean Engineering
• /
• v.8 no.4
• /
• pp.330-343
• /
• 2016

This paper presents results of an experimental investigation of vortex-induced vibration (VIV) of a flexibly mounted and rigid cylinder with two-degrees-of-freedom with respect to varying ratio of in-line natural frequency to cross-flow natural frequency, $f^*$, at a fixed low mass ratio. Combined in-line and cross-flow motion was observed in a sub-critical Reynolds number range. Three-dimensional displacement meter and tension meter were used to measure dynamic responses of the model. To validate the results and the experiment system, x and y response amplitudes and ratio of oscillation frequency to cross-flow natural frequency were compared with other experimental results. It has been found that the higher harmonics, such as third and more vibration components, can occur on a certain part of steel catenary riser under a condition of dual resonance mode. In the present work, however, due to the limitation of a size of circulating water channel, the whole test of a whole configuration of the riser at an adequate scale for VIV phenomenon was not able to be conducted. Instead, we have modeled a rigid cylinder and assumed that the cylinder is a part of steel catenary riser where the higher harmonic motions could occur. Through the experiment, we have found that even though the cylinder was assumed to be rigid, the occurrence of the higher harmonic motions was observed in a small reduced velocity ($V_r$) range, where the influence of the in-line response is relatively large. The transition of the vortex shedding mode from one to another was examined by using time history of x and y directional displacement over all experimental cases. We also observed the influence of in-line restoring force power spectral density with $f^*$.

• Journal of the Korean GEO-environmental Society
• /
• v.6 no.1
• /
• pp.73-82
• /
• 2005

Generally, to evaluate a slope stability of the geosynthetic reinforced soil slope, the modified version of limit equilibrium method can be used. In most cases, resisting effects of reinforcement are dealt with considering an increased shear strength on the potential slip surface. However, it is not clear that the methods satisfy all three equilibrium equations. In this study, the new slope stability analysis method in which not only reinforcing effects of geosynthetics can be considered but also all three equilibrium equation can be satisfied is proposed. A number of illustrative examples, including published load test of large-scale reinforced retaining wall and centrifuge model tests on the geotextile reinforced soil slopes, are also analyzed. As a result, it is shown that the newly suggested method produces a relatively accurate factor of safety.

• Journal of Korean Tunnelling and Underground Space Association
• /
• v.11 no.4
• /
• pp.419-425
• /
• 2009

Recently, the number of shallow tunnel construction increases to improve the structural safety and environment-friendliness. In Semi-Cut and Cover Method, ground is excavated to the crown arch level and braced rib arch is set to backfill before the excavation of lower face. Semi-Cut and Cover Method is proposed to solve the problems occurred by the conventional Cut and Cover Method, such as unstability, high-cost and the large cutting slope to be reinforced. In this paper, the behaviors of Braced Rib Arch in shallow tunnel excavated by semi-cut and cover method was studied. Model tests in 1:10 Scale were performed in real construction sequences. The distance between supports of rib arch was 1.8 m and the length of spacer was 1.0 m. the size of test pit was 4.0 m (width)$\times$3.3 m (length) 4.0 m (height) in dimension. Tests results show that backfill load acting on arch was smaller than that in the conventional Open-Cut Method.

• Korean Journal of Cognitive Science
• /
• v.34 no.3
• /
• pp.197-226
• /
• 2023

Generative Text Summarization is one of the Natural Language Processing tasks. It generates a short abbreviated summary while preserving the content of the long text. ROUGE is a widely used lexical-overlap based metric for text summarization models in generative summarization benchmarks. Although it shows very high performance, the studies report that 30% of the generated summary and the text are still inconsistent. This paper proposes a methodology for evaluating the performance of the summary model without using the correct summary. AggreFACT is a human-annotated dataset that classifies the types of errors in neural text summarization models. Among all the test candidates, the two cases, generation summary, and when errors occurred throughout the summary showed the highest correlation results. We observed that the proposed evaluation score showed a high correlation with models finetuned with BART and PEGASUS, which is pretrained with a large-scale Transformer structure.

• Smart Media Journal
• /
• v.9 no.4
• /
• pp.26-35
• /
• 2020

Attempts to predict future stock prices have been studied steadily since the past. However, unlike general time-series data, financial time-series data has various obstacles to making predictions such as non-stationarity, long-term dependence, and non-linearity. In addition, variables of a wide range of data have limitations in the selection by humans, and the model should be able to automatically extract variables well. In this paper, we propose a 'sliding time step normalization' method that can normalize non-stationary data and LSTM autoencoder to compress variables from all variables. and 'moving transfer learning', which divides periods and performs transfer learning. In addition, the experiment shows that the performance is superior when using as many variables as possible through the neural network rather than using only 100 major financial variables and by using 'sliding time step normalization' to normalize the non-stationarity of data in all sections, it is shown to be effective in improving performance. 'moving transfer learning' shows that it is effective in improving the performance in long test intervals by evaluating the performance of the model and performing transfer learning in the test interval for each step.

• Smart Structures and Systems
• /
• v.29 no.1
• /
• pp.1-16
• /
• 2022

Despite recent breakthroughs in deep learning and computer vision fields, the pixel-wise identification of tiny objects in high-resolution images with complex disturbances remains challenging. This study proposes a modified U-net for tiny crack segmentation in real-world steel-box-girder bridges. The modified U-net adopts the common U-net framework and a novel Self-Attention-Self-Adaption (SASA) neuron as the fundamental computing element. The Self-Attention module applies softmax and gate operations to obtain the attention vector. It enables the neuron to focus on the most significant receptive fields when processing large-scale feature maps. The Self-Adaption module consists of a multiplayer perceptron subnet and achieves deeper feature extraction inside a single neuron. For data augmentation, a grid-based crack random elastic deformation (CRED) algorithm is designed to enrich the diversities and irregular shapes of distributed cracks. Grid-based uniform control nodes are first set on both input images and binary labels, random offsets are then employed on these control nodes, and bilinear interpolation is performed for the rest pixels. The proposed SASA neuron and CRED algorithm are simultaneously deployed to train the modified U-net. 200 raw images with a high resolution of 4928 × 3264 are collected, 160 for training and the rest 40 for the test. 512 × 512 patches are generated from the original images by a sliding window with an overlap of 256 as inputs. Results show that the average IoU between the recognized and ground-truth cracks reaches 0.409, which is 29.8% higher than the regular U-net. A five-fold cross-validation study is performed to verify that the proposed method is robust to different training and test images. Ablation experiments further demonstrate the effectiveness of the proposed SASA neuron and CRED algorithm. Promotions of the average IoU individually utilizing the SASA and CRED module add up to the final promotion of the full model, indicating that the SASA and CRED modules contribute to the different stages of model and data in the training process.

• Journal of Nuclear Fuel Cycle and Waste Technology(JNFCWT)
• /
• v.9 no.1
• /
• pp.23-32
• /
• 2011

An engineering scale test, which is called KENTEX, was carried out to understand and to analyze the coupled thermal, hydrological and mechanical phenomena in the engineered barrier system(EBS) of Korean reference disposal system. Using the experimental data obtained from KENTEX, the water saturation processes in bentonite could be analyzed. From the comparison between the model calculation using ABAQUS and the experimental results, the difference of the water content between them in the unsaturating part was large because the drying phenomena due to moisture redistribution by the temperature gradient could not be included in the model. In the saturating part, the difference of the water content between them was decreased gradually and showed to be small in the full saturation. And the time of about 95% saturation could be estimated about 500 days from the model calculation and experimental results. Also it could be known that the moisture redistribution in the unsaturated part could not be affected on the saturation time of bentonite in the repository. Therefore, it is considered that this model could be used to quantitatively predict the water saturation time in bentonite as EBS for the disposal system.

• KSCE Journal of Civil and Environmental Engineering Research
• /
• v.36 no.3
• /
• pp.521-528
• /
• 2016

The tollbooths of a main motorway toll plaza are usually operated proactively responding to the variations of traffic demands of two-type vehicles, i.e. cars and the other (heavy) vehicles, respectively. In this vein, it is one of key elements to forecast accurate traffic volumes for the two vehicle types in advanced tollgate operation. Unfortunately, it is not easy for existing univariate short-term prediction techniques to simultaneously generate the two-vehicle-type traffic demands in literature. These practical and academic backgrounds make it one of attractive research topics in Intelligent Transportation System (ITS) forecasting area to forecast the future traffic volumes of the two-type vehicles at an acceptable level of accuracy. In order to address the shortcomings of univariate short-term prediction techniques, a Multiple In-and-Out (MIO) forecasting model to simultaneously generate the two-type traffic volumes is introduced in this article. The MIO model based on a non-parametric approach is devised under the on-line access conditions of large-scale historical data. In a feasible test with actual data, the proposed model outperformed Kalman filtering, one of a widely-used univariate models, in terms of prediction accuracy in spite of multivariate prediction scheme.

• Journal of KIISE:Databases
• /
• v.35 no.6
• /
• pp.505-517
• /
• 2008

This paper proposes a rewriting algorithm of OWL-DL ontology query in SPARQL. Currently, to obtain inference results of given SPARQL queries, Web ontology repositories construct inference ontology models and match the SPARQL queries with the models. However, an inference model requires much larger space than its original base model, and reusability of the model is not available for other inferrable SPARQL queries. Therefore, the aforementioned approach is not suitable for large scale SPARQL query processing. To resolve tills issue, this paper proposes a novel SPARQL query rewriting algorithm that can obtain results by rewriting SPARQL queries and accomplishing query operations against the base ontology model. To achieve this goal, we first define OWL-DL inference rules and apply them on rewriting graph pattern in queries. The paper categorizes the inference rules and discusses on how these rules affect the query rewriting. To show the advantages of our proposal, a prototype system based on lena is implemented. For comparative evaluation, we conduct an experiment with a set of test queries and compare of our proposal with the previous approach. The evaluation result showed the proposed algorithm supports an improved performance in efficiency of the inferrable SPARQL query processing without loss of completeness and soundness.