• The Journal of the Convergence on Culture Technology
• /
• v.9 no.1
• /
• pp.649-654
• /
• 2023

This paper proposes the design of a neural network structure search model using graph convolutional neural networks. Deep learning has a problem of not being able to verify whether the designed model has a structure with optimized performance due to the nature of learning as a black box. The neural network structure search model is composed of a recurrent neural network that creates a model and a convolutional neural network that is the generated network. Conventional neural network structure search models use recurrent neural networks, but in this paper, we propose GC-NAS, which uses graph convolutional neural networks instead of recurrent neural networks to create convolutional neural network models. The proposed GC-NAS uses the Layer Extraction Block to explore depth, and the Hyper Parameter Prediction Block to explore spatial and temporal information (hyper parameters) based on depth information in parallel. Therefore, since the depth information is reflected, the search area is wider, and the purpose of the search area of the model is clear by conducting a parallel search with depth information, so it is judged to be superior in theoretical structure compared to GC-NAS. GC-NAS is expected to solve the problem of the high-dimensional time axis and the range of spatial search of recurrent neural networks in the existing neural network structure search model through the graph convolutional neural network block and graph generation algorithm. In addition, we hope that the GC-NAS proposed in this paper will serve as an opportunity for active research on the application of graph convolutional neural networks to neural network structure search.

• Proceedings of the Korea Water Resources Association Conference
• /
• 2015.05a
• /
• pp.557-557
• /
• 2015

To overcome main drawbacks of parametric models, k-nearest neighbor resampling (KNNR) is suggested for water quality analysis involving geographic information. However, with KNNR nonparametric model, Geographic information is not properly handled. In the current study, to manipulate geographic information properly, we introduce a depth function which is a novel statistical concept in the classical KNNR model for stormwater quality simulation. An application is presented for a case study of the total suspended solids throughout the entire United States. Total suspended solids concentration data of stormwater demonstrated that the proposed model significantly improves the simulation performance rather than the existing KNNR model.

• Smart Structures and Systems
• /
• v.33 no.5
• /
• pp.325-332
• /
• 2024

This study proposes a novel long short-term memory (LSTM)-based approach for predicting carbonation depth, with the aim of enhancing the durability evaluation of concrete structures. Conventional carbonation depth prediction relies on statistical methodologies using carbonation influencing factors and in-situ carbonation depth data. However, applying in-situ data for predictive modeling faces challenges due to the lack of time-series data. To address this limitation, an LSTM-based carbonation depth prediction technique is proposed. First, training data are generated through random sampling from the distribution of carbonation velocity coefficients, which are calculated from in-situ carbonation depth data. Subsequently, a Bayesian theorem is applied to tailor the training data for each target bridge, which are depending on surrounding environmental conditions. Ultimately, the LSTM model predicts the time-dependent carbonation depth data for the target bridge. To examine the feasibility of this technique, a carbonation depth dataset from 3,960 in-situ bridges was used for training, and untrained time-series data from the Miho River bridge in the Republic of Korea were used for experimental validation. The results of the experimental validation demonstrate a significant reduction in prediction error from 8.19% to 1.75% compared with the conventional statistical method. Furthermore, the LSTM prediction result can be enhanced by sequentially updating the LSTM model using actual time-series measurement data.

• Journal of information and communication convergence engineering
• /
• v.9 no.2
• /
• pp.197-200
• /
• 2011

In this paper, we propose a multi-view stereoscopic image synthesis algorithm for 3DTV system using depth information with an RGB texture from a depth camera. The proposed algorithm synthesizes multi-view images which a virtual convergence camera model could generate. Experimental results showed that the performance of the proposed algorithm is better than those of conventional methods.

• Structural Engineering and Mechanics
• /
• v.18 no.2
• /
• pp.263-276
• /
• 2004

The purpose of this paper is to determine the subsoil depth affected from the load on the plate resting on elastic foundation using stress distribution within the subsoil that will be occurred depending on the loading and dimension of the plate. An iterative method is developed in order to determine the effective depth of the subsoil under the plate. Numerical examples from the technical literature are solved by means of the method suggested herein and displacements, bending moments and shear forces are presented in graphical and tabular forms to evaluate the effects of the limit depth considered in the study. Results showed the efficiency and simplicity of the present approach for the plate resting on an elastic foundation.

• Current Optics and Photonics
• /
• v.3 no.5
• /
• pp.458-465
• /
• 2019

The thermal effect and the light output of a laser crystal under different pumping depths were reported., Based on the thermal model of a single-ended pumped Tm:YAP crystal, the thermal stress coupled model used Comsol to theoretically calculate the effect of changing the pump spot size and pump depth on crystal heat distribution and stress distribution. The experimental results showed that the laser output power first increased and then decreased with increasing pump spot size. As the depth of focus increased, the laser output power first increased and then decreased. The experimental results were consistent with the theoretical simulation results. The theory of pump spot radius and depth of focus in this paper provided an effective simulation method for mitigating thermal effects, and provided theoretical supports for laser crystals to obtain higher laser output power.

• Journal of Ocean Engineering and Technology
• /
• v.36 no.2
• /
• pp.83-90
• /
• 2022

This paper presents the minimum submergence depth of an underwater vehicle that can remove the effect of free surface on the resistance of the underwater vehicle. The total resistance of the underwater vehicle in fully submerged modes comprises only viscous pressure and friction resistances, and no wave resistance should be present, based on the free surface effect. In a model test performed in this study, the resistance is measured in the range of 2 to 10 kn (1.03-5.14 m/s) under depth conditions of 850 mm (2.6D) and 1250 mm (3.8D), respectively, and the residual resistance coefficients are compared. Subsequently, resistance analysis is performed via computational fluid dynamics (CFD) simulation to investigate the free surface effect based on various submergence depths. First, the numerical analysis results in the absence of free surface conditions and the model test results are compared to show the tendency of the resistance coefficients and the reliability of the CFD simulation results. Subsequently, numerical analysis results of submergence depth presented in a reference paper are compared with the model test results. These two sets of results confirm that the resistance increased due to the free surface effect as the high speed and depth approach the free surface. Therefore, to identify a fully submerged depth that is not affected by the free surface effect, case studies for various depths are conducted via numerical analysis, and a correlation for the fully submerged depth based on the Froude number of an underwater vehicle is derived.

• Proceedings of the KSRS Conference
• /
• 2003.11a
• /
• pp.304-306
• /
• 2003

Oceanic pycnocline depth is usually obtained from in situ measurements. As ocean internal waves occur on and propagate along oceanic pycnocline, it is possible to estimate the depth remotely. This paper presents a method for retrieving pycnocline depth from synthetic aperture radar (SAR) imagery where internal waves are visible. This model is constructed by combining a two-layer ocean model and a nonlinear internal wave model. It is also assumed that the observed groups of internal wave packets on SAR imagery are generated by local semidiurnal tides. Case study in East China Sea shows a good agreement with in situ CTD data.

• Advances in concrete construction
• /
• v.2 no.3
• /
• pp.229-247
• /
• 2014

A numerical study of the influence of shear-span/depth ratio on the cohesive crack fracture parameters and double - K fracture parameters of concrete is carried out in this paper. For the study the standard bending specimen geometry loaded with four point bending test is used. For four point loading, the shear - span/depth ratio is varied as 0.4, 1 and 1.75 and the ao/D ratio is varied from 0.2, 0.3 and 0.4 for laboratory specimens having size range from 100 - 500 mm. The input parameters for determining the double - K fracture parameters are taken from the developed fictitious crack model. It is found that the cohesive crack fracture parameters are independent of shear-span/depth ratio. Further, the unstable fracture toughness of double-K fracture model is independent of shear-span/depth ratio whereas, the initial cracking toughness of the material is dependent on the shear-span/depth ratio.

• Journal of Broadcast Engineering
• /
• v.23 no.5
• /
• pp.614-621
• /
• 2018

This paper proposes a method of restoring corrupted depth image captured by depth camera through unsupervised learning using generative adversarial network (GAN). The proposed method generates restored face depth images using 3D morphable model convolutional neural network (3DMM CNN) with large-scale CelebFaces Attribute (CelebA) and FaceWarehouse dataset for training deep convolutional generative adversarial network (DCGAN). The generator and discriminator equip with Wasserstein distance for loss function by utilizing minimax game. Then the DCGAN restore the loss of captured facial depth images by performing another learning procedure using trained generator and new loss function.