• Journal of the Computational Structural Engineering Institute of Korea
• /
• v.34 no.3
• /
• pp.167-174
• /
• 2021

Recent investigation into the integrity of nuclear containment buildings has highlighted the importance of developing an elaborate diagnostic method to evaluate the distribution and size of cavities inside concrete walls. As part of developing such a method, this paper presents a finite element approach to modeling elastic waves propagating in the containment building walls of a nuclear power plant. We introduce a perfectly matched layer (PML) wave-absorbing boundary to limit the large-scale nuclear containment wall to the region of interest. The formulation results in a semi-discrete form with symmetric damping and stiffness matrices. The transient elastic wave equations for a mixed unsplit-field PML were solved for displacement and stresses in the time domain. Numerical results show that the sensitivity of displacement, velocity, acceleration, and stresses is large depending on the size and location of the cavity. The dynamic response of the wall slightly differs depending on the existence of the containment liner plate. The results of this study can be applied to a full-waveform inversion approach for characterizing cavities inside a containment wall.

• Journal of the Computational Structural Engineering Institute of Korea
• /
• v.36 no.6
• /
• pp.355-364
• /
• 2023

The arbitrary Lagrangian-Eulerian (ALE) method has been extensively researched owing to its capability to accurately predict the propagation of blast shock waves. Although the use of the ALE method for dynamic analysis can produce unreliable results depending on the mesh size of the finite element, few studies have explored the relationship between the mesh size for the air domain and the accuracy of numerical analysis. In this study, we propose a procedure to calculate the optimal mesh size based on the mean squared error between the maximum blast pressure values obtained from numerical simulations and experiments. Furthermore, we analyze the relationship between the weight of explosive material (TNT) and the optimal mesh size of the air domain. The findings from this study can contribute to estimating the optimal mesh size in blast simulations with various explosion weights and promote the development of advanced blast numerical analysis models.

• Journal of Korean Society of Transportation
• /
• v.29 no.5
• /
• pp.121-138
• /
• 2011

More affordable and available cutting-edge technologies (e.g., wireless vehicle communication) are regarded as a possible alternative to the fixed infrastructure-based traffic information system requiring the expensive infrastructure investments and mostly implemented in the uninterrupted freeway network with limited spatial system expansion. This paper develops an advanced decentralized traveler information System (ATIS) using vehicle-to-vehicle (V2V) communication system whose performance (drivers' travel time savings) are enhanced by three complementary functions (autonomous automatic incident detection algorithm, reliable sample size function, and driver behavior model) and evaluates it in the typical $6{\times}6$ urban grid network with non-recurrent traffic state (traffic incident) with the varying key parameters (traffic flow, communication radio range, and penetration ratio), employing the off-the-shelf microscopic simulation model (VISSIM) under the ideal vehicle communication environment. Simulation outputs indicate that as the three key parameters are increased more participating vehicles are involved for traffic data propagation in the less communication groups at the faster data dissemination speed. Also, participating vehicles saved their travel time by dynamically updating the up-to-date traffic states and searching for the new route. Focusing on the travel time difference of (instant) re-routing vehicles, lower traffic flow cases saved more time than higher traffic flow ones. This is because a relatively small number of vehicles in 300vph case re-route during the most system-efficient time period (the early time of the traffic incident) but more vehicles in 514vph case re-route during less system-efficient time period, even after the incident is resolved. Also, normally re-routings on the network-entering links saved more travel time than any other places inside the network except the case where the direct effect of traffic incident triggers vehicle re-routings during the effective incident time period and the location and direction of the incident link determines the spatial distribution of re-routing vehicles.

• Journal of Dental Rehabilitation and Applied Science
• /
• v.26 no.2
• /
• pp.121-143
• /
• 2010

To investigate the effect of implant types and bone resorption on the fracture characteristics. 4 types of Osstem$^{(R)}$Implant were chosen and classified into external parallel, internal parallel, external taper, internal taper groups. Finite elements analysis was conducted with ANSYS Multi Physics software. Fatigue fracture test was performed by connecting the mold to the dynamic load fatigue testing machine with maximum load of 600N and minimum load of 60N. The entire fatigue test was performed with frequency of 14Hz and fractured specimens were observed with Hitachi S-3000 H scanning electron microscope. The results were as follows: 1. In the fatigue test of 2 mm exposed implants group, Tapered type and external connected type had higher fatigue life. 2. In the fatigue test of 4 mm exposed implants group, Parallel type and external connected types had higher fatigue life. 3. The fracture patterns of all 4 mm exposed implant system appeared transversely near the dead space of the fixture. With a exposing level of 2 mm, all internally connected implant systems were fractured transversely at the platform of fixture facing the abutment. but externally connected ones were fractured at the fillet of abutment body and hexa of fixture or near the dead space of the fixture. 4. Many fatigue striations were observed near the crack initiation and propagation sites. The cleavage with facet or dimple fractures appeared at the final fracture sites. 5. Effective stress of buccal site with compressive stress is higher than that of lingual site with tensile stress, and effective stress acting on the fixture is higher than that of the abutment screw. Also, maximum effective stress acting on the parallel type fixtures is higher. It is careful to use the internal type implant system in posterior area.

• Journal of Intelligence and Information Systems
• /
• v.22 no.2
• /
• pp.127-142
• /
• 2016

Deep learning model is a kind of neural networks that allows multiple hidden layers. There are various deep learning architectures such as convolutional neural networks, deep belief networks and recurrent neural networks. Those have been applied to fields like computer vision, automatic speech recognition, natural language processing, audio recognition and bioinformatics where they have been shown to produce state-of-the-art results on various tasks. Among those architectures, convolutional neural networks and recurrent neural networks are classified as the supervised learning model. And in recent years, those supervised learning models have gained more popularity than unsupervised learning models such as deep belief networks, because supervised learning models have shown fashionable applications in such fields mentioned above. Deep learning models can be trained with backpropagation algorithm. Backpropagation is an abbreviation for "backward propagation of errors" and a common method of training artificial neural networks used in conjunction with an optimization method such as gradient descent. The method calculates the gradient of an error function with respect to all the weights in the network. The gradient is fed to the optimization method which in turn uses it to update the weights, in an attempt to minimize the error function. Convolutional neural networks use a special architecture which is particularly well-adapted to classify images. Using this architecture makes convolutional networks fast to train. This, in turn, helps us train deep, muti-layer networks, which are very good at classifying images. These days, deep convolutional networks are used in most neural networks for image recognition. Convolutional neural networks use three basic ideas: local receptive fields, shared weights, and pooling. By local receptive fields, we mean that each neuron in the first(or any) hidden layer will be connected to a small region of the input(or previous layer's) neurons. Shared weights mean that we're going to use the same weights and bias for each of the local receptive field. This means that all the neurons in the hidden layer detect exactly the same feature, just at different locations in the input image. In addition to the convolutional layers just described, convolutional neural networks also contain pooling layers. Pooling layers are usually used immediately after convolutional layers. What the pooling layers do is to simplify the information in the output from the convolutional layer. Recent convolutional network architectures have 10 to 20 hidden layers and billions of connections between units. Training deep learning networks has taken weeks several years ago, but thanks to progress in GPU and algorithm enhancement, training time has reduced to several hours. Neural networks with time-varying behavior are known as recurrent neural networks or RNNs. A recurrent neural network is a class of artificial neural network where connections between units form a directed cycle. This creates an internal state of the network which allows it to exhibit dynamic temporal behavior. Unlike feedforward neural networks, RNNs can use their internal memory to process arbitrary sequences of inputs. Early RNN models turned out to be very difficult to train, harder even than deep feedforward networks. The reason is the unstable gradient problem such as vanishing gradient and exploding gradient. The gradient can get smaller and smaller as it is propagated back through layers. This makes learning in early layers extremely slow. The problem actually gets worse in RNNs, since gradients aren't just propagated backward through layers, they're propagated backward through time. If the network runs for a long time, that can make the gradient extremely unstable and hard to learn from. It has been possible to incorporate an idea known as long short-term memory units (LSTMs) into RNNs. LSTMs make it much easier to get good results when training RNNs, and many recent papers make use of LSTMs or related ideas.