• Korean Journal of Agricultural and Forest Meteorology
• /
• v.19 no.3
• /
• pp.153-163
• /
• 2017

The objective of this study is to enhance the model's speed of estimating weather variables (e.g., minimum/maximum temperature, sunshine hour, PRISM (Parameter-elevation Regression on Independent Slopes Model) based precipitation), which are applied to the Agrometeorological Early Warning System (http://www.agmet.kr). The current process of weather estimation is operated on high-performance multi-core CPUs that have 8 physical cores and 16 logical threads. Nonetheless, the server is not even dedicated to the handling of a single county, indicating that very high overhead is involved in calculating the 10 counties of the Seomjin River Basin. In order to reduce such overhead, several cache and parallelization techniques were used to measure the performance and to check the applicability. Results are as follows: (1) for simple calculations such as Growing Degree Days accumulation, the time required for Input and Output (I/O) is significantly greater than that for calculation, suggesting the need of a technique which reduces disk I/O bottlenecks; (2) when there are many I/O, it is advantageous to distribute them on several servers. However, each server must have a cache for input data so that it does not compete for the same resource; and (3) GPU-based parallel processing method is most suitable for models such as PRISM with large computation loads.

• 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.

• Journal of Korean Tunnelling and Underground Space Association
• /
• v.24 no.5
• /
• pp.431-449
• /
• 2022

As many tunnels generally have been constructed, various experiences and techniques have been accumulated for tunnel design as well as tunnel construction. Hence, there are not a few cases that, for some usual tunnel design works, it is sufficient to perform the design by only modifying or supplementing previous similar design cases unless a tunnel has a unique structure or in geological conditions. In particular, for a tunnel blast design, it is reasonable to refer to previous similar design cases because the blast design in the stage of design is a preliminary design, considering that it is general to perform additional blast design through test blasts prior to the start of tunnel excavation. Meanwhile, entering the industry 4.0 era, artificial intelligence (AI) of which availability is surging across whole industry sector is broadly utilized to tunnel and blasting. For a drill and blast tunnel, AI is mainly applied for the estimation of blast vibration and rock mass classification, etc. however, there are few cases where it is applied to blast pattern design. Thus, this study attempts to automate tunnel blast design by means of machine learning, a branch of artificial intelligence. For this, the data related to a blast design was collected from 25 tunnel design reports for learning as well as 2 additional reports for the test, and from which 4 design parameters, i.e., rock mass class, road type and cross sectional area of upper section as well as bench section as input data as well as16 design elements, i.e., blast cut type, specific charge, the number of drill holes, and spacing and burden for each blast hole group, etc. as output. Based on this design data, three machine learning models, i.e., XGBoost, ANN, SVM, were tested and XGBoost was chosen as the best model and the results show a generally similar trend to an actual design when assumed design parameters were input. It is not enough yet to perform the whole blast design using the results from this study, however, it is planned that additional studies will be carried out to make it possible to put it to practical use after collecting more sufficient blast design data and supplementing detailed machine learning processes.

• Journal of Soil and Groundwater Environment
• /
• v.13 no.1
• /
• pp.77-91
• /
• 2008

In this study, we developed a technique of applying DRASTIC, which is the most widely used tool for estimation of groundwater vulnerability to the aqueous phase contaminant infiltrated from the surface, and a groundwater flow model jointly to assess groundwater contamination potential. The developed technique is then applied to Buyeo-eup area in Buyeo-gun, Chungcheongnam-do, Korea. The input thematic data of a depth to water required in DRASTIC model is known to be the most sensitive to the output while only a few observations at a few time schedules are generally available. To overcome this practical shortcoming, both steady-state and transient groundwater level distributions are simulated using a finite difference numerical model, MODFLOW. In the application for the assessment of groundwater vulnerability, it is found that the vulnerability results from the numerical simulation of a groundwater level is much more practical compared to cokriging methods. Those advantages are, first, the results from the simulation enable a practitioner to see the temporally comprehensive vulnerabilities. The second merit of the technique is that the method considers wide variety of engaging data such as field-observed hydrogeologic parameters as well as geographic relief. The depth to water generated through geostatistical methods in the conventional method is unable to incorporate temporally variable data, that is, the seasonal variation of a recharge rate. As a result, we found that the vulnerability out of both the geostatistical method and the steady-state groundwater flow simulation are in similar patterns. By applying the transient simulation results to DRASTIC model, we also found that the vulnerability shows sharp seasonal variation due to the change of groundwater recharge. The change of the vulnerability is found to be most peculiar during summer with the highest recharge rate and winter with the lowest. Our research indicates that numerical modeling can be a useful tool for temporal as well as spatial interpolation of the depth to water when the number of the observed data is inadequate for the vulnerability assessments through the conventional techniques.

• Journal of Korea Port Economic Association
• /
• v.32 no.1
• /
• pp.29-58
• /
• 2016

The purpose of this paper is to show the clustering trend and the comparison of empirical results and is to choose the clustering ports for 3 Korean ports(Busan, Incheon and Gwangyang Ports) by using the bootstrapped DEA(Data Envelopment Analysis) and game Cross-efficiency models for 38 Asian ports during the period 2003-2013 with 4 input variables(birth length, depth, total area, and number of cranes) and 1 output variable(container TEU). The main empirical results of this paper are as follows. First, bootstrapped DEA efficiency of SW and LT is 0.7660, 0.7341 respectively. Clustering results of the bootstrapped DEA analysis show that 3 Korean ports [ Busan (6.46%), Incheon (3.92%), and Gwangyang (2.78%)] can increase the efficiency in the SW model, but the LT model shows clustering values of -1.86%, -0.124%, and 2.11% for Busan, Gwangyang, and Incheon respectively. Second, the game cross-efficiency model suggests that Korean ports should be clustered with Hong Kong, Shanghi, Guangzhou, Ningbo, Port Klang, Singapore, Kaosiung, Keelong, and Bangkok ports. This clustering enhances the efficiency of Gwangyang by 0.131%, and decreases that of Busan by-1.08%, and that of Incheon by -0.009%. Third, the efficiency ranking comparison between the two models using the Wilcoxon Signed-rank Test was matched with the average level of SW (72.83 %) and LT (68.91%). The policy implication of this paper is that Korean port policy planners should introduce the bootstrapped DEA, and game cross-efficiency models when clustering is needed among Asian ports for enhancing the efficiency of inputs and outputs. Also, the results of SWOT(Strength, Weakness, Opportunity, and Threat) analysis among the clustering ports should be considered.

• Transactions of the Korean Society of Mechanical Engineers A
• /
• v.34 no.5
• /
• pp.619-628
• /
• 2010

Shape Memory Alloy (SMA) has recently received attention in developing implantable surgical equipments and it is expected to lead the future medical device market by adequately imitating surgeons' flexible and delicate hand movement. However, SMA actuators have not been used widely because of their nonlinear behavior called hysteresis, which makes their control difficult. Hence, we propose a parameter, $t_1$, which is necessary for temperature control, by analyzing the open-loop step response between current and temperature and by comparing it with the values of linear differential equations. $t_1$ is a pole of the transfer function in the invariant linear model in which the input and output are current and temperature, respectively; hence, $t_1$ is found to be related to the state variable used for temperature control. When considering the parameter under heat treatment conditions, $T_{max}$ was found to assume the lowest value, and $t_1$ was irrelevant to the heat treatment.

• The Journal of the Korea Contents Association
• /
• v.16 no.8
• /
• pp.651-660
• /
• 2016

This study analyzed the contribution of the national economy, China's textile industry by 2010 I-O Table issued by the Chinese Bureau of Statistics 2013. The results shows that the production inducement coefficient of China's textile industry is the column total 3.6228 and in line total 3.5452, is a key industry that leads the industry in China. Second, the index of the power of dispersion of the Chinese textile industry is 1.1982, index of the sensitivity of dispersion is 1.1725. Third, income inducement coefficient of China's textile industry 0.5228, tax inducement coefficient 0.1522, a value-added inducement coefficient 1. Especially China's textile industry induce 2993.6 trillion yuan(textile industry of 8.6 trillion yuan, up 3.0%) in the national production, value-added inducement 97.1 trillion yuan (textile industry 1.7 trillion yuan, up 2.0%), income inducement 42.8 trillion yuan (textile industry 0.9 one trillion yuan, 2.0%), also tax inducement 15.4 trillion yuan (textile industry 0.3 one trillion yuan, 2.0%).

• Journal of the Institute of Electronics Engineers of Korea SP
• /
• v.44 no.2 s.314
• /
• pp.46-54
• /
• 2007

In current printer industry, four or more colorants are added for color gamut extension because the gamut of printer is smaller than other devices. However, these additional colorants make a redundancy problem that several combinations of colorants reproduced same color stimulus in colorimetric inverse characterization process. Thus, we propose a method of colorimetric inverse characterization using color correlation between colorant's amount. First, for analyzing the combination of colorants which represent the same color stimulus, we estimate the color stimulus for all combination of colorants by Cellular Yule-Nielsen Spectral Neugebauer printer model. The combination of colorants which has higher color correlation factor comparing combinations of colorant around itself in color space is selected. It can reduced the color difference from the tetrahedral interpolation process which is estimation of the output value(colorants combination) for arbitrary input(color stimulus). The selected combinations of colorants and their color stimulus are stored to the lookup table. In experiment, the CMYKGO printer was used. As a result, the dark region of color gamut was extended and the color tone was more naturally represented.

• Journal of Korean Society of Forest Science
• /
• v.102 no.4
• /
• pp.506-514
• /
• 2013

This study examined the feasibility on the construction of a wood industrialization service center for a wood industry cluster establishment in Jeollanam-do. Construction of the wood industrialization service center is based on a discount rate of 3.5%, an investment period of 4 years, a business operations period of 16 years and an investment cost of 24600 million won; the total amount of the net present value, the cost-benefit ratio and the internal rate of return were assumed to be 2.579 million won, 2.51%, and 10.1%, respectively. In addition, the production inducement coefficient, the induced production effect, the income-induced coefficient, the income inducement effect, the employment inducement coefficient, and the employment inducement effect were estimated 1.4345, 35287 million won, 0.1655, 4000.7 million won, and 0.4665, 1,145 people, in the effects of the wood related industries using the multi-regional input-output model, respectively. Financial independence of operating income to cover its own costs incurred in accordance with the operating project might be practicable.

• Journal of Energy Engineering
• /
• v.24 no.3
• /
• pp.164-172
• /
• 2015

As the petroleum products produced from the Oil refinery industry (ORI), a national key industry in Korea, are supplied to other industries as an intermediate goods, the supply shortage of ORI has a large impact on the national economy. This paper attempts to analyze the supply shortage effects which are defined as the negative impact of one won of supply failure in the ORI on the production of other industries. To this end, an inter-industry analysis using an input-output (I-O) table describing inter-industry flow of intermediate goods is applied. More concretely, the supply-driven model is employed over the period 1990-2012. In addition, the results are compared with those for shipbuilding, semiconductor, and steel industries. The results show that the supply shortage effects are computed to be 0.9205 won when using 2012 I-O table. More specifically, the supply shortage effects on chemical products and transportation industries are computed to be 0.2113 and 0.1140, which are relatively large, The supply shortage effect of ORI is smaller than that of steel industry (1.4131 won), but larger than that of shipbuilding industry (0.0586 won) and that of semiconductor industry (0.1111 won).