• Journal of KIISE:Software and Applications
• /
• v.35 no.1
• /
• pp.33-40
• /
• 2008

In this paper, we propose a method to enhance adaptability in a dialogue system using the reinforcement learning that reduces response errors by trials and error-search similar to a human dialogue process. The adaptive dialogue strategy means that the dialogue system improves users' satisfaction and dialogue efficiency by loaming users' dialogue styles. To apply the reinforcement learning to the dialogue system, we use a main-dialogue span and sub-dialogue spans as the mathematic application units, and evaluate system usability by using features; success or failure, completion time, and error rate in sub-dialogue and the satisfaction in main-dialogue. In addition, we classify users' groups into beginners and experts to increase users' convenience in training steps. Then, we apply reinforcement learning policies according to users' groups. In the experiments, we evaluated the performance of the proposed method on the individual reinforcement learning policy and group's reinforcement learning policy.

• Proceedings of the Korea Concrete Institute Conference
• /
• 1996.10a
• /
• pp.480-485
• /
• 1996

A numerical method for perdicting the behavior of a reinforced concrete column under biaxial loading is proposed, using the layered finite element method. Concrete is assumed to exhibit strain softening and steel reinforcement is elastic-plastic. The bending theory assumptions are used and bond slip of reinforcement is meglected. To perdict the entire load-deformation characteristics, displacement control method is used. This method consider not only combined effect due to axial load and bending moment but also that due to bending moments. Predicted behaviors of reinforced concrete columns under biaxial loading through the numerical method proposed in this study show good agreements with test results.

• Journal of Korean Tunnelling and Underground Space Association
• /
• v.25 no.2
• /
• pp.43-63
• /
• 2023

With the concentration of the population in the city center and the saturation of the structures on the ground, the development of the underground structures becomes important and the construction of an adjoining tunnel that can reduce the overall problems is respected. In addition, it is necessary to apply the reinforcement construction method for the pillar part of the adjacent tunnel that can secure stability, economy and workability of the site. In this study, the tunnel pillar reinforcement method using prestress and grouting was reviewed. There are various reinforcement methods that can compensate for the problems of the side tunnel, but as the tunnel pillar construction method using prestress and grouting is judged to be excellent in field applicability, stability and economic feasibility, theoretical and numerical analysis of the actual behavior mechanism are conducted. Numerical analysis is divided into PC stranded wire + steel pipe reinforcement grouting + prestress (Case 1), pillar part tie bolt reinforcement (Case 2), pillar part non-reinforcement (Case 3) under the same ground conditions, and the maximum value of the celling displacement, internal displacement, and member force, the stability was confirmed. Through numerical analysis, it was confirmed that Case 1 which reinforced the PC stranded wire, was the best construction method and if it is verified and supplemented through field experiments later, it will be possible to derive superior results in terms of displacement control and member force than the currently applied reinforcement method was judged.

• Journal of the Korean Geotechnical Society
• /
• v.15 no.3
• /
• pp.161-176
• /
• 1999

A powerful numerical method that can be used for modeling rock-structure interaction is the Discontinuous Deformation Analysis (D D A) method developed by Shi in 1988. In this method, rock masses are treated as systems of finite and deformable blocks. Large rock mass deformations and block movements are allowed. Although various extensions of the D D A method have been proposed in the literature, the method is not capable of modeling water-block interaction, sequential loading or unloading and rock reinforcement; three features that are needed when modeling surface or underground excavation in fractured rock. This paper presents three new extensions to the D D A method. The extensions consist of hydro-mechanical coupling between rock blocks and steady water flow in fractures, sequential loading or unloading, and rock reinforcement by rockbolts, shotcrete or concrete lining. Examples of application of the D D A method with the new extensions are presented. Simulations of the underground excavation of the \ulcornerUnju Tunnel\ulcorner in Korea were carried out to evaluate the influence of fracture flow, excavation sequence and reinforcement on the tunnel stability. The results of the present study indicate that fracture flow and improper selection of excavation sequence could have a destabilizing effect on the tunnel stability. On the other hand, reinforcement by rockbolts and shotcrete can stabilize the tunnel. It is found that, in general, the D D A program with the three new extensions can now be used as a practical tool in the design of underground structures. In particular, phases of construction (excavation, reinforcement) can now be simulated more realistically.

• Journal of IKEEE
• /
• v.23 no.4
• /
• pp.1150-1156
• /
• 2019

This paper explores a model-free value-based approach for solving survival gridworld problem. Survival gridworld problem opens up a challenge involving taking risks to gain better rewards. Classic value-based approach in model-free reinforcement learning assumes minimal risk decisions. The proposed method involves a hybrid on-policy and off-policy updates to experience roll-outs using a modified Q-based update equation that introduces a parametric linear rectifier and motivational discount. The significance of this approach is it allows model-free training of agents that take into account risk factors and motivated exploration to gain better path decisions. Experimentations suggest that the proposed method achieved better exploration and path selection resulting to higher episode scores than classic off-policy and on-policy Q-based updates.

• Transactions of the Korean Society of Machine Tool Engineers
• /
• v.16 no.1
• /
• pp.69-74
• /
• 2007

Topology optimization of the inner reinforcement for a vehicle's hood has been performed by evolutionary structural optimization(ESO) using a smoothing scheme. The purpose of this study is to obtain optimal topology of the inner reinforcement for a vehicle's hood considering the static stiffness of bending and torsion simultaneously. To do this, the multiobjective optimization technique was implemented. Optimal topologies were obtained by the ESO method. From several combinations of weighting factors, a Pareto-optimal solution was obtained. Also, a smoothing scheme was implemented to suppress the checkerboard pattern in the procedure of topology optimization. It is concluded that ESO method with a smoothing scheme is effectively applied to topology optimization of the inner reinforcement of a vehicle's hood considering the static stiffness of bending and torsion.

• Proceedings of the Korean Operations and Management Science Society Conference
• /
• 2006.05a
• /
• pp.1235-1239
• /
• 2006

Fractal manufacturing system (FrMS) distinguishes itself from other manufacturing systems by the fact that there is a fractal repeated at every scale. A fractal is a volatile organization which consists of goal-oriented agents referred to as AIR-units (autonomous and intelligent resource units). AIR-units unrestrictedly reconfigure fractals in accordance with their own goals. Their goals can be dynamically changed along with the environmental status. Since goals of AIR-units are represented as fuzzy models, an AIR-unit itself is a fuzzy logic controller. This paper presents a goal regulation mechanism in the FrMS. In particular, a reinforcement learning method is adopted as a regulating mechanism of the fuzzy goal model, which uses only weak reinforcement signal. Goal regulation is achieved by building a feedforward neural network to estimate compatibility level of current goals, which can then adaptively improve compatibility by using the gradient descent method. Goal-oriented features of AIR-units are also presented.

• Structural Monitoring and Maintenance
• /
• v.2 no.2
• /
• pp.133-143
• /
• 2015

Civil engineering infrastructure is aging and requires cost-effective maintenance strategies to enable infrastructure systems operate reliably and sustainably. This paper presents an approach for determining risk-cost balanced repair strategy of corrosion damaged reinforced concrete structures with consideration of uncertainty in structural resistance deterioration. On the basis of analytical models of cover concrete cracking evolution and bond strength degradation due to reinforcement corrosion, the effect of reinforcement corrosion on residual load carrying capacity of corroded reinforced concrete structures is investigated. A stochastic deterioration model based on gamma process is adopted to evaluate the probability of failure of structural bearing capacity over the lifetime. Optimal repair planning and maintenance strategies during the service life are determined by balancing the cost for maintenance and the risk of structural failure. The method proposed in this study is then demonstrated by numerical investigations for a concrete structure subjected to reinforcement corrosion. The obtained results show that the proposed method can provide a risk cost optimised repair schedule during the service life of corroded concrete structures.

• Journal of Semiconductor Engineering
• /
• v.2 no.1
• /
• pp.125-129
• /
• 2021

As the interest of deep learning, techniques to control the color of images in image processing field are evolving together. However, there is no clear standard for color, and it is not easy to find a way to represent only the color itself like the color-palette. In this paper, we propose a novel color palette extraction system by chroma fine-tuning with reinforcement learning. It helps to recognize the color combination to represent an input image. First, we use RGBY images to create feature maps by transferring the backbone network with well-trained model-weight which is verified at super resolution convolutional neural networks. Second, feature maps are trained to 3 fully connected layers for the color-palette generation with a generative adversarial network (GAN). Third, we use the reinforcement learning method which only changes chroma information of the GAN-output by slightly moving each Y component of YCbCr color gamut of pixel values up and down. The proposed method outperforms existing color palette extraction methods as given the accuracy of 0.9140.

• Nuclear Engineering and Technology
• /
• v.55 no.10
• /
• pp.3515-3524
• /
• 2023

A reinforcement learning framework is proposed for the control problem of outlet steam pressure of the once-through steam generator(OTSG) in this paper. The double-layer controller using Proximal Policy Optimization(PPO) algorithm is applied in the control structure of the OTSG. The PPO algorithm can train the neural networks continuously according to the process of interaction with the environment and then the trained controller can realize better control for the OTSG. Meanwhile, reinforcement learning has the characteristic of difficult application in real-world objects, this paper proposes an innovative pretraining method to solve this problem. The difficulty in the application of reinforcement learning lies in training. The optimal strategy of each step is summed up through trial and error, and the training cost is very high. In this paper, the LSTM model is adopted as the training environment for pretraining, which saves training time and improves efficiency. The experimental results show that this method can realize the self-adjustment of control parameters under various working conditions, and the control effect has the advantages of small overshoot, fast stabilization speed, and strong adaptive ability.