Joo-Young Park

Improvement of Fatigue Model of Concrete Pavement Slabs Using Environmental Loading

Concrete slab curls and warps due to the uneven distribution of temperature and moisture and as the result, internal stress develops within the slab. Therefore, environmental loads must be considered in addition to the traffic loads to predict the lifespan of the concrete pavement more accurately. The strength of the concrete slab is gradually decreases to a certain level at which fatigue cracking is generated by the repetitive traffic and environmental loadings. In this study, a new fatigue regression model was developed based on the results from previously performed studies. To verify the model, another laboratory flexural fatigue test program which was not used in the model development, was conducted and compared with the predictions of other existing models. Each fatigue model was applied to analysis logic of cumulative fatigue damage of concrete pavement developed in the study. The sensitivity of cumulative fatigue damage calculated by each model was analyzed for the design factors such as slab thickness, joint spacing, complex modulus of subgrade reaction and the load transfer at joints. As the result, the model developed in this study could reflect environmental loading more reasonably by improving other existing models which consider R, minimum/maximum stress ratio.

Optimal design for concrete pavement situated above box culvert: experimental and numerical study

Abstract This study examined the pavement distress data obtained from the field as well as the effects of the soil cover depth, joint position and reinforced slab length on the behaviour of concrete pavement using the finite element method. An ultimate strength model with a similarity of 1/2 was designed and constructed. Finite element analysis incorporating an implicit direct numerical integration scheme was conducted to obtain an optimal design for the pavement structure above a box culvert. An accelerated pavement test was conducted for validation, in which a realistic traffic load was imposed on the pavement model constructed in the laboratory. The strain histories were measured at each section and compared with those of numerical analysis. The optimal joint position and reinforced slab length that can minimise the damage to the pavement structure above a box culvert were also determined.

Development of Mechanistic-empirical Joint Spacing Design Method for Concrete Pavements

Tensile stress occurs and random crack develops in concrete pavement slab when it contracts by variation of temperature and humidity. The tensile stress decreases and the random crack is minimized by sawcutting the slab and inducing the crack with regular spacing. The random crack, joint damage, decrease of load transfer efficiency are caused by too wide joint spacing while too narrow joint spacing leads to increase of construction cost and decrease of comfort. A mechanistic-empirical joint spacing design method for the concrete pavement was developed in this study. Structurally and environmentally weakest sections were found among the sections showing good performance, and design strengths were determined by finite element analysis on the sections. The joint width for which the load transfer efficiency is suddenly lowered was determined as allowable joint with referring to existing research results. The maximum joint spacing for which the maximum tensile stress calculated by the finite element analysis did not exceed the design strength were found. And the maximum joint width expected by the maximum joint spacing were compared to the allowable joint width. The new method developed in this study was applied to two zones of Hamyang-Woolsan Expressway being designed. The same joint spacing as a test section constructed by 8.0m of joint spacing wider than usual was calculated by the design method. Very low cracking measured at 6 years after opening of the test section verified the design method developed in this study.

Control of Crawling Robot using Actor-Critic Fuzzy Reinforcement Learning

Abstract Recently, reinforcement learning methods have drawn much intere sts in the area of machine learning. Dominant ap-proaches in researches for the reinforcement learning include t he value-function approach, the policy search approach, and the actor-critic approach, among which pertinent to this pa per are algorithms studied for problems with con-tinuous states and continuous actions along the line of the act or-critic strategy. In particular, this paper focuses on presenting a method combining th e so-called ACFRL(actor-critic fuzzy reinforcement learning), which is an ac-tor-critic type reinforcement learning based on fuzzy theory, t ogether with the RLS-NAC which is based on the RLS filters and natural actor-critic methods. The presented method is applied to a control problem for crawling robots, and some results are reported from comparison of learning performan ce. Key Words : Fuzzy Model, Actor-Critic Method, RLS-NAC, Crawling Robot 1. 서 론 최근에 이론적 가치 및 실용적 가능성 측면에서 크게 주목받고 있는 기계학습(machine learning)의 한 분야인 강화학습(reinforcement learning)은 인간 또는 동물이 경험을 통해 지식을 습득해 나아가는 과정을 모사하여 사용자가 원하는 목표를 달성하는 방법론이다. 강화학습은 시행착오를 통해 문제의 해결책을 찾아가기 때문에 수식적인 모델링으로 해(solution)를 얻기 힘든 복잡한 비선형 시스템(nonlinear system)을 제어 하는데 장점이 있다. 그리고 제어 분야[1-2] 뿐만 아니라 게임[3], 금융[4] 등 여러 분야에 성공적으로 적용되었으며, 점차적으로 범용성을 넓혀가고 있다.강화학습은 이산(discrete)인 저차원 공간을 가지는 문제에서는 쉽게 적용되지만, 하드웨어가 결합된 연속적(continuous)인 고차원 시스템에서는 계산속도 문제와 시스템이 가지는 공간을 정확하게 표현할 수 없는 문제 등으로 인해 학습이 성공적으로 이루어지지 않는 경우가 종종 발생한다[5].그래서 이와 같은 시스템이 가지는 공간을 표현하는 공간의 일반화(generalization)에 관한 문제에 대해서는 이전부터 광범위하게 연구가 진행되어 오고 있다. 이에 대한 예로 타일 코딩(tile-coding)이나 신경망(neural network) 등의 방법과 이를 서로 융합하는 다양한 방법 등을 들 수 있다[5]. 본 논문에서는 연속적인 고차원 공간에 강화학습을 적용하는 것을 목표로 하여 액터-크리틱 방법(actor-critic method), 퍼지 이론, RLS(recursive least-squares) 필터 등을 종합적으로 사용하는 방법론을 고려한다. 강화학습 관련 연구에서 가장 유력하게 사용되어 온 방법들로는 가치함수(value function)를 활용하는 기법, 정책(제어규칙, poli-cy) 탐색 기법 및 액터-크리틱 기법 등이 있는데, 본 논문