• 한국공간구조학회논문집
• /
• 제22권2호
• /
• pp.39-46
• /
• 2022

Recently, machine learning is widely used to solve optimization problems in various engineering fields. In this study, machine learning is applied to development of a control algorithm for a smart control device for reduction of seismic responses. For this purpose, Deep Q-network (DQN) out of reinforcement learning algorithms was employed to develop control algorithm. A single degree of freedom (SDOF) structure with a smart tuned mass damper (TMD) was used as an example structure. A smart TMD system was composed of MR (magnetorheological) damper instead of passive damper. Reward design of reinforcement learning mainly affects the control performance of the smart TMD. Various hyper-parameters were investigated to optimize the control performance of DQN-based control algorithm. Usually, decrease of the time step for numerical simulation is desirable to increase the accuracy of simulation results. However, the numerical simulation results presented that decrease of the time step for reward calculation might decrease the control performance of DQN-based control algorithm. Therefore, a proper time step for reward calculation should be selected in a DQN training process.

• Structural Engineering and Mechanics
• /
• 제41권5호
• /
• pp.675-689
• /
• 2012

Most connections of steel structures exhibit flexible behaviour under cyclic loading. The flexible connections can be assumed as nonlinear rotational springs attached to the ends of each beam. The nonlinear behaviour of the connections can be considered by suitable moment-rotation relationship. Time-history analysis by direct integration method can be used as a powerful technique to determine the nonlinear dynamic response of the structure. In conventional numerical integration, the response is evaluated for a series of short time increments. The limitations on the size of time intervals can be removed by using Chen and Robinson improved time history analysis method, in which the integrated displacements are used as the new variables in integrated equation of motion. The proposed method permits longer time intervals and reduces the computational works. In this paper the nonlinearity behaviour of the structure is summarized on the connections, and the step by step improved time-history analysis is used to calculate the dynamic response of the structure. Several numerical calculations which indicate the applicability and advantages of the proposed methodology are presented. These calculations illustrate the importance of the effect of the nonlinear behaviour of the flexible connections in the calculation of the dynamic response of steel frames.

• 산업경영시스템학회지
• /
• 제47권2호
• /
• pp.48-56
• /
• 2024

We examine a single machine scheduling problem with step-improving jobs in which job processing times decrease step-wisely over time according to their starting times. The objective is to minimize total completion time which is defined as the sum of completion times of jobs. The total completion time is frequently considered as an objective because it is highly related to the total time spent by jobs in the system as well as work-in-progress. Many applications of this problem can be observed in the real world such as data gathering networks, system upgrades or technological shock, and production lines operated with part-time workers in each shift. Our goal is to develop a scheduling algorithm that can provide an optimal solution. For this, we present an efficient branch and bound algorithm with an assignment-based node design and tight lower bounds that can prune branch and bound nodes at early stages and accordingly reduce the computation time. In numerical experiments well designed to consider various scenarios, it is shown that the proposed algorithm outperforms the existing method and can solve practical problems within reasonable computation time.

• Korea-Australia Rheology Journal
• /
• 제12권1호
• /
• pp.83-92
• /
• 2000

In resin transfer molding (RTM) process, preplaced fiber mat is set up in a mold and thermoset resin is injected into the mold. An important interest in RTM process is to minimize cycle time without sacrificing part quality or increasing cost. In this study, the numerical simulation and optimization process in filling stage were conducted in order to determine the optimum gate locations. Control volume finite element method (CVFEM) was used in this numerical analysis with the coordinate transformation method to analyze the complex 3-dimensional structure. Experiments were performed to monitor the flow front to validate simulation results. The results of numerical simulation predicted well the experimental results with every single, simultaneous and sequential injection procedure. We performed the optimization analysis for the sequential injection procedure to minimize fill time. The complex geometry of an automobile bumper core was chosen. Genetic algorithm was used in order to determine the optimum gate locations with regard to 3-step sequential injection case. These results could provide the information of the optimum gate locations in each injection step and could predict fill time and flow front.

• 제어로봇시스템학회:학술대회논문집
• /
• 제어로봇시스템학회 1992년도 한국자동제어학술회의논문집(국제학술편); KOEX, Seoul; 19-21 Oct. 1992
• /
• pp.370-375
• /
• 1992

An algorithm for the notion planning of the robotic hand is proposed to generate finite displacements and changes in orientation of objects by considering sliding effects between the fingertips and the object at contact points. Specifically, an optimization problem is firstly solved to find minimum contact forces and minimum joint velocities to impart a desired motion to the object at each time step. Then the instantaneous relative velocity at the contact point is found by determining velocities of the fingertip and the velocity of the object at the contact point. Finally time derivatives of the surface variables and contact angle of the fingertip and the object at the present time step is computed using the Montana's contact equation to find the contact parameters of the fingertip and the object at the next time step. To show the validity of the proposed algorithm, a numerical example is illustrated by employing the robotic hand manipulating a sphere with three fingers each of which has four joints.

• 제어로봇시스템학회:학술대회논문집
• /
• 제어로봇시스템학회 1996년도 한국자동제어학술회의논문집(국내학술편); 포항공과대학교, 포항; 24-26 Oct. 1996
• /
• pp.85-88
• /
• 1996

We develop an efficient technique of controlling chaos using M-step ahead prediction with the OGY method. It has smaller transient time than the OGY method, and prevents burst phenomena that occur in noisy environment. This technique is very simple and needs small memory compared with targeting algorithms. Numerical examples show that the proposed algorithm has good performance, especially in noisy environment.

• Journal of Ship and Ocean Technology
• /
• 제8권3호
• /
• pp.8-17
• /
• 2004

A numerical method is developed for computing the free surface flows around a transom stern of a ship at a high Froude number. At high speed, the flow may be detached from the flat transom stern. In the limit of the high Froude number, the problem becomes a planning problem. In the present study, we make the finite-element computations for a transom stern flows around a wedge-shaped floating ship. The numerical method is based on the Hamilton's principle. The problem is formulated as an initial value problem with nonlinear free surface conditions. In the numerical procedures, the domain was discretized into a set of finite elements and the numerical quadrature was used for the functional equation. The time integrations of the nonlinear free surface condition are made iteratively at each time step. A set of large algebraic equations is solved by GMRES(Generalized Minimal RESidual, Saad and Schultz 1986) method which is proven very efficient. The computed results are compared with previous numerical results obtained by others.

• Journal of applied mathematics & informatics
• /
• 제40권3_4호
• /
• pp.603-618
• /
• 2022

In this paper, a numerical solution of the generalized diffusion equation with a delay has been obtained by a numerical technique based on the Galerkin finite element method by applying the cubic B-spline basis functions. The time discretization process is carried out using the forward Euler method. The numerical scheme is required to preserve the delay-independent asymptotic stability with an additional restriction on time and spatial step sizes. Both the theoretical and computational rates of convergence of the numerical method have been examined and found to be in agreement. As it can be observed from the numerical results given in tables and graphs, the proposed method approximates the exact solution very well. The accuracy of the numerical scheme is confirmed by computing L₂ and L_∞ error norms.

• 한국화재소방학회논문지
• /
• 제30권1호
• /
• pp.31-36
• /
• 2016

본 연구는 에폭시 계열 내화 도료의 열분해 특성을 파악하기 위해 열중량분석과 수치해석을 동시에 수행하였다. 에폭시 계열 내화 도료의 질량 감소 양상을 수치적으로 구현하기 위한 계산 모델이 소개되었으며, 계산 모델은 내화 도료의 여러 열분해과정을 단순화하여 4 단계의 순차적인 열분해 형태로 만들었다. 반응 속도는 Arrhenious 형태로 모델 되었고, 열중량분석을 통해 획득된 열분해 현상을 수치적으로 모사하기 위하여 화학 반응 매개변수들이 최적화되었다. 실험 결과 두 단계(two-step)와 세 단계(three-step)의 급격한 질량 감소가 질소와 공기 분위기에서 각각 나타났다. 또한 도료가 공기 분위기에 노출되었을 때 산소의 참여로 발생하는 복합적인 화학 반응들이 내화도료의 안정화를 도와 $200{\sim}500^{\circ}C$ 범위에서 질량 감소율을 낮추었다. 수치해석 결과는 실험과 비교하여 전체적으로 잘 일치하였으며 수치 모델에 포함되지 않은 $800^{\circ}C$ 이후를 제외하면 3% 미만의 오차를 보였다.

• Structural Engineering and Mechanics
• /
• 제3권2호
• /
• pp.185-192
• /
• 1995

A numerical algorithm for plane stress and shell elasto-plasticity is presented in this paper. The proposed strain decomposition (SD) algorithm is an elastic predictor/plastic corrector algorithm, and in the context of operator splitting, is a return mapping algorithm. However, it differs significantly from other return mapping algorithms in that only the necessary response functions are used without invoking their gradients, and the stress increment is updated only at the end of the time step. This makes the proposed SD algorithm more suitable for materials with complex yield surfaces and will guard against error accumulation during the time step. Comparative analyses of structural systems using the proposed strain decomposition (SD) algorithm and the iterative radial return (IRR) algorithm are presented. The results demonstrate the accuracy and usefulness of the proposed algorithm.