• 제어로봇시스템학회:학술대회논문집
• /
• 2005.06a
• /
• pp.920-924
• /
• 2005

This paper presents adaptive learning data of evolvable neural networks (ENNs) for time series prediction of nonlinear dynamic systems. ENNs are a special class of neural networks that adopt the concept of biological evolution as a mechanism of adaptation or learning. ENNs can adapt to an environment as well as changes in the environment. ENNs used in this paper are L-system and DNA coding based ENNs. The ENNs adopt the evolution of simultaneous network architecture and weights using indirect encoding. In general just previous data are used for training the predictor that predicts future data. However the characteristics of data and appropriate size of learning data are usually unknown. Therefore we propose adaptive change of learning data size to predict the future data effectively. In order to verify the effectiveness of our scheme, we apply it to chaotic time series predictions of Mackey-Glass data.

• Proceedings of the Korean Institute of Intelligent Systems Conference
• /
• 1997.10a
• /
• pp.115-120
• /
• 1997

System identification is the task of inferring a mathematical description of a dynamic system from a series of measurements of the system. There are several motives for establishing mathematical descriptions of dynamic systems. Typical applications encompass simulation, prediction, fault diagnostics, and control system design. The paper demonstrates that neural networks can be used effective for the identification of nonlinear dynamical systems. The content of this paper concerns dynamic neural network models, where not all inputs to and outputs from the networks are measurable. Only one model type is treated, the well-known Innovation State Space model(Kalman Predictor). The identification is based only on input/output measurements, so in fact a non-linear Extended Kalman Filter problem is solved. Even for linear models this is a non-linear problem without any assurance of convergence, and in spite of this fact an attempt is made to apply the principles from linear models, an extend them to non-linear models. Computer simulation results reveal that the identification scheme suggested are practically feasible.

• Korean Journal of Air-Conditioning and Refrigeration Engineering
• /
• v.8 no.1
• /
• pp.65-75
• /
• 1996

A pilot plant, which is simplified the hot water heating control system of a large scale residential building, is used to investigate the effects of control methods and operating conditions on the system performance and to compare control characteristics. Digital variable structure controller(DVSC) and digital PI controller are implemented to control the speed of the circulating pump for the pilot plant using PC. For the DVSC, a control algorithm is suggested, which using a nonlinear sliding surface and a PID sliding surface outside and inside of output error boundary layer, respectively. Smith predictor algorithm is used for the compensation of long dead time. The suggested DVSC yields improved control performance compared with existing DVSC using linear sliding surface only. the system responses with the suggested DVSC shows good responses without overshoot for various operating conditions and robust under external disturbances compared with digital PI controller.

• Journal of the Korean Institute of Telematics and Electronics B
• /
• v.32B no.11
• /
• pp.1489-1495
• /
• 1995

In this paper, we propose recurrent neural prediction models (RNPM), recurrent neural networks trained as a nonlinear predictor of speech, as a new connectionist model for speech recognition. RNPM modulates its mapping effectively by internal representation, and it requires no time alignment algorithm. Therefore, computational load at the recognition stage is reduced substantially compared with the well known predictive neural networks (PNN), and the size of the required memory is much smaller. And, RNPM does not suffer from the problem of deciding the time varying target function. In the speaker dependent and independent speech recognition experiments under the various conditions, the proposed model was comparable in recognition performance to the PNN, while retaining the above merits that PNN doesn't have.

• Journal of Korean Society of Water and Wastewater
• /
• v.13 no.3
• /
• pp.42-55
• /
• 1999

A mathematical model for biological nutrient removal in a sequencing batch reactor process, which is based on the IAWQ Activated Sludge Model No. 2 with a few modifications, has been developed. Twenty water quality components and twenty three kinetic equations are incorporated in the model. The model is structured in the matrix form based on the law of mass conservation using stoichiometry and kinetic equations. Stoichiometric coefficients and kinetic parameters included in the model equations are chosen from the literature. A multistep predictor-corrector algorithm of variable step-size is adopted for solving the vector nonlinear ordinary differential equations. The simulation for experimental results is conducted to evaluate the validity of the model and to calibrate coefficients and parameters. The simulation using the model well represents the experimental results from laboratory. The mathematical model developed in this study may be utilized for the design and operation of a sequencing batch reactor process under the steady and unsteady-state at various environmental conditions.

• Journal of the Society of Naval Architects of Korea
• /
• v.33 no.2
• /
• pp.1-12
• /
• 1996

Hydrodynamic characteristics of shipping water on deck are investigated by using a simplified two-dimensional model. Formulation of the shipping water on deck leads to a nonlinear hyperbolic system of equations based on the shallow-water wave theory. Time-domain solution of these equations are obtained numerically using a finite difference method which adopts predictor-corrector method for time-marching and 2nd upwind differencing method for convection term calculation. To confirm the validity of the present numerical method, we calculated some shallow-water wave problems accompanying a bore and compared the obtained results with the analytic solutions. We found good agreements between them. Though the calculation results of shipping water on deck, we show that the shipping water flows into the deck as a rarefying wave arid grows into a bore after colliding with a deck structure. Also we examined the effects of acceleration and slope of deck and found that they have significant influences on the flow of shipping water.

• Geomechanics and Engineering
• /
• v.12 no.3
• /
• pp.441-464
• /
• 2017

The determination of the mixture parameters of stabilization has become a great concern in geotechnical applications. This paper presents an effort about the application of artificial intelligence (AI) techniques including radial basis neural network (RBNN), multi-layer perceptrons (MLP), generalized regression neural network (GRNN) and adaptive neuro-fuzzy inference system (ANFIS) in order to predict the unconfined compressive strength (UCS) of silty soil stabilized with bottom ash (BA), jute fiber (JF) and steel fiber (SF) under different freeze-thaw cycles (FTC). The dosages of the stabilizers and number of freeze-thaw cycles were employed as input (predictor) variables and the UCS values as output variable. For understanding the dominant parameter of the predictor variables on the UCS of stabilized soil, a sensitivity analysis has also been performed. The performance measures of root mean square error (RMSE), mean absolute error (MAE) and determination coefficient ($R^2$) were used for the evaluations of the prediction accuracy and applicability of the employed models. The results indicate that the predictions due to all AI techniques employed are significantly correlated with the measured UCS ($p{\leq}0.05$). They also perform better predictions than nonlinear regression (NLR) in terms of the performance measures. It is found from the model performances that RBNN approach within AI techniques yields the highest satisfactory results (RMSE = 55.4 kPa, MAE = 45.1 kPa, and $R^2=0.988$). The sensitivity analysis demonstrates that the JF inclusion within the input predictors is the most effective parameter on the UCS responses, followed by FTC.

• Wind and Structures
• /
• v.26 no.1
• /
• pp.25-34
• /
• 2018

This paper presents a beam finite element model of a vibrate wind blade in large elastic deformation subjected to the aerodynamic, centrifugal, gyroscopic and gravity loads. The gyroscopic loads applied to the blade are induced by her simultaneous vibration and rotation. The proposed beam finite element model is based on a simplex interpolation method and it is mainly intended to the numerical analysis of wind blades vibration in large elastic deformation. For this purpose, the theory of the sheared beams and the finite element method are combined to develop the algebraic equations system governing the three-dimensional motion of blade vibration. The applicability of the theoretical approach is elucidated through an original case study. Also, the static deformation of the used wind blade is assessed by appropriate software using a solid finite element model in order to show the effectiveness of the obtained results. To simulate the nonlinear dynamic response of wind blade, the predictor-corrector Newmark scheme is applied and the stability of numerical process is approved during a large time of blade functioning. Finally, the influence of the modified geometrical stiffness on the amplitudes and frequencies of the wind blade vibration induced by the sinusoidal excitation of gravity is analyzed.

• Structural Engineering and Mechanics
• /
• v.67 no.6
• /
• pp.629-641
• /
• 2018

The paper describes the development of a two-dimensional (2D) co-rotational nonlinear beam finite element that includes advanced path-following capabilities for detecting bifurcation instability in elasto-plasticity of steel elements subjected to fire without introducing imperfections. The advantage is twofold: i) no need to assume the magnitude of the imperfections and consequent reduction of the model complexity; ii) the presence of possible critical points is checked at each converged time step based on the actual load and stiffness distribution in the structure that is affected by the temperature field in the elements. In this way, the buckling modes at elevated temperature, that may be different from the ones at ambient temperature, can be properly taken into account. Moreover, an improved displacement predictor for estimating the displacement field allowed significant reduction of the computational cost. A co-rotational framework was exploited for describing the beam kinematic. In order to highlight the potential practical implications of the developed finite element, a parametric analysis was performed to investigate how the beam element compares both with the EN1993-1-2 buckling curve and with experimental tests on axially compressed steel members. Validation against experimental data and numerical outcomes obtained with commercial software is thoroughly described.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.27 no.6A
• /
• pp.562-567
• /
• 2002

Noise predictive maximum likelihood(NPML) detector embeds noise predictions/ whitening process in branch metric calculation of Viterbi detector and improves the reliability of branch metric computation. Therefore, PRML detector with a noise predictor achieves some performance improvement and has an advantage of low complexity. This paper shows that NP(1221)ML system through noise predictive PR-equalized signal has less complexity and better performance than high order PR(12321)ML system in high density perpendicular magnetic recording. The simulation results are evaluated using (1) random sequence and (2) run length limited (1,7) sequence, and they are applied to linear channel and nonlinear channel with normalized linear density $1.0{\leq}K_p{\leq}3.0$.