• Journal of Advanced Marine Engineering and Technology
• /
• v.22 no.3
• /
• pp.320-327
• /
• 1998

The purpose of this study is to design the intelligent speed control system for marine diesel engine by combining the Model Matching Method and the Nominal Model Tracking Method. Recently for the speed control of a diesel engine some methods using the advanced control techniques such as LQ control Fuzzy control or H$\infty$ control etc. have been reported. However most of speed controllers of a marine diesel engine developed are still using the PID control algorithm But the performance of a marine diesel engine depends highly on the parameter setting of the PID controllers. The authors proposed already a new method to tune efficiently the PID parameters by the Model Mathcing Method typically taking a marine diesel engine as a non-oscillatory second-order system. It was confirmed that the previously proposed method is superior to Ziegler & Nichols's method through simulations under the assumption that the parameters of a diesel engine are exactly known. But actually it is very difficult to find out the exact model of the diesel engine. Therefore when the model and the actual diesel engine are unmatched as an alternative to enhance the speed control characteristics this paper proposes a Model Refernce Adaptive Speed Control system of a diesel engine in which PID control system for the model of a diesel engine is adopted as the nominal model and a Fuzzy controller is adopted as the adaptive controller, And in the nominal model parameters of a diesel engine are adjusted using the Model Matching Method. it is confirmed that the proposed method gives better performance than the case of using only Model Matching Method through the analysis of the characteristics of indicial responses.

• Proceedings of the KSME Conference
• /
• 2008.11a
• /
• pp.678-683
• /
• 2008

Moment-based reliability analysis is the method to calculate reliability using Pearson System with first-four raw moments obtained from simulation model. But it is too expensive to calculate first four moments from complicate simulation model. To overcome this drawback the MD(multiplicative decomposition) method which approximates simulation model to kriging metamodel and calculates first four raw moments explicitly with multiplicative decomposition techniques. In general, kriging metamodel is an interpolation model that is decomposed of global model and local model. The global model, in general, can be used as the constant global model, the 1st order global model, or the 2nd order global model. In this paper, the influences of global models on the accuracy and robustness of raw moments are examined and compared. Finally, we suggest the best global model which can provide exact and robust raw moments using MD method.

• International Journal of Highway Engineering
• /
• v.18 no.4
• /
• pp.63-68
• /
• 2016

PURPOSES : The nonlinear model of fatigue cracking is typically used for determining the maintenance period. However, this requires that the model parameters be known. In this study, the particle filter (PF) method was used to determine various statistical parameters such as the mean and standard deviation values for the nonlinear model of fatigue cracking. METHODS : The PF method was used to determine various statistical parameters for the nonlinear model of fatigue cracking, such as the mean and standard deviation. RESULTS : On comparing the values obtained using the PF method and the least square (LS) method, it was found that PF method was suitable for determining the statistical parameters to be used in the nonlinear model of fatigue cracking. CONCLUSIONS : The values obtained using the PF method were as accurate as those obtained using the LS method. Furthermore, reliability design can be applied because the statistical parameters of mean and standard deviation can be obtained through the PF method.

• Speech Sciences
• /
• v.9 no.2
• /
• pp.135-146
• /
• 2002

Generally a speaker verification system improves its system recognition ratio by regularizing log likelihood ratio, using a speaker model and its background speaker model that are required to be verified. The speaker-based cohort method is one of the methods that are widely used for selecting background speaker model. Recently, Gaussian-based cohort model has been suggested as a virtually synthesized cohort model, and unlike a speaker-based model, this is the method that chooses only the probability distributions close to basic speaker's probability distribution among the several neighboring speakers' probability distributions and thereby synthesizes a new virtual speaker model. It shows more excellent results than the existing speaker-based method. This study compared the existing speaker-based background speaker models and virtual speaker models and then constructed new virtual background speaker model groups which combined them in a certain ratio. For this, this study constructed a speaker verification system that uses GMM (Gaussin Mixture Model), and found that the suggested method of selecting virtual background speaker model shows more improved performance.

• Journal of Environmental Science International
• /
• v.2 no.3
• /
• pp.193-199
• /
• 1993

This study has two objectives. One is developing the runoff model for Hoe-Dong Reservoir basin located at the upstream of Su-Young River in Pusan. To develop the runoff model, basic hydrological parameters - curve number to find effective rainfall, and storage coefficient, etc. - should be estimated. In this study, the effective rainfall was calculated by the SCS method, and the storage coefficient used in the Clark watershed routing was cited from the report of P.E.B. The other is the derivation of transfer function for Hoe-Dong Reservoir basin. The linear, discrete, input-output model which contained six parameters was selected, and the parameters were estimated by the least square method and the correlation function method, respectively. Throughout this study, rainfall and flood discharge data were based on the field observation in 1981.8.22 - 8.23 (typhoon Gladys). It was observed that the Clark watershed routing regenerated the flood hydrograph of typhoon Gladys very well, and this fact showed that the estimated hydrological parameters were relatively correct. Also, the calculated hydrograph by the linear, discrete, input-output model showed good agreement with the regenerated hydrograph at Hoe-Dong Dam site, so this model can be applicable to other small urban areas. Key Words : runoff, effective rainfall, SCS method, clark watershed iou상ng, hydrological parameters, parameter estimation, least square method, correlation function method, input-output model, typhoon gladys.

• Structural Engineering and Mechanics
• /
• v.61 no.6
• /
• pp.807-816
• /
• 2017

In this paper, we introduce an efficient new model reduction method, named the automated static condensation method, which is developed for the local analysis of large finite element models. The algebraic multilevel substructuring procedure is modified appropriately, and then applied to the original static condensation method. The retained substructure, which is the local finite element model to be analyzed, is defined, and then the remaining part of the global model is automatically partitioned into many omitted substructures in an algebraic perspective. For an efficient condensation procedure, a substructural tree diagram and substructural sets are established. Using these, the omitted substructures are sequentially condensed into the retained substructure to construct the reduced model. Using several large practical engineering problems, the performance of the proposed method is demonstrated in terms of its solution accuracy and computational efficiency, compared to the original static condensation method and the superelement technique.

• Industrial Engineering and Management Systems
• /
• v.7 no.1
• /
• pp.44-50
• /
• 2008

Focusing on the exponential smoothing method equivalent to (1, 1) order ARMA model equation, a new method of estimating smoothing constant using exponential smoothing method is proposed. This study goes beyond the usual method of arbitrarily selecting a smoothing constant. First, an estimation of the ARMA model parameter was made and then, the smoothing constants. The empirical example shows that the theoretical solution satisfies minimum variance of forecasting error. The new method was also applied to the stock market price of electrical machinery industry (6 major companies in Japan) and forecasting was accomplished. Comparing the results of the two methods, the new method appears to be better than the ARIMA model. The result of the new method is apparently good in 4 company data and is nearly the same in 2 company data. The example provided shows that the new method is much simpler to handle than ARIMA model. Therefore, the proposed method would be better in these general cases. The effectiveness of this method should be examined in various cases.

• Management Science and Financial Engineering
• /
• v.21 no.2
• /
• pp.25-30
• /
• 2015

This short article compares different solution methods for a basic RBC model (Hansen, 1985). We solve and simulate the model using two main algorithms: the methods of perturbation and projection, respectively. One novelty is that we offer a type of the hybrid method: we compute easily a second-order approximation to decision rules and use that approximation as an initial guess for finding Chebyshev polynomials. We also find that the second-order perturbation method is most competitive in terms of accuracy for standard RBC model.

• Journal of Power System Engineering
• /
• v.11 no.3
• /
• pp.66-72
• /
• 2007

This study proposed new method to decide the reference model necessary for design PID controller. In generally, control design problems using the reference model have the following two factors. One factor is that numerical model of the controlled system can be obtained extremely, and the other is that specification for the closed-loop dynamic performance is pure moderate. Therefore, the control design procedure is essentially based on the partial reference model matching which offers a reasonable method to simplify the design and the controller configuration under the controlled system uncertainty. ITAE(Integral of time-multiplied absolute error) performance index and Kitamori method etc. which were used a reference model method had a limit to settling time and rising time of reference model that it arrived to steady state response according to the controlled system. On this study, if it only knew peak time of overshoot and settling time by measurement signal of the controlled system, it can be made the reference model easily. We proposed new method to improve performance index of the reference model superior to existing reference model index and illustrate the numerical simulation results to show the effectiveness of proposed control method design.

• Communications for Statistical Applications and Methods
• /
• v.30 no.5
• /
• pp.501-516
• /
• 2023

Knowledge distillation (KD) refers to extracting knowledge from a large and complex model (teacher) and transferring it to a relatively small model (student). This can be done by training the teacher model to obtain the activation function values of the hidden or the output layers and then retraining the student model using the same training data with the obtained values. Recently, relational KD (RKD) has been proposed to extract knowledge about relative differences in training data. This method improved the performance of the student model compared to conventional KDs. In this paper, we propose a new method for RKD by introducing a new loss function for RKD. The proposed loss function is defined using the area difference between the teacher model and the student model in a specific hidden layer, and it is shown that the model can be successfully compressed, and the generalization performance of the model can be improved. We demonstrate that the accuracy of the model applying the method proposed in the study of model compression of audio data is up to 1.8% higher than that of the existing method. For the study of model generalization, we demonstrate that the model has up to 0.5% better performance in accuracy when introducing the RKD method to self-KD using image data.