• The Transactions of The Korean Institute of Electrical Engineers
• /
• v.67 no.1
• /
• pp.114-123
• /
• 2018

This paper concerns a design and learning method of softmax function neural networks based on K-means clustering. The partial discharge data Information is preliminarily processed through simulation using an Epoxy Mica Coupling sensor and an internal Phase Resolved Partial Discharge Analysis algorithm. The obtained information is processed according to the characteristics of the pattern using a Motor Insulation Monitoring System program. At this time, the processed data are total 4 types that void discharge, corona discharge, surface discharge and slot discharge. The partial discharge data with high dimensional input variables are secondarily processed by principal component analysis method and reduced with keeping the characteristics of pattern as low dimensional input variables. And therefore, the pattern classifier processing speed exhibits improved effects. In addition, in the process of extracting the partial discharge data through the MIMS program, the magnitude of amplitude is divided into the maximum value and the average value, and two pattern characteristics are set and compared and analyzed. In the first half of the proposed partial discharge pattern classifier, the input and hidden layers are classified by using the K-means clustering method and the output of the hidden layer is obtained. In the latter part, the cross entropy error function is used for parameter learning between the hidden layer and the output layer. The final output layer is output as a normalized probability value between 0 and 1 using the softmax function. The advantage of using the softmax function is that it allows access and application of multiple class problems and stochastic interpretation. First of all, there is an advantage that one output value affects the remaining output value and its accompanying learning is accelerated. Also, to solve the overfitting problem, L2-normalization is applied. To prove the superiority of the proposed pattern classifier, we compare and analyze the classification rate with conventional radial basis function neural networks.

• Journal of the Korea Computer Graphics Society
• /
• v.18 no.4
• /
• pp.1-8
• /
• 2012

We present improved deformation energy to enhance the visual quality of a shape deformation technique, where we preserve the local structure of an input planar shape. The deformation energy, in general, consists of several constraints such as Laplacian coordinate constraint to preserve the quality of deformed silhouette edges, mean value coordinates and edge length constraints to preserve the quality of deformed internal shape, and user-specified position constraints to control the shape deformation. When the positions of user-specified vertices change, shape deformation techniques compute the positions of the other vertices by means of nonlinear least squares optimization to minimize the deformation energy. When a user-specified vertex changes its position rapidly, it is frequently observed that the visual quality of the deformed shape decrease rapidly, which is mainly caused by unnecessary enlargement of the Laplacian vectors and unnecessary change of the edge directions along the boundary of the shape. In this paper, we propose improved deformation energy by prohibiting the Laplacian and edge length constraints from changing unnecessarily. The proposed deformation energy incorporated with well-known optimization technique can enhance the visual quality of shape deformation along the silhouette and within the interior of the planar shape while sacrificing only a little execution time.

• Transactions of the Korean Society of Mechanical Engineers B
• /
• v.35 no.6
• /
• pp.649-656
• /
• 2011

A model for the numerical simulation of lithium-ion batteries (LIBs) is developed for use in battery cell design, with a view to improving the performances of such batteries. The model uses Newman-type electrochemical and transfer $theories^{(1,2)}$ to describe the behavior of the lithium-ion cell, together with the Levenberg-Marquardt optimization scheme to estimate the performance or design parameters in nonlinear problems. The mathematical model can provide an insight into the mechanism of LIB behavior during the charging/discharging process, and can therefore help to predict cell performance. Furthermore, by means of least-squares fitting to experimental discharge curves measured at room temperature, we were able to obtain the values of transport and kinetic parameters that are usually difficult to measure. By comparing the calculated data with the life-test discharge curves (SB LiMotive cell), we found that the capacity fade is strongly dependent on the decrease in the reaction area of active materials in the anode and cathode, as well as on the electrolyte diffusivity.

• Journal of Advanced Marine Engineering and Technology
• /
• v.40 no.7
• /
• pp.647-654
• /
• 2016

Indoor localization based on visible-light communication using the received signal strength intensity (RSSI) has been widely studied because of its high accuracy compared with other wireless localization methods. However, because the RSSI can vary according to the inclination and azimuth of the receiver, a large error can occur, even at the same position. In this paper, we propose a visible-light communication-based 3-D indoor positioning algorithm using the Gauss-Newton technique in order to reduce the errors caused by the change in the inclination of the receiver. The proposed system reduces the amount of computations by selecting the initial position of the receiver through the linear least-squares method (LSM), which is applied to the RSSIs, and improves the position accuracy by applying the Gauss-Newton technique to the 3-D nonlinear model that contains the RSSIs acquired by the changes in the azimuth and inclination of the receiver. In order to verify the validity of the proposed algorithm in an indoor space with dimensions of $6{\times}6{\times}3m$ where 16 LED lights are installed, we compare and analyze the errors of the conventional linear LSM-based trilateration technique and the proposed algorithm according to the changes in the inclination and azimuth of the receiver. The experimental results show that the location accuracy of the proposed algorithm is improved by 82.5% compared to the conventional LSM-based trilateration technique.

• Korean journal of applied entomology
• /
• v.35 no.2
• /
• pp.119-125
• /
• 1996

Temperature-dependent development study for overwintered eggs of Pseudococcus comstocki (Kuwana) wasconducted to develop a forecasting model for egg hatch date. Hatch times of overwintered eggs were comparedat five constant temperatures (10, 15, 20, 25, 27$^{\circ}$C) and different collection dates. A nonlinear, four-parameterdevelopmental model with high temperature inhibition accurately described (R2=0.9948) mean developmentalrates of all temperatures. Variation in developmental times was modeled(~~=0.972w9)it h a cumulative Weibullfunction. Least-squares linear regression (rate=O.O06358[Temp.]-0.07566)d escribed development in the linearregion (15-25$^{\circ}$C) of the development curve. The low development threshold temperature was estimated 11.9"Cand 154.14 degree-days were required for complete development. The linear degree-day model (thermal summation)and rate summation model (Wagner et al. 1985) were validated using field phenology data. In degreedaymodels, mean-minus-base method, sine wave method, and rectangle method were used in estimation of dailythermal units. Mean-minus-base method was 18 to 28d late, sine wave method was 11 to 14d late, rectanglemethod was 3 to 5d late, and rate summation model was 2 to 3d late in predicting 50% hatch of overwinteredeggs. hatch of overwintered eggs.

• Wind and Structures
• /
• v.30 no.4
• /
• pp.433-450
• /
• 2020

The strong turbulence characteristic of typhoon not only will significantly change flow field characteristics surrounding the large-scale wind turbine and aerodynamic force distribution on surface, but also may cause morphological evolution of coast dune and thereby form sand storms. A 5MW horizontal-axis wind turbine in a wind power plant of southeastern coastal areas in China was chosen to investigate the distribution law of additional loads caused by wind-sand coupling movement of coast dune at landing of strong typhoons. Firstly, a mesoscale Weather Research and Forecasting (WRF) mode was introduced in for high spatial resolution simulation of typhoon "Megi". Wind speed profile on the boundary layer of typhoon was gained through fitting based on nonlinear least squares and then it was integrated into the user-defined function (UDF) as an entry condition of small-scaled CFD numerical simulation. On this basis, a synchronous iterative modeling of wind field and sand particle combination was carried out by using a continuous phase and discrete phase. Influencing laws of typhoon and normal wind on moving characteristics of sand particles, equivalent pressure distribution mode of structural surface and characteristics of lift resistance coefficient were compared. Results demonstrated that: Compared with normal wind, mesoscale typhoon intensifies the 3D aerodynamic distribution mode on structural surface of wind turbine significantly. Different from wind loads, sand loads mainly impact on 30° ranges at two sides of the lower windward region on the tower. The ratio between sand loads and wind load reaches 3.937% and the maximum sand pressure coefficient is 0.09. The coupling impact effect of strong typhoon and large sand particles is more significant, in which the resistance coefficient of tower is increased by 9.80% to the maximum extent. The maximum resistance coefficient in typhoon field is 13.79% higher than that in the normal wind field.

• Journal of Intelligence and Information Systems
• /
• v.25 no.4
• /
• pp.53-65
• /
• 2019

Object tracking is one of important steps to achieve video-based surveillance systems. Object tracking is considered as an essential task similar to object detection and recognition. In order to perform object tracking, various machine learning methods (e.g., least-squares, perceptron and support vector machine) can be applied for different designs of tracking systems. In general, generative methods (e.g., principal component analysis) were utilized due to its simplicity and effectiveness. However, the generative methods were only focused on modeling the target object. Due to this limitation, discriminative methods (e.g., binary classification) were adopted to distinguish the target object and the background. Among the machine learning methods for binary classification, total error rate minimization can be used as one of successful machine learning methods for binary classification. The total error rate minimization can achieve a global minimum due to a quadratic approximation to a step function while other methods (e.g., support vector machine) seek local minima using nonlinear functions (e.g., hinge loss function). Due to this quadratic approximation, the total error rate minimization could obtain appropriate properties in solving optimization problems for binary classification. However, this total error rate minimization was based on a batch mode setting. The batch mode setting can be limited to several applications under offline learning. Due to limited computing resources, offline learning could not handle large scale data sets. Compared to offline learning, online learning can update its solution without storing all training samples in learning process. Due to increment of large scale data sets, online learning becomes one of essential properties for various applications. Since object tracking needs to handle data samples in real time, online learning based total error rate minimization methods are necessary to efficiently address object tracking problems. Due to the need of the online learning, an online learning based total error rate minimization method was developed. However, an approximately reweighted technique was developed. Although the approximation technique is utilized, this online version of the total error rate minimization could achieve good performances in biometric applications. However, this method is assumed that the total error rate minimization can be asymptotically achieved when only the number of training samples is infinite. Although there is the assumption to achieve the total error rate minimization, the approximation issue can continuously accumulate learning errors according to increment of training samples. Due to this reason, the approximated online learning solution can then lead a wrong solution. The wrong solution can make significant errors when it is applied to surveillance systems. In this paper, we propose an exactly reweighted technique to recursively update the solution of the total error rate minimization in online learning manner. Compared to the approximately reweighted online total error rate minimization, an exactly reweighted online total error rate minimization is achieved. The proposed exact online learning method based on the total error rate minimization is then applied to object tracking problems. In our object tracking system, particle filtering is adopted. In particle filtering, our observation model is consisted of both generative and discriminative methods to leverage the advantages between generative and discriminative properties. In our experiments, our proposed object tracking system achieves promising performances on 8 public video sequences over competing object tracking systems. The paired t-test is also reported to evaluate its quality of the results. Our proposed online learning method can be extended under the deep learning architecture which can cover the shallow and deep networks. Moreover, online learning methods, that need the exact reweighting process, can use our proposed reweighting technique. In addition to object tracking, the proposed online learning method can be easily applied to object detection and recognition. Therefore, our proposed methods can contribute to online learning community and object tracking, detection and recognition communities.