• Journal of the Korean Society for Railway
• /
• v.15 no.2
• /
• pp.135-140
• /
• 2012

Interior Permanent Magnet Synchronous motors (IPMSM) with concentrated winding are superior to distributed winding in the power density point of view. But it causes huge amount of eddy current losses on the permanent magnet. This paper presents the optimal permanent magnet V-shape on the rotor of an interior permanent magnet synchronous motor to reduce the core losses and improve the performance. Each eddy current loss on permanent magnet has been investigated in detail by using FEM (Finite Element Method) instead of equivalent magnetic circuit network method in order to consider saturation and non-linear magnetic property. Simulation-based design of experiment is also applied to avoid large number of analyses according to each design parameter and consider expected interactions among parameters. Consequently, the optimal design to reduce the core loss on the permanent magnet while maintaining or improving motor performance is proposed by an optimization algorithm using regression equation derived and lastly, it is verified by FEM.

• KSCE Journal of Civil and Environmental Engineering Research
• /
• v.34 no.6
• /
• pp.1779-1786
• /
• 2014

Artificial Intelligence-based techniques have been applied to problems where mathematical relations can not be presented due to complicatedness of the physical process. A representative example in hydraulics is the local scour around bridge piers. This study presents a GEP model for predicting the local scour around bridge piers. The model is trained by 64 laboratory data to build the regression equation, and the constructed model is verified against 33 laboratory data. Comparisons between the models with dimensional and normalized variables reveals that the GEP model with dimensional variables predicts better. The proposed model is now applied to two field datasets. It is found that the MAPE of the scour depths predicted by the GEP model increases compared with the predictions of local scours in laboratory scale. In addition, the model performance increases significantly when the model is trained by the field dataset rather than the laboratory dataset. The findings suggest that apart from the ANN model, GEP model is a sound and reliable model for predicting local scour depth.

• The Journal of Korean Institute of Communications and Information Sciences
• /
• v.39C no.9
• /
• pp.811-819
• /
• 2014

Activity recognition is a key component in identifying the context of a user for providing services based on the application such as medical, entertainment and tactical scenarios. Instead of applying numerous sensor devices, as observed in many previous investigations, we are proposing the use of smartphone with its built-in multimodal sensors as an unobtrusive sensor device for recognition of six physical daily activities. As an improvement to previous works, accelerometer, gyroscope and magnetometer data are fused to recognize activities more reliably. The evaluation indicates that the IBK classifier using window size of 2s with 50% overlapping yields the highest accuracy (i.e., up to 99.33%). To achieve this peak accuracy, simple time-domain and frequency-domain features were extracted from raw sensor data of the smartphone.

• Smart Structures and Systems
• /
• v.20 no.5
• /
• pp.619-636
• /
• 2017

Magnetorheological (MR) damper is a type of controllable device widely used in vibration mitigation. This device is highly nonlinear, and exhibits strongly hysteretic behavior that is dependent on both the motion imposed on the device and the strength of the surrounding electromagnetic field. An accurate model for understanding and predicting the nonlinear damping force of the MR damper is crucial for its control applications. The MR damper models are often identified off-line by conducting regression analysis using data collected under constant voltage. In this study, a MR damper model is integrated with a model for the power supply unit (PSU) to consider the dynamic behavior of the PSU, and then a real-time nonlinear model updating technique is proposed to accurately identify this integrated MR damper model with the efficiency that cannot be offered by off-line methods. The unscented Kalman filter is implemented as the updating algorithm on a cyber-physical model updating platform. Using this platform, the experimental study is conducted to identify MR damper models in real-time, under in-service conditions with time-varying current levels. For comparison purposes, both off-line and real-time updating methods are applied in the experimental study. The results demonstrate that all the updated models can provide good identification accuracy, but the error comparison shows the real-time updated models yield smaller relative errors than the off-line updated model. In addition, the real-time state estimates obtained during the model updating can be used as feedback for potential nonlinear control design for MR dampers.

• Journal of Sasang Constitutional Medicine
• /
• v.27 no.2
• /
• pp.211-221
• /
• 2015

Objectives The aim of this study is to develop a Korea Sasang Constitutional Diagnostic Questionnaire (KS-15) with practicality and validity. Methods The data of body shape, temperament and symptoms were collected from 11 nationwide multi-centers in Korean Medical Clinics between Nov. 2009 to Aug. 2012. KS-15 algorithm was developed through Multinomial Logistic Regression and was applied for test group to verify it's classificatory function. Results The weight value of variables of KS-15 were different between male and female. The concordance of KS-15 with the expert in training group was 58.34% (72.8% in Tae-eumin, 54.8% in Soeumin, and 44.1% in Soyangin). The concordance of KS-15 with the expert in test group was 63.1% (63% in Tae-eumin, 83.3% in Soeumin, and 51.9% in Soyangin). Conclusions KS-15 seems to have suitable implement with practicality and validity. Further study is necessary to verify it accuracy and cut-off.

• Journal of Engineering Education Research
• /
• v.13 no.4
• /
• pp.51-59
• /
• 2010

In order to manage the lectures efficiently in the university and improve the educational outcome, the process is needed that make diagnosis of the present educational outcome of each classes on a lecture and find factors of educational outcome. In most studies for finding the factors of the efficient lecture, statistical methods such as association analysis, regression analysis are used usually, and recently decision tree analysis is employed, too. The decision tree analysis have the merits that is easy to understand a result model, and to be easy to apply for the decision making, but have the weaknesses that is not strong for characteristic of input data such as multicollinearity. This paper indicates the weaknesses of decision tree analysis, and suggests the experimental solution using multiple decision tree algorithm to supplement these problems. The experimental result shows that the suggested method is more effective in finding the reliable factors of the educational outcome.

• Journal of Biosystems Engineering
• /
• v.20 no.1
• /
• pp.95-103
• /
• 1995

A series of study has been conducted to develop a multi-purpose fruit grader using a black ＆ white image processing system equipped with a 550 nm interference filter. A device and high performance algorithms were developed for sizing and color grading of Fuji apple in the previous study. In this study an emphasis was put on finding correlations between weights of several kinds of fruits and their area fractions(AF), and on compensating the blurring effect upon sizing and color grading by conveying speed of fruit. Also, the effect of orientation and direction of fruit on conveyor during image forming was analyzed to identify any difficulty (or utilizing an automatic fruit feeder. The results are summarized as follows. 1. The correlation coefficients(r) between the weights of fruits and their image sizes were 0.984~0.996 for apples, 0.983~0.990 for peachs, 0.995 for tomato, 0.986 for sweet persimmon and 0.970~0.993 for pears. 2. It was possible to grade fruits by color with the area weighted mean gray values(AWMGV) based on the mean gray valves of direct image and the compensated values of reflected image of a fruit, and also possible to sort fruits by size with AF. Accuracies in sizing and color grading ranged over 81.0% ~95.0% and 82.0% ~89.7% respectively as compared with results from sizing by electronic weight scale and grading by expert. 3. The blurring effect on the sizing and color grading depending on conveying speed was identified and regression equations were derived. 4. It was found that errors in sizing and coloring grading due to the change in direction and orientation of Fuji apple on the conveyor were not significant as far as the stem end of apple keeping upward.

• The Transactions of The Korean Institute of Electrical Engineers
• /
• v.66 no.12
• /
• pp.1772-1781
• /
• 2017

In this study, the design methodology for alleviating the overfitting problem of Polynomial Neural Networks(PNN) is realized with the aid of two kinds techniques such as L2 regularization and Sum of Squared Coefficients (SSC). The PNN is widely used as a kind of mathematical modeling methods such as the identification of linear system by input/output data and the regression analysis modeling method for prediction problem. PNN is an algorithm that obtains preferred network structure by generating consecutive layers as well as nodes by using a multivariate polynomial subexpression. It has much fewer nodes and more flexible adaptability than existing neural network algorithms. However, such algorithms lead to overfitting problems due to noise sensitivity as well as excessive trainning while generation of successive network layers. To alleviate such overfitting problem and also effectively design its ensuing deep network structure, two techniques are introduced. That is we use the two techniques of both SSC(Sum of Squared Coefficients) and $L_2$ regularization for consecutive generation of each layer's nodes as well as each layer in order to construct the deep PNN structure. The technique of $L_2$ regularization is used for the minimum coefficient estimation by adding penalty term to cost function. $L_2$ regularization is a kind of representative methods of reducing the influence of noise by flattening the solution space and also lessening coefficient size. The technique for the SSC is implemented for the minimization of Sum of Squared Coefficients of polynomial instead of using the square of errors. In the sequel, the overfitting problem of the deep PNN structure is stabilized by the proposed method. This study leads to the possibility of deep network structure design as well as big data processing and also the superiority of the network performance through experiments is shown.

• Proceedings of the Korean Vacuum Society Conference
• /
• 2010.02a
• /
• pp.413-413
• /
• 2010

In semiconductor manufacturing field, all equipments have various sensors to diagnosis the situations of processes. For increasing the accuracy of diagnosis, hundreds of sensors are emplyed. As sensors provide millions of data, the process diagnosis from them are unrealistic. Besides, in some cases, the results from some data which have same conditions are different. We want to find some information, such as data and knowledge, from the data. Nowadays, fault detection and classification (FDC) has been concerned to increasing the yield. Certain faults and no-faults can be classified by various FDC tools. The uncertainty in semiconductor manufacturing, no-faulty in faulty and faulty in no-faulty, has been caused the productivity to decreased. From the uncertainty, the rough set theory is a viable approach for extraction of meaningful knowledge and making predictions. Reduction of data sets, finding hidden data patterns, and generation of decision rules contrasts other approaches such as regression analysis and neural networks. In this research, a RGB sensor was used for diagnosis plasma instead of optical emission spectroscopy (OES). RGB data has just three variables (red, green and blue), while OES data has thousands of variables. RGB data, however, is difficult to analyze by human's eyes. Same outputs in a variable show different outcomes. In other words, RGB data includes the uncertainty. In this research, by rough set theory, decision rules were generated. In decision rules, we could find the hidden data patterns from the uncertainty. RGB sensor can diagnosis the change of plasma condition as over 90% accuracy by the rough set theory. Although we only present a preliminary research result, in this paper, we will continuously develop uncertainty problem solving data mining algorithm for the application of semiconductor process diagnosis.

• Transactions of the Korean Society of Mechanical Engineers A
• /
• v.41 no.8
• /
• pp.691-701
• /
• 2017

The barbell squat is a fundamental physical exercise for strengthening the lower body and core muscles. It is an integral part of training and conditioning programs in sports, rehabilitation, and fitness. In this paper, we proposed a virtual test framework for squat exercises using a Smith machine to simulate joint torques and muscle forces, based on an integrated product-human model and motion synthesis algorithms. We built a muscular skeletal human model with boundary conditions modeling the interactions between the human body and a machine or the ground. To validate the model, EMG, external forces, and squat motions were captured through physical experiments by varying the foot position. A regression-based motion synthesis algorithm was developed based on the captured squat motions to generate a new motion for a given foot position. The proposed approach is expected to reduce the need for physical experiments in the development of training programs.