Dohyun Kim

Design Analysis and Experimental Evaluation of an MR Fluid Clutch

An MRC (Magneto-Rheological Clutch), a device to transmit torque by shear stress of MR fluids, has the property that its power transmissibility changes quickly in response to control signal. In this study, we consider methods to predict performance of an MRC. First, we anticipate the performance of an MRC with a simplified mathematical model and second, we predict the performance in consideration of the applied magnetic field and viscosity distribution of fluids caused by the field. Between the two methods, compared with experimental results, it is shown that the numerical method is closer to reality than the simplified one.

Microfluidic Western blotting: Cationic surfactant based protein sizing integrated with electrostatic immobilization

We report on fully-functional Western blotting demonstrated in an automated microfluidic format. The novel microfluidic device incorporates contiguous polyacrylamide (PA) gel regions tailored for different biochemical/physical functionality in a microchamber. Our photopatterned polymer design strategy enables assay performance not possible with traditional bench-top Western blotting: rapid completion of multiple Western-blotting steps (∼2 hours), minute sample consumption (< 10 ng), and no manual intervention. Using this format, three key Western blotting steps are seamlessly integrated in a single microfluidic device: (1) accurate protein sizing and separation via cationic detergent based polyacrylamide gel electrophoresis (Cat-PAGE), (2) immobilization of proteins after separation using electrostatic interaction with PA gel matrix, and (3) antibody-based detection and quantitation of immobilized protein targets. As a means to characterize system performance, a protein ladder consisting of protein G, ovalbumin (OVA), bovine serum albumin (BSA), and α-actinin is separated and then immobilized with mobility information preserved. Subsequent immunoblotting is also demonstrated. We see this format as forming the basis for unmatched protein blotting as is relevant to basic life sciences research.

Inverse transitions in a spin-glass model on a scale-free network.

In this paper, we will investigate critical phenomena by considering a model spin glass on scale-free networks. For this purpose, we consider the Ghatak-Sherrington (GS) model, a spin-1 spin-glass model with a crystal field, instead of the usual Ising-type model. Scale-free networks on which the GS model is placed are constructed from the static model, in which the number of vertices is fixed from the beginning. On the basis of the replica-symmetric solution, we obtain the analytical solutions, i.e., free energy and order parameters, and we derive the various phase diagrams consisting of the paramagnetic, ferromagnetic, and spin-glass phases as functions of temperature T, the degree exponent λ, the mean degree K, and the fraction of the ferromagnetic interactions ρ. Since the present model is based on the GS model, which considers the three states (S = 0, ± 1), the S = 0 state plays a crucial role in the λ-dependent critical behavior: glass transition temperature T(g) has a finite value, even when 2 < λ < 3. In addition, when the crystal field becomes nonzero, the present model clearly exhibits three types of inverse transitions, which occur when an ordered phase is more entropic than a disordered one.

Robust kernel-based regression with bounded influence for outliers

The kernel-based regression (KBR) method, such as support vector machine for regression (SVR) is a well-established methodology for estimating the nonlinear functional relationship between the response variable and predictor variables. KBR methods can be very sensitive to influential observations that in turn have a noticeable impact on the model coefficients. The robustness of KBR methods has recently been the subject of wide-scale investigations with the aim of obtaining a regression estimator insensitive to outlying observations. However, existing robust KBR (RKBR) methods only consider Y-space outliers and, consequently, are sensitive to X-space outliers. As a result, even a single anomalous outlying observation in X-space may greatly affect the estimator. In order to resolve this issue, we propose a new RKBR method that gives reliable result even if a training data set is contaminated with both Y-space and X-space outliers. The proposed method utilizes a weighting scheme based on the hat matrix that resembles the generalized M-estimator (GM-estimator) of conventional robust linear analysis. The diagonal elements of hat matrix in kernel-induced feature space are used as leverage measures to downweight the effects of potential X-space outliers. We show that the kernelized hat diagonal elements can be obtained via eigen decomposition of the kernel matrix. The regularized version of kernelized hat diagonal elements is also proposed to deal with the case of the kernel matrix having full rank where the kernelized hat diagonal elements are not suitable for leverage. We have shown that two kernelized leverage measures, namely, the kernel hat diagonal element and the regularized one, are related to statistical distance measures in the feature space. We also develop an efficiently kernelized training algorithm for the parameter estimation based on iteratively reweighted least squares (IRLS) method. The experimental results from simulated examples and real data sets demonstrate the robustness of our proposed method compared with conventional approaches.

New configuration of bidirectional intelligent semiconductor transformer with high-frequency AC-DC converter

This paper proposes a new configuration of bidirectional intelligent semiconductor transformer with rating of 1.9kV/220V, 2kVA. The proposed transformer consists of high-voltage high-frequency AC-DC rectifier, and low-voltage DC-DC and DC-AC converters. The operational feasibility of proposed transformer was verified by computer simulation with PSIM software. Based on the simulation results, a hardware prototype with rating of 1.9kV/220V, 2kVA was built and tested in the lab to confirm the feasibility of hardware implementation. Using three units of this transformer, a 3-phase transformer with rating of 3.3kV/380V, 6kVA can be built. The proposed transformer could be applicable for implementing the smart grid.

Operation Mode Development and Evaluation for Grid-Tied PMSG Wind Power System Combined with Battery Energy Storage

This paper describes the operation mode development for the grid-tied PMSG(permanent magnet synchronous generator) wind power system combined with a battery energy storage. The development of operation modes was carried out through simulations with PSCAD/EMTDC software and experiments with a 10kW hardware prototype. The detailed simulation models for PMSG wind power system and battery energy storage were developed using user-defined models programed with C-code. A 10kW hardware simulator was built and tested in connection with the local load and the utility power. The simulation and experimental results confirm that the grid-tied PMSG wind power system combined with battery energy storage can supply highly reliable power to the local load in various operation modes.

Leakage Current Behaviors of Ba0.5Sr0.5TiO3 Thin Films on Pt, RuO2, and Pt/RuO2 Bottom Electrodes

Ba 0.5 Sr 0.5 TiO 3 thin films were grown by rf-magnetron sputtering. Pt and RuO 2 films were used as bottom electrodes of the Ba 0.5 Sr 0.5 TiO 3 thin films. Further, a Pt/RuO 2 hybrid electrode was adopted as an electrode for the film. Structural properties of the Ba 0.5 Sr 0.5 TiO 3 thin films were found to be closely related to the type of the bottom electrodes. Dielectric constants of the films on Pt, RuO 2 , and Pt/RuO 2 were measured as 500, 320, and 450 in the range of 100 Hz ˜ 1 MHz, respectively. Leakage current characteristics of the films were also found to be strongly dependent on the type of the electrodes and their microstructures.

Embedded variable selection method using signomial classification

We propose two variable selection methods using signomial classification. We attempt to select, among a set of the input variables, the variables that lead to the best performance of the classifier. One method repeatedly removes variables based on backward selection, whereas the second method directly selects a set of variables by solving an optimization problem. The proposed methods conduct variable selection considering nonlinear interactions of variables and obtain a signomial classifier with the selected variables. Computational results show that the proposed methods more effectively selects desirable variables for predicting output and provide the classifiers with better or comparable test error rates, as compared with existing methods.

Investigation of biochemical property changes in activation-induced CD8+ T cell apoptosis using Raman spectroscopy

Abstract. The study was to investigate the changes in biochemical properties of activated mature CD8+ T cells related to apoptosis at a molecular level. We confirmed the activation and apoptosis of CD8+ T cells by fluorescence-activated cell sorting and atomic force microscopy and then performed Raman spectral measurements on activated mature CD8+ T cells and cellular deoxyribose nucleic acid (DNA). In the activated mature CD8+ T cells, there were increases in protein spectra at 1002 and 1234  cm−1. In particular, to assess the apoptosis-related DNA spectral signatures, we investigated the spectra of the cellular DNA isolated from resting and activated mature CD8+ T cells. Raman spectra at 765 to 786  cm−1 and 1053 to 1087  cm−1 were decreased in activated mature DNA. In addition, we analyzed Raman spectrum using the multivariate statistical method including principal component analysis. Raman spectra of activated mature DNA are especially well-discriminated from those of resting DNA. Our findings regarding the biochemical and structural changes associated with apoptosis in activated mature T cells and cellular DNA according to Raman spectroscopy provide important insights into allospecific immune responses generated after organ transplantation, and may be useful for therapeutic manipulation of the immune response.

A Network Structural Approach to the Link Prediction Problem

The link prediction problem is an emerging real-life social network problem in which data mining techniques have played a critical role. It arises in many practical applications such as recommender systems, information retrieval, and marketing analysis of social networks. We propose a new mathematical programming approach for predicting a future network using estimated node degree distribution identified from historical data. The link prediction problem is formulated as an integer programming problem that maximizes the sum of link scores probabilities with respect to the estimated node degree distribution. The performance of the proposed framework is tested on real-life social networks, and the computational results show that the proposed approach can improve the performance of previously published link prediction methods.