• Industrial Engineering and Management Systems
• /
• v.4 no.1
• /
• pp.54-67
• /
• 2005

This paper provides an evaluation of an optimization-based, multiple-input double-output (MIDO) run-to-run (R2R) control scheme for general semiconductor manufacturing processes. The controller in this research, termed adaptive dual response optimizing controller (ADROC), can serve as a process optimizer as well as a recipe regulator between consecutive runs of wafer fabrication. In evaluation, it is assumed that the equipment model could be appropriately described by a pair of second-order polynomial functions in terms of a set of controllable variables. Of practical relevance is to consider a drifting effect in the equipment model since in common semiconductor practice the process tends to drift due to machine aging and tool wearing. We select a typical application of R2R control to chemical mechanical planarization (CMP) in semiconductor manufacturing in this evaluation, and there are five different CMP process scenarios demonstrated, including mean shift, variance increase, and IMA disturbances. For the controller, ADROC, an on-line estimation technique is implemented in a self-tuning (ST) control manner for the adaptation purpose. Subsequently, an ad hoc global optimization algorithm based on the dual response approach, arising from the response surface methodology (RSM) literature, is used to seek the optimum recipe within the acceptability region for the execution of next run. The main components of ADROC are described and its control performance is assessed. It reveals from the evaluation that ADROC can provide excellent control actions for the MIDO R2R situations even though the process exhibits complicated, nonlinear interaction effects between control variables, and the drifting disturbances.

• The Transactions of The Korean Institute of Electrical Engineers
• /
• v.66 no.2
• /
• pp.416-424
• /
• 2017

In this study, we propose a design of digits recognition system based on RBFNNs through a comparative study of pre-processing algorithms in order to recognize digits in handwritten. Histogram of Oriented Gradient(HOG) is used to get the features of digits in the proposed digits recognition system. In the pre-processing part, a dimensional reduction is executed by using Principal Component Analysis(PCA) and (2D)2PCA which are widely adopted methods in order to minimize a loss of the information during the reduction process of feature space. Also, The architecture of radial basis function neural networks consists of three functional modules such as condition, conclusion, and inference part. In the condition part, the input space is partitioned with the use of fuzzy clustering realized by means of the Fuzzy C-Means algorithm. Also, it is used instead of gaussian function to consider the characteristic of input data. In the conclusion part, the connection weights are used as the extended type of polynomial expression such as constant, linear, quadratic and modified quadratic. By using MNIST handwritten digit benchmarking database, experimental results show the effectiveness and efficiency of proposed digit recognition system when compared with other studies.

• 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.

• Journal of the Korean Institute of Illuminating and Electrical Installation Engineers
• /
• v.20 no.6
• /
• pp.55-62
• /
• 2006

For the fault diagnosis of three-phase induction motors, we construct a experimental unit and then develop a diagnosis algorithm based on pattern recognition. The experimental unit consists of machinery module for induction motor drive and data acquisition module to obtain the fault signal. As the first step for diagnosis procedure, preprocessing is performed to make the acquired current simplified and normalized. To simplify the data, three-phase current is transformed into the magnitude of Concordia vector. As the next step, feature extraction is performed by kernel principal component analysis(KPCA) and linear discriminant analysis(LDA). Finally, we used the classifier based on radial basis function(RBF) network. To show the effectiveness, the proposed diagnostic system has been intensively tested with the various data acquired under different electrical and mechanical faults with varying load.

• Journal of the Korean Institute of Navigation
• /
• v.7 no.2
• /
• pp.1-45
• /
• 1983

This paper concerns the applications of the Kalman filter to navigation and the develment of computer programs of the navigational calculations. Methods to apply the Kalman filter to celestial fix, fix by cross bearing and cocked hat are proposed, and numerical simulations under various noise conditiions are conducted. The accuracy of the optimal positions obtained by the Kalman filter is compared with that of the fixed positiions by radial error method. In the case of celestial fix, an algorithm to estimate the optimal positions by using the linear Kalman filter is presented. The optimal positions by the Kalman filter are compared with the running fixes and with the most probable positions obtained from a single line of position. It is confirmed that the resutls of the proposed method are more accurate than the others. In practical piloting, bearings are generally measured intermittently and the measurement process is nonlinear. It is, therefore, difficult for us to apply the Kalman filter to fix by cross bearing. In order to be used in such an unfavorable case, the extended Kalman filter is revised and the aplicability of the revised extended Kalman filter is checked by numerical simulation under various noise conditions. In a cocked hat, an inside or outside fix is dependent only upon azimuth spread, if the error of each line of position is assumed to be equal both in magnitude and sign. A new technique of selecting a ship's position between an inside fix and an outside fix in a cocked hat by using fix determinant derived from the equation of three lines of position is also presented. The relations among the optimal position by Kalman filter, incentre (or excentre) and random error centtre of the cocked hat are discussed theoretically and the accuracy of the optimal position is compared with that of the others by numerical simulation.

• Korean Journal of Hydrosciences
• /
• v.7
• /
• pp.37-49
• /
• 1996

This study aims at suggesting an alternative to improve flood controling capacity according to the cument design criteria for the existing Soyanggang Multi-purpose Dam which was constructed 20 years ago as the largest dam in Korea. The peak inflow of the adopted probable maximum flood (PMF) at the time of construction was 13,500 $m^3$/s. However, the newly estimated peak inflow of the PMF is 18,000 $m^3$/s which is 1.34 times bigger than the original one. This is considered to be due to the accumulation of the reliable flood and storm event records after construction, and due to the increasing tendency of the local flood peaks according to the influence of world-wide weather change. The new estimation of the probable maximum precipitation (PMP) was based on the hydro-meteorological method suggested by the guideline of the World Meteorological Organization (WMO). The unit hydrograph which was applied for the estimation of PMF was derived through linear programming algorithm by minimizing the sum of absolute deviations of the calculated and recorded flood hydrographs. In order to adopt the newly estimated PMF as a design flood, following four alternatives were compared : (1) allocation of more flood control space by lowering the normal high water level, (2) construction of a new spillway in addition to the existing spillway, (3) construction of a new dam which has relevant flood control storage at the upstream of the Soyanggang dam, (4) raising the existing dam crest. The preliminary evaluation of these alternatives resulted in that the second alternative is most economic and feasible. So as to stably cope with the newly estimated PMF by meeting all the current functions of the multipurpose dam, a detailed study of an additional spillway tunnel has to be followed.

• Journal of the korean society of oceanography
• /
• v.32 no.4
• /
• pp.163-170
• /
• 1997

Some CMP gathers acquired from shallow marine seismic reflection survey in offshore Korea do not show the hyperbolic trend of moveout. It originated from so-called swell effect of source and streamer, which are towed under rough sea surface during the data acquisition. The observed time deviations of NMO-corrected traces can be entirely ascribed to the swell effect. To correct these time deviations, a residual statics is introduced using Genetic Algorithms (GA) into the swell correction. A new class of global optimization methods known as GA has recently been developed in the field of Artificial Intelligence and has a resemblance with the genetic evolution of biological systems. The basic idea in using GA as an optimization method is to represent a population of possible solutions or models in a chromosome-type encoding and manipulate these encoded models through simulated reproduction, crossover and mutation. GA parameters used in this paper are as follows: population size Q=40, probability of multiple-point crossover P$_c$=0.6, linear relationship of mutation probability P$_m$ from 0.002 to 0.004, and gray code representation are adopted. The number of the model participating in tournament selection (nt) is 3, and the number of expected copies desired for the best population member in the scaling of fitness is 1.5. With above parameters, an optimization run was iterated for 101 generations. The combination of above parameters are found to be optimal for the convergence of the algorithm. The resulting reflection events in every NMO-corrected CMP gather show good alignment and enhanced quality stack section.

• Journal of the Korean Society of Fisheries and Ocean Technology
• /
• v.27 no.4
• /
• pp.303-312
• /
• 1991

This paper describes the general formulation for the in-plane flexural free vibration analysis of two layered structure by the transfer influence coefficient method. The structure is regared as a distributed mass system with lumped mass and inertia moments, massless linear and rotational springs, and joints elements of releases and rolls at which the displacements are discontinuous in each layer. The results of the simple numerical examples on a personal computer demonstrate the validity of the present method, that is, the numerical high accuracy, the high speed, the flexibility for programming of the present algorithm, compared with the transfer matrix method.

• The KIPS Transactions:PartD
• /
• v.16D no.1
• /
• pp.11-26
• /
• 2009

In order to diagnose cardiovascular disease, we proposed EP-based(emerging pattern- based) classification technique using multi-parametric diagnosis indexes. We analyzed linear/nonlinear features of HRV for three recumbent postures and extracted four diagnosis indexes from ST-segments to apply the multi-parametric diagnosis indexes. In this paper, classification model using essential emerging patterns for diagnosing disease was applied. This classification technique discovers disease patterns of patient group and these emerging patterns are frequent in patients with cardiovascular disease but are not frequent in the normal group. To evaluate proposed classification algorithm, 120 patients with AP (angina pectrois), 13 patients with ACS(acute coronary syndrome) and 128 normal people data were used. As a result of classification, when multi-parametric indexes were used, the percent accuracy in classifying three groups was turned out to be about 88.3%.

• The KIPS Transactions:PartB
• /
• v.13B no.5 s.108
• /
• pp.525-532
• /
• 2006

This paper proposes a new class of image restoration on the Ising modeled binary 'Takbon' image by the flexible line-fitting ICM(Iterated conditional modes) method. Basically 'Takbon' image need be divided into two extreme regions, information and background one due to its stroke combinations. The main idea is the line process, comparing with the conventional ICM approaches which were based on partially rectangular structured point process. For calculating geometrical mechanism, we have defined line-fitting functions at each current pixel array which form the set of linear lines with gradients and lengths. By applying the Bayes' decision to this set, the region of the current pixel is decided as one of the binary levels. In this case, their statistical reiteration for distinct tracking between intra and extra region offers a criterion to decide the attachment at each step. Finally simulations using the binary 'Takbon' image are provided to demonstrate the effectiveness of our new algorithm