• Transactions of the Korean Society of Mechanical Engineers A
• /
• v.25 no.3
• /
• pp.537-547
• /
• 2001

The paper proposes the fuzzy logic global approximate optimization strategies in optimal sizing of automotive A-pillar trim with rib structures for occupant head protection. Two different strategies referred to as evolutionary fuzzy modeling (EFM) and neuro-fuzzy modeling (NFM) are implemented in the context of global approximate optimization. EFM and NFM are based on soft computing paradigms utilizing fuzzy systems, neural networks and evolutionary computing techniques. Such approximation methods may have their promising characteristics in a case where the inherent nonlinearity in analysis model should be accommodated over the entire design space and the training data is not sufficiently provided. The objective of structural design is to determine the dimensions of rib in A-pillar, minimizing the equivalent head injury criterion HIC(d). The paper describes the head-form modeling and head impact simulation using LS-DYNA3D, and the approximation procedures including fuzzy rule generation, membership function selection and inference process for EFM and NFM, and subsequently presents their generalization capabilities in terms of number of fuzzy rules and training data.

• Information Systems Review
• /
• v.12 no.2
• /
• pp.75-88
• /
• 2010

This paper proposed a method that segmented customers utilizing SOM(Self-organizing Map) and predicted the customers' response of a marketing promotion for each customer's segments. Our proposed method focused on predicting the response of customers dividing into customers' segment whereas most studies have predicted the response of customers all at once. We deployed logistic regression, neural networks, and support vector machines to predict customers' response that is a kind of dichotomous classification while the integrated approach was utilized to improve the performance of the prediction model. Sample data including 45 variables regarding demographic data about 600 customers, transaction data, and promotion activities were applied to the proposed method presenting classification matrix and the comparative analyses of each data mining techniques. We could draw some significant promotion strategies for segmented customers applying our proposed method to sample data.

• Phonetics and Speech Sciences
• /
• v.12 no.3
• /
• pp.55-63
• /
• 2020

Voice conversion can be applied to various voice processing applications. It can also play an important role in data augmentation for speech recognition. The conventional method uses the architecture of voice conversion with speech synthesis, with Mel filter bank as the main parameter. Mel filter bank is well-suited for quick computation of neural networks but cannot be converted into a high-quality waveform without the aid of a vocoder. Further, it is not effective in terms of obtaining data for speech recognition. In this paper, we focus on performing voice-to-voice conversion using only the raw spectrum. We propose a deep learning model based on the transformer network, which quickly learns the voice conversion properties using an attention mechanism between source and target spectral components. The experiments were performed on TIDIGITS data, a series of numbers spoken by an English speaker. The conversion voices were evaluated for naturalness and similarity using mean opinion score (MOS) obtained from 30 participants. Our final results yielded 3.52±0.22 for naturalness and 3.89±0.19 for similarity.

• Bulletin of the Korean Chemical Society
• /
• v.31 no.4
• /
• pp.808-814
• /
• 2010

Activities of nanostructure HZSM-5 and Co-ZSM-5 catalysts (with different Co-loading) for catalytic conversion of ethyl acetate and toluene were studied. The catalysts were prepared by wet impregnation method and were characterized by XRD, BET, SEM, TEM and ICP-AES techniques. Catalytic studies were carried out inside a U-shaped fixed bed reactor under atmospheric pressure and different temperatures. Toluene showed lower reactivity than ethyl acetate for conversion on Co-ZSM-5 catalysts. The effect of Co loading on conversion was prominent at temperatures below $400^{\circ}C$ and $450^{\circ}C$ for ethyl acetate and toluene respectively. In a binary mixture of organic compounds, toluene and ethyl acetate showed an inhibition and promotional behaviors respectively, in which the conversion of toluene was decreased at temperatures above $350^{\circ}C$. Inhibition effect of water vapor was negligible at temperatures above $400^{\circ}C$. An artificial neural networks model was developed to predict the conversion efficiency of ethyl acetate on Co-ZSM-5 catalysts based on experimental data. Predicted results showed a good agreement with experimental results. ANN modeling predicted the order of studied variable effects on ethyl acetate conversion, which was as follows: reaction temperature (50%) > ethyl acetate inlet concentration (25.085%) > content of Co loading (24.915%).

• Smart Structures and Systems
• /
• v.22 no.4
• /
• pp.413-424
• /
• 2018

The factors affecting the shear strength of the angle shear connectors in the steel-concrete composite beams can play an important role to estimate the efficacy of a composite beam. Therefore, the current study has aimed to verify the output of shear capacity of angle shear connector according to the input provided by Support Vector Machine (SVM) coupled with Firefly Algorithm (FFA). SVM parameters have been optimized through the use of FFA, while genetic programming (GP) and artificial neural networks (ANN) have been applied to estimate and predict the SVM-FFA models' results. Following these results, GP and ANN have been applied to develop the prediction accuracy and generalization capability of SVM-FFA. Therefore, SVM-FFA could be performed as a novel model with predictive strategy in the shear capacity estimation of angle shear connectors. According to the results, the Firefly algorithm has produced a generalized performance and be learnt faster than the conventional learning algorithms.

• The Journal of Korea Institute of Information, Electronics, and Communication Technology
• /
• v.13 no.4
• /
• pp.283-292
• /
• 2020

Artificial intelligence, which is one of the representative images of the 4th industrial revolution, has been highly recognized since 2016. This paper analyzed domestic paper trends for 'Artificial Intelligence', 'Machine Learning', and 'Deep Learning' among the domestic papers provided by the Korea Academic Education and Information Service. There are approximately 10,000 searched papers, and word count analysis, topic modeling and semantic network is used to analyze paper's trends. As a result of analyzing the extracted papers, compared to 2015, in 2016, it increased 600% in the field of artificial intelligence, 176% in machine learning, and 316% in the field of deep learning. In machine learning, a support vector machine model has been studied, and in deep learning, convolutional neural networks using TensorFlow are widely used in deep learning. This paper can provide help in setting future research directions in the fields of 'artificial intelligence', 'machine learning', and 'deep learning'.

• Proceedings of the Korean Society of Precision Engineering Conference
• /
• 2005.10a
• /
• pp.754-758
• /
• 2005

Optimization of process variables such as arc current, welding voltage and welding speed in terms of the weld characteristics desired is the key step in achieving high quality and improving performance characteristics without increasing the cost. Consequently, incorrect settings of those process variables give rise to deviations in the welding characteristics from the desired bead geometry. Therefore, trainee welders are referred to the tabulated information relating different metal types and thickness as to recommend the desired values of process variables. Basically, the bead geometry plays an important role in determining the mechanical properties of the weld. So that it is very important to select the process variables for obtaining optimal bead geometry. However, it is difficult for the traditional identification methods to provide an accurate model because the optimized welding process is non-linear and time-dependent. In this paper, the possibilities of the Infra-red sensor in sensing and control of the bead geometry in the automated welding process are presented. Infra-red sensor is a well-known method to deal with the problems with a high degree of fuzziness so that the sensor is employed to build the relationship between process variables and the quality characteristic the proposed above respectively. Based on several neural networks, the mathematical models are derived from extensive experiments with different welding parameters and complex geometrical features. The developed system enables to select the optimal welding parameters and control the desired weld dimensions during arc welding process.

• Proceedings of the Korea Water Resources Association Conference
• /
• 2019.05a
• /
• pp.363-363
• /
• 2019

활성화 함수(activation function)는 기계학습(machine learning)의 학습과정에 비선형성을 도입하여 심층적인 학습을 용이하게 하고 예측의 정확도를 높이는 중요한 요소 중 하나이다(Roy et al., 2019). 일반적으로 기계학습에서 사용되고 있는 활성화 함수의 종류에는 계단 함수(step function), 시그모이드 함수(sigmoid 함수), 쌍곡 탄젠트 함수(hyperbolic tangent function), ReLU 함수(Rectified Linear Unit function) 등이 있으며, 예측의 정확도 향상을 위하여 다양한 형태의 활성화 함수가 제시되고 있다. 본 연구에서는 기계학습을 통하여 수위예측 시 정확도 향상을 위하여 Hybrid 활성화 함수를 제안하였다. 연구대상지는 조수간만의 영향을 받는 한강을 대상으로 선정하였으며, 2009년 ~ 2018년까지 10년간의 수문자료를 활용하였다. 수위예측 알고리즘은 Python 내 Tensorflow의 RNN (Recurrent Neural Networks) 모델을 이용하였으며, 강수량, 수위, 조위, 댐 방류량, 하천 유량의 수문자료를 학습시켜 3시간 및 6시간 후의 수위를 예측하였다. 예측정확도 향상을 위하여 입력 데이터는 정규화(Normalization)를 시켰으며, 민감도 분석을 통하여 신경망모델의 은닉층 개수, 학습률의 최적 값을 도출하였다. Hybrid 활성화 함수는 쌍곡 탄젠트 함수와 ReLU 함수를 혼합한 형태로 각각의 가중치($w_1,w_2,w_1+w_2=1$)를 변경하여 정확도를 평가하였다. 그 결과 가중치의 비($w_1/w_2$)에 따라서 예측 결과의 RMSE(Roote Mean Square Error)가 최소가 되고 NSE (Nash-Sutcliffe model Efficiency coefficient)가 최대가 되는 지점과 Peak 수위의 예측정확도가 최대가 되는 지점을 확인할 수 있었다. 본 연구는 현재 Data modeling을 통한 수위예측의 정확도 향상을 위해 기초가 되는 연구이나, 향후 다양한 형태의 활성화 함수를 제안하여 정확도를 향상시킨다면 예측 결과를 통하여 침수예보에 대한 의사결정이 가능할 것으로 기대된다.

• Journal of the Korean Society of Propulsion Engineers
• /
• v.17 no.2
• /
• pp.71-83
• /
• 2013

Recently, the advanced condition monitoring methods such as the model-based method and the artificial intelligent method have been applied to maximize the availability as well as to minimize the maintenance cost of the aircraft gas turbines. Among them the non-linear GPA(Gas Path Analysis) method and the GA(Genetic Algorithms) have lots of advantages to diagnose the engines compared to other advanced condition monitoring methods such as the linear GPA, fuzzy logic and neural networks. Therefore this work applies both the non-linear GPA and the GA to diagnose AE3007 turbofan engine for an aircraft, and in case of having sensor noise and bias it is confirmed that the GA is better than the GPA through the comparison of two methods.

• The Journal of the Acoustical Society of Korea
• /
• v.40 no.5
• /
• pp.466-472
• /
• 2021

Speaker Change Detection (SCD) refers to finding the moment when the main speaker changes from one person to the next in a speech conversation. In speaker change detection, difficulties arise due to overlapping speakers, inaccuracy in the information labeling, and data imbalance. To solve these problems, TIMIT corpus widely used in speech recognition have been concatenated artificially to obtain a sufficient amount of training data, and the detection of changing speaker has performed after identifying overlapping speakers. In this paper, we propose an speaker change detection system that considers the speaker overlapping. We evaluated and verified the performance using various approaches. As a result, a detection system similar to the X-Vector structure was proposed to remove the speaker overlapping region, while the Bi-LSTM method was selected to model the speaker change system. The experimental results show a relative performance improvement of 4.6 % and 13.8 % respectively, compared to the baseline system. Additionally, we determined that a robust speaker change detection system can be built by conducting related studies based on the experimental results, taking into consideration text and speaker information.