• Journal of the Korea Convergence Society
• /
• v.9 no.4
• /
• pp.47-55
• /
• 2018

In this paper, we propose a novel face recognition(FR) method that takes advantage of combining weighted deep local features extracted from multiple Deep Convolutional Neural Networks(DCNNs) learned with a set of facial local regions. In the proposed method, the so-called weighed deep local features are generated from multiple DCNNs each trained with a particular face local region and the corresponding weight represents the importance of local region in terms of improving FR performance. Our weighted deep local features are applied to Joint Bayesian metric learning in conjunction with Nearest Neighbor(NN) Classifier for the purpose of FR. Systematic and comparative experiments show that our proposed method is robust to variations in pose, illumination, and expression. Also, experimental results demonstrate that our method is feasible for improving face recognition performance.

• The KIPS Transactions:PartB
• /
• v.13B no.4 s.107
• /
• pp.429-438
• /
• 2006

Neural Networks(NNs) are applied for solving a wide variety of nonlinear problems in several areas, such as image processing, pattern recognition etc. Although NN can be simulated by using software, many potential NN applications required real-time processing. Thus they need to be implemented as hardware. The hardware implementation of multi-layer perceptrons(MLPs) in several kind of NNs usually uses a fixed-point arithmetic due to a simple logic operation and a shorter processing time compared to the floating-point arithmetic. However, the fixed-point arithmetic-based MLP has a drawback which is not able to apply the MLP software that use floating-point arithmetic. We propose a design method for MLPs which has the floating-point arithmetic-based fully-pipelining architecture. It has a processing speed that is proportional to the number of the hidden nodes. The number of input and output nodes of MLPs are generally constrained by given problems, but the number of hidden nodes can be optimized by user experiences. Thus our design method is using optimized number of hidden nodes in order to improve the processing speed, especially in field of a repeated processing such as image processing, pattern recognition, etc.

• Advances in concrete construction
• /
• v.10 no.6
• /
• pp.479-488
• /
• 2020

The experimental and numerical works were carried out on high performance fiber reinforced concrete (HPFRC) with w/cm ratios ranging from 0.25 to 0.40, fiber volume fraction (Vf)=0-1.5% and 10% silica fume replacement. Improvements in compressive and flexural strengths obtained for HPFRC are moderate and significant, respectively, Empirical equations developed for the compressive strength and flexural strength of HPFRC as a function of fiber volume fraction. A relation between flexural strength and compressive strength of HPFRC with R=0.78 was developed. Due to the complex mix proportions and non-linear relationship between the mix proportions and properties, models with reliable predictive capabilities are not developed and also research on HPFRC was empirical. In this paper due to the inadequacy of present method, a back propagation-neural network (BP-NN) was employed to estimate the 28-day compressive strength of HPFRC mixes. BP-NN model was built to implement the highly non-linear relationship between the mix proportions and their properties. This paper describes the data sets collected, training of ANNs and comparison of the experimental results obtained for various mixtures. On statistical analyses of collected data, a multiple linear regression (MLR) model with R2=0.78 was developed for the prediction of compressive strength of HPFRC mixes, and average absolute error (AAE) obtained is 6.5%. On validation of the data sets by NNs, the error range was within 2% of the actual values. ANN model has given the significant degree of accuracy and reliability compared to the MLR model. ANN approach can be effectively used to estimate the 28-day compressive strength of fibrous concrete mixes and is practical.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.22 no.4
• /
• pp.611-620
• /
• 2021

The maintenance industry is mainly progressing based on condition-based maintenance after corrective maintenance and preventive maintenance. In condition-based maintenance, maintenance is performed at the optimum time based on the condition of equipment. In order to find the optimal maintenance point, it is important to accurately understand the condition of the equipment, especially the remaining useful life. Thus, using simulation data (C-MAPSS), a prediction model is proposed to predict the remaining useful life of a turbofan engine. For the modeling process, a C-MAPSS dataset was preprocessed, transformed, and predicted. Data pre-processing was performed through piecewise RUL, moving average filters, and standardization. The remaining useful life was predicted using principal component analysis and the k-NN method. In order to derive the optimal performance, the number of principal components and the number of neighbor data for the k-NN method were determined through 5-fold cross validation. The validity of the prediction results was analyzed through a scoring function while considering the usefulness of prior prediction and the incompatibility of post prediction. In addition, the usefulness of the RUL prediction model was proven through comparison with the prediction performance of other neural network-based algorithms.

• Structural Monitoring and Maintenance
• /
• v.5 no.1
• /
• pp.51-65
• /
• 2018

A number of sensing techniques have been implemented for detecting defects in civil infrastructures instead of onsite human inspections in structural health monitoring. However, the issue of faults in sensors has not received much attention. This issue may lead to incorrect interpretation of data and false alarms. To overcome these challenges, this article presents a deep learning-based method with a new architecture of Stateful Long Short Term Memory Neural Networks (S-LSTM NN) for detecting sensor fault without going into details of the fault features. As LSTMs are capable of learning data features automatically, and the proposed method works without an accurate mathematical model. The detection of four types of sensor faults are studied in this paper. Non-stationary acceleration responses of a three-span continuous bridge when under operational conditions are studied. A deep network model is applied to the measured bridge data with estimation to detect the sensor fault. Another set of sensor output data is used to supervise the network parameters and backpropagation algorithm to fine tune the parameters to establish a deep self-coding network model. The response residuals between the true value and the predicted value of the deep S-LSTM network was statistically analyzed to determine the fault threshold of sensor. Experimental study with a cable-stayed bridge further indicated that the proposed method is robust in the detection of the sensor fault.

• Proceedings of the KIEE Conference
• /
• 2006.10a
• /
• pp.145-146
• /
• 2006

본 논문은 부분방전(PD: Partial Discharge)의 패턴인식 확률 극대화를 목적으로 신경망(NN: Neural Network) 파라미터 중에서 은닉층 뉴런의 수, 모멘텀(momentum)의 Step size와 Decay rate 를 최적화하기 위하여 유전 알고리즘(GA: Genetic Algonthm)을 적응하였다. 실험적 연구의 대상으로서, GIS(Gas Insulated Switchgear)사고의 주요 원인으로 보고되어있는 결함들을 인위적으로 모의한 16개 Test cell을 이용하여 부분방전을 발생시켰다. 부분방전 신호는 본 연구팀이 개발한 센서를 이용하여 검출되어 데이터베이스가 구축되어 그로부터 추출된 학습 데이터들의 학습에 다음과 같은 5가지 신경망 모델이 적응되었다: Multilayer Perception (MLP), Jordan-Elman Network (JEN), Recurrent Network (RN), Self-Organizing Feature Map (SOFM), Time-Lag Recurrent Network (TLRN). 유전 알고리즘 적용 효율성을 분석하기 위하여 동일한 데이터를 이용하여 다음과 같은 두 가지 방법을 적용한 결과를 상호 비교하였다. 우선 상기 선택된 모델만 적용하였고 다근 하나는 상기 모델과 Genetic Algorithm이 동시에 적용되었다. 모든 모델에 대하여 학습오차와 패턴 분류 확률을 비교한 결과, 유전 알고리즘 적응 시 부분방전 패턴인식 확률이 향상되었음이 확인되어 향후 신뢰성 있는 GIS 부분방전 진단기술에 활용될 수 있을 것으로 사료된다.

• Advances in concrete construction
• /
• v.13 no.5
• /
• pp.385-394
• /
• 2022

Recently, an increasing number of cutting-edged studies have shown that designing a smart active control for real-time implementation requires piles of hard-work criteria in the design process, including performance controllers to reduce the tracking errors and tolerance to external interference and measure system disturbed perturbations. This article proposes an effective artificial-intelligence method using these rigorous criteria, which can be translated into general control plants for the management of civil engineering installations. To facilitate the calculation, an efficient solution process based on linear matrix (LMI) inequality has been introduced to verify the relevance of the proposed method, and extensive simulators have been carried out for the numerical constructive model in the seismic stimulation of the active rigidity. Additionally, a fuzzy model of the neural network based system (NN) is developed using an interconnected method for LDI (linear differential) representation determined for arbitrary dynamics. This expression is constructed with a nonlinear sector which converts the nonlinear model into a multiple linear deformation of the linear model and a new state sufficient to guarantee the asymptomatic stability of the Lyapunov function of the linear matrix inequality. In the control design, we incorporated H Infinity optimized development algorithm and performance analysis stability. Finally, there is a numerical practical example with simulations to show the results. The implication results in the RMS response with as well as without tuned mass damper (TMD) of the benchmark building under the external excitation, the El-Centro Earthquake, in which it also showed the simulation using evolved bat algorithmic LMI fuzzy controllers in term of RMS in acceleration and displacement of the building.

• International Journal of Aeronautical and Space Sciences
• /
• v.12 no.1
• /
• pp.78-83
• /
• 2011

Traditional autopilot design requires an accurate aerodynamic model and relies on a gain schedule to account for system nonlinearities. This paper presents the control architecture applied to a dynamic model inversion at a single flight condition with an on-line neural network (NN) in order to regulate errors caused by approximate inversion. This eliminates the need for an extensive design process and accurate aerodynamic data. The simulation results using a developed full nonlinear 6 degree of freedom model are presented. This paper also presents the stability evaluation for control systems to which NNs were applied. Although feedback can accommodate uncertainty to meet system performance specifications, uncertainty can also affect the stability of the control system. The importance of robustness has long been recognized and stability margins were developed to quantify it. However, the traditional stability margin techniques based on linear control theory can not be applied to control systems upon which a representative non-linear control method, such as NNs, has been applied. This paper presents an alternative stability margin technique for NNs applied to control systems based on the system responses to an inserted gain multiplier or time delay element.

• Smart Structures and Systems
• /
• v.29 no.5
• /
• pp.677-691
• /
• 2022

Water cycle algorithm (WCA) has been a very effective optimization technique for complex engineering problems. This study employs the WCA for simultaneous prediction of heating load (L_H) and cooling load (L_C) in residential buildings. This algorithm is responsible for optimally tuning a neural network (NN). Utilizing 614 records, the behavior of the L_H and L_C is explored and the captured knowledge is then used to predict for 154 unanalyzed building conditions. Since the WCA is a population-based algorithm, different numbers of the searching agents were tested to find the most optimum configuration. It was observed that the best solution is discovered by 500 agents. A comparison with five newly-developed benchmark optimizers, namely equilibrium optimizer (EO), multi-tracker optimization algorithm (MTOA), slime mould algorithm (SMA), multi-verse optimizer (MVO), and electromagnetic field optimization (EFO) revealed that the WCANN predicts the desired parameters with considerably larger accuracy. Obtained root mean square errors (1.4866, 2.1296, 2.8279, 2.5727, 2.5337, and 2.3029 for the L_H and 2.1767, 2.6459, 3.1821, 2.9732, 2.9616, and 2.6890 for the L_C) indicated that the most reliable prediction was presented by the proposed model. The EFONN, however, provided a more time-effective solution. Lastly, an explicit predictive formula was elicited from the WCANN.

• Journal of Korea Game Society
• /
• v.6 no.2
• /
• pp.13-22
• /
• 2006

As the development of games continues, the intelligence of NPC is becoming more and more important. Nowadays, the NPCs of MMORPGS are not only capable of simple actions like moving and attacking players, but also utilizing variety of skills and tactics as human-players do. This study suggests a method that grants characters used in RPG(Role-Playing Game) an ability of training and adaptation using Neural network and Genetic Algorithm. In this study, a simple game-play model is constructed to test how suggested intellect characters could train and adapt themselves to game rules and tactics. In the game-play model, three types of characters(Tanker, Dealer, Healer) are used. Intellect character group constructed by NN and GA, and trained by combats against enemy character group constructed by FSM. As the result of test, the proposed intellect characters group acquire an appropriate combat tactics by themselves according to their abilities and those of enemies, and adapt change of game rule.