• Structural Monitoring and Maintenance
• /
• v.6 no.3
• /
• pp.219-235
• /
• 2019

Scheduled inspections of common crossings are one of the main cost drivers of railway maintenance. Prognostics and health management (PHM) approach and modern monitoring means offer many possibilities in the optimization of inspections and maintenance. The present paper deals with data driven prognosis of the common crossing remaining useful life (RUL) that is based on an inertial monitoring system. The problem of scheduled inspections system for common crossings is outlined and analysed. The proposed analysis of inertial signals with the maximal overlap discrete wavelet packet transform (MODWPT) and Shannon entropy (SE) estimates enable to extract the spectral features. The relevant features for the acceleration components are selected with application of Lasso (Least absolute shrinkage and selection operator) regularization. The features are fused with time domain information about the longitudinal position of wheels impact and train velocities by multivariate regression. The fused structural health (SH) indicator has a significant correlation to the lifetime of crossing. The RUL prognosis is performed on the linear degradation stochastic model with recursive Bayesian update. Prognosis testing metrics show the promising results for common crossing inspection scheduling improvement.

• Smart Structures and Systems
• /
• v.30 no.3
• /
• pp.303-315
• /
• 2022

The degradation of wheel tread may result in serious hazards in the railway operation system. Therefore, timely wheel defect diagnosis of in-service trains to avoid tragic events is of particular importance. The focus of this study is to develop a novel wheel defect detection approach based on the relevance vector machine (RVM) which enables online detection of potentially defective wheels with trackside monitoring data acquired under different running-speed conditions. With the dynamic strain responses collected by a trackside monitoring system, the cumulative Fourier amplitudes (CFA) characterizing the effect of individual wheels are extracted to formulate multiple probabilistic regression models (MPRMs) in terms of multi-kernel RVM, which accommodate both variables of vibration frequency and running speed. Compared with the general single-kernel RVM-based model, the proposed multi-kernel MPRM approach bears better local and global representation ability and generalization performance, which are prerequisite for reliable wheel defect detection by means of data acquired under different running-speed conditions. After formulating the MPRMs, we adopt a Bayesian null hypothesis indicator for wheel defect identification and quantification, and the proposed method is demonstrated by utilizing real-world monitoring data acquired by an FBG-based trackside monitoring system deployed on a high-speed trial railway. The results testify the validity of the proposed method for wheel defect detection under different running-speed conditions.

• Journal of Information Processing Systems
• /
• v.18 no.2
• /
• pp.188-196
• /
• 2022

Recently, neural architecture search (NAS) has received increasing attention as it can replace human experts in designing the architecture of neural networks for different tasks and has achieved remarkable results in many challenging tasks. In this study, a path-based computation neural architecture encoder (PCE) was proposed. Our PCE first encodes the computation of information on each path in a neural network, and then aggregates the encodings on all paths together through an attention mechanism, simulating the process of information computation along paths in a neural network and encoding the computation on the neural network instead of the structure of the graph, which is more consistent with the computational properties of neural networks. We performed an extensive comparison with eight encoding methods on two commonly used NAS search spaces (NAS-Bench-101 and NAS-Bench-201), which included a comparison of the predictive capabilities of performance predictors and search capabilities based on two search strategies (reinforcement learning-based and Bayesian optimization-based) when equipped with different encoders. Experimental evaluation shows that PCE is an efficient encoding method that effectively ranks and predicts neural architecture performance, thereby improving the search efficiency of neural architectures.

• Proceedings of the Korea Water Resources Association Conference
• /
• 2022.05a
• /
• pp.347-347
• /
• 2022

최근 데이터 사이언스의 비약적인 발전과 함께 다양한 형태의 딥러닝 알고리즘이 개발되어 수자원 분야에도 적용되고 있다. 이 연구에서는 LSTM(Long Short-Term Memory) 네트워크와 BO-LSTM이라는 베이지안 최적화(BO) 기술을 결합하여 일단위 앙상블 미래 댐유입량을 projection하는 딥 러닝 모델을 제안하였다. BO-LSTM 하이퍼파라미터 및 손실 함수는 베이지안 최적화 기법을 통해 훈련 및 최적화되며, BO 접근법은 모델의 하이퍼파라미터와 손실 함수를 높은 정확도로 빠르게 최적화할 수 있었다(R=0.92 및 NSE=0.85). 또한 미래 댐 유입량을 예측하기 위한 LSTM의 구조는 Forecasting 모형과 Proiection 모형으로 구분하여 두 모형의 장단점을 분석하였으며, 본 연구의 결과로부터 데이터 처리 단계가 모델 훈련의 효율성을 높이고 노이즈를 줄이는 데 효과적이고 미래 예측에 있어 LSTM 구조에 따른 영향을 확인할 수 있었다. 본 연구는 소양강 유역, 2020-2100년 기간 동안의 미래 예측에 적용되었다. 전반적으로, CIMIP6 데이터에 따르면 10%에서 50%의 미래 유입량 증가가 발생하는 것으로 확인되었으며, 이는 미래 강수량의 증가의 폭과 유사함을 확인하였다. 유입량 산정에 있어 신뢰할 수 있는 예측은 저수지 운영, 계획 및 관리에 있어 정책 입안자와 운영자에게 도움이 될 것입니다.

• Journal of Intelligence and Information Systems
• /
• v.29 no.1
• /
• pp.107-119
• /
• 2023

The technology that recognizes a soldier's motion and movement status has recently attracted large attention as a combination of wearable technology and artificial intelligence, which is expected to upend the paradigm of troop management. The accuracy of state determination should be maintained at a high-end level to make sure of the expected vital functions both in a training situation; an evaluation and solution provision for each individual's motion, and in a combat situation; overall enhancement in managing troops. However, when input data is given as a timer series or sequence, existing feedforward networks would show overt limitations in maximizing classification performance. Since human behavior data (3-axis accelerations and 3-axis angular velocities) handled for military motion recognition requires the process of analyzing its time-dependent characteristics, this study proposes a high-performance data-driven classifier which utilizes the long-short term memory to identify the order dependence of acquired data, learning to classify eight representative military operations (Sitting, Standing, Walking, Running, Ascending, Descending, Low Crawl, and High Crawl). Since the accuracy is highly dependent on a network's learning conditions and variables, manual adjustment may neither be cost-effective nor guarantee optimal results during learning. Therefore, in this study, we optimized hyperparameters using Bayesian optimization for maximized generalization performance. As a result, the final architecture could reduce the error rate by 62.56% compared to the existing network with a similar number of learnable parameters, with the final accuracy of 98.39% for various military operations.

• Journal of KIISE:Software and Applications
• /
• v.30 no.3_4
• /
• pp.326-338
• /
• 2003

The objective of a neural network design and model selection is to construct an optimal network with a good generalization performance. However, training data include noises, and the number of training data is not sufficient, which results in the difference between the true probability distribution and the empirical one. The difference makes the teaming parameters to over-fit only to training data and to deviate from the true distribution of data, which is called the overfitting phenomenon. The overfilled neural network shows good approximations for the training data, but gives bad predictions to untrained new data. As the complexity of the neural network increases, this overfitting phenomenon also becomes more severe. In this paper, by taking statistical viewpoint, we proposed an integrative process for neural network design and model selection method in order to improve generalization performance. At first, by using the natural gradient learning with adaptive regularization, we try to obtain optimal parameters that are not overfilled to training data with fast convergence. By adopting the natural pruning to the obtained optimal parameters, we generate several candidates of network model with different sizes. Finally, we select an optimal model among candidate models based on the Bayesian Information Criteria. Through the computer simulation on benchmark problems, we confirm the generalization and structure optimization performance of the proposed integrative process of teaming and model selection.

• Journal of Korea Water Resources Association
• /
• v.54 no.11
• /
• pp.955-968
• /
• 2021

A water resource plan is routinely based on a natural flow and can be estimated using observed streamflow data or a long-term continuous rainfall-runoff model. However, the watershed with the natural flow is very limited to the upstream area of the dam. In particular, for the ungauged watershed, a rainfall-runoff model is established for the gauged watershed, and the model is then applied to the ungauged watershed by transferring the associated parameters. In this study, the GR4J rainfall-runoff model is mainly used to regionalize the parameters that are estimated from the 14 dam watershed via an optimization process. In terms of optimizing the parameters, the Bayesian approach was applied to consider the uncertainty of parameters quantitatively, and a number of parameter samples obtained from the posterior distribution were used for the regionalization. Here, the relationship between the estimated parameters and the topographical factors was first identified, and the dependencies between them are effectively modeled by a Copula function approach to obtain the regionalized parameters. The predicted streamflow with the use of regionalized parameters showed a good agreement with that of the observed with a correlation of about 0.8. It was found that the proposed regionalized framework is able to effectively simulate streamflow for the ungauged watersheds by the use of the regionalized parameters, along with the associated uncertainty, informed by the basin characteristics.

• Journal of Intelligence and Information Systems
• /
• v.25 no.2
• /
• pp.39-55
• /
• 2019

Recently banks and large financial institutions have introduced lots of Robo-Advisor products. Robo-Advisor is a Robot to produce the optimal asset allocation portfolio for investors by using the financial engineering algorithms without any human intervention. Since the first introduction in Wall Street in 2008, the market size has grown to 60 billion dollars and is expected to expand to 2,000 billion dollars by 2020. Since Robo-Advisor algorithms suggest asset allocation output to investors, mathematical or statistical asset allocation strategies are applied. Mean variance optimization model developed by Markowitz is the typical asset allocation model. The model is a simple but quite intuitive portfolio strategy. For example, assets are allocated in order to minimize the risk on the portfolio while maximizing the expected return on the portfolio using optimization techniques. Despite its theoretical background, both academics and practitioners find that the standard mean variance optimization portfolio is very sensitive to the expected returns calculated by past price data. Corner solutions are often found to be allocated only to a few assets. The Black-Litterman Optimization model overcomes these problems by choosing a neutral Capital Asset Pricing Model equilibrium point. Implied equilibrium returns of each asset are derived from equilibrium market portfolio through reverse optimization. The Black-Litterman model uses a Bayesian approach to combine the subjective views on the price forecast of one or more assets with implied equilibrium returns, resulting a new estimates of risk and expected returns. These new estimates can produce optimal portfolio by the well-known Markowitz mean-variance optimization algorithm. If the investor does not have any views on his asset classes, the Black-Litterman optimization model produce the same portfolio as the market portfolio. What if the subjective views are incorrect? A survey on reports of stocks performance recommended by securities analysts show very poor results. Therefore the incorrect views combined with implied equilibrium returns may produce very poor portfolio output to the Black-Litterman model users. This paper suggests an objective investor views model based on Support Vector Machines(SVM), which have showed good performance results in stock price forecasting. SVM is a discriminative classifier defined by a separating hyper plane. The linear, radial basis and polynomial kernel functions are used to learn the hyper planes. Input variables for the SVM are returns, standard deviations, Stochastics %K and price parity degree for each asset class. SVM output returns expected stock price movements and their probabilities, which are used as input variables in the intelligent views model. The stock price movements are categorized by three phases; down, neutral and up. The expected stock returns make P matrix and their probability results are used in Q matrix. Implied equilibrium returns vector is combined with the intelligent views matrix, resulting the Black-Litterman optimal portfolio. For comparisons, Markowitz mean-variance optimization model and risk parity model are used. The value weighted market portfolio and equal weighted market portfolio are used as benchmark indexes. We collect the 8 KOSPI 200 sector indexes from January 2008 to December 2018 including 132 monthly index values. Training period is from 2008 to 2015 and testing period is from 2016 to 2018. Our suggested intelligent view model combined with implied equilibrium returns produced the optimal Black-Litterman portfolio. The out of sample period portfolio showed better performance compared with the well-known Markowitz mean-variance optimization portfolio, risk parity portfolio and market portfolio. The total return from 3 year-period Black-Litterman portfolio records 6.4%, which is the highest value. The maximum draw down is -20.8%, which is also the lowest value. Sharpe Ratio shows the highest value, 0.17. It measures the return to risk ratio. Overall, our suggested view model shows the possibility of replacing subjective analysts's views with objective view model for practitioners to apply the Robo-Advisor asset allocation algorithms in the real trading fields.

• Journal of the Korean Institute of Intelligent Systems
• /
• v.14 no.2
• /
• pp.130-135
• /
• 2004

This paper presents a new design methods of the short-term load forecasting system (STLFS) using the data mining. The structure of the proposed STLFS is divided into two parts: the Takagi-Sugeno (T-S) fuzzy model-based classifier and predictor The proposed classifier is composed of the Gaussian fuzzy sets in the premise part and the linearized Bayesian classifier in the consequent part. The related parameters of the classifier are easily obtained from the statistic information of the training set. The proposed predictor takes form of the convex combination of the linear time series predictors for each inputs. The problem of estimating the consequent parameters is formulated by the convex optimization problem, which is to minimize the norm distance between the real load and the output of the linear time series estimator. The problem of estimating the premise parameters is to find the parameter value minimizing the error between the real load and the overall output. Finally, to show the feasibility of the proposed method, this paper provides the short-term load forecasting example.

• Journal of Korea Water Resources Association
• /
• v.54 no.8
• /
• pp.617-628
• /
• 2021

Drought is a major natural disaster that causes serious social and economic losses. Local drought forecasts can provide important information for drought preparedness. In this study, we propose a new machine learning model that predicts drought by using historical drought indices and meteorological data from 10 sites from 1981 to 2020 in the southeastern part of the Korean Peninsula, Busan-Ulsan-Gyeongnam. Using Bayesian optimization techniques, a hyper-parameter-tuned Random Forest, XGBoost, and Light GBM model were constructed to predict the evaporative demand drought index on a 6-month time scale after 1-month. The model performance was compared by constructing a single site model and a regional model, respectively. In addition, the possibility of improving the model performance was examined by constructing a fine-tuned model using data from a individual site based on the regional model.