• Smart Structures and Systems
• /
• v.22 no.2
• /
• pp.185-192
• /
• 2018

Participants in the Asia Pacific Conference of the Prognostics and Health Management Society 2017 (PHMAP 2017) Data Challenge were given measured vibration signals from motor-driven gearboxes used in pulverizers. Using this information, participants were requested to predict failure dates and the faulty components. The measured signals were affected by significant noise and disturbance, as the pulverizers in the provided data worked under actual operating conditions. This paper thus presents a fault prediction method for a motor-driven gearbox in a pulverizer system that can perform under external noise and disturbance conditions. First, two fault features, an RMS value in the higher frequency zones (HRMS) and an amplitude of a period for high-speed shaft in the quefrency domain ($QA_{HSS}$), were extracted based on frequency analysis using the higher and lower sampling rate data. The two features were then applied to each pulverizer based on results of frequency responses to impact loadings. Then, a regression analysis was used to predict the failure date using the two extracted features. A weighted regression analysis was used to compensate for the imbalance of the features in the given period. In addition, the faulty components in the motor-driven gearboxes were predicted based on the modulated frequency components. The score predicted by the proposed approach was ranked first in the PHMAP 2017 Data Challenge.

• Journal of the Korean Operations Research and Management Science Society
• /
• v.34 no.2
• /
• pp.35-54
• /
• 2009

The recent economic crisis not only reduces the profit of department stores but also incurs the significance losses caused by the increasing late-payment rate of credit cards. Under this pressure, the scope of credit prediction needs to be broadened from the simple prediction of whether this customer has a good credit or not to the accurate prediction of how much profit can be gained from this customer. This study classifies the delinquent customers of credit card in a Korean department store into homogeneous clusters. Using this information, this study analyzes the repayment patterns for each cluster and develops the credit prediction system to manage the delinquent customers. The model presented by this study uses Kohonen network, which is one of artificial neural networks of data mining technique, to cluster the credit delinquent customers into clusters. Cox proportional hazard model is also used, which is one of survival analysis used in medical statistics, to analyze the repayment patterns of the delinquent customers in each cluster. The presented model estimates the repayment period of delinquent customers for each cluster and introduces the influencing variables on the repayment pattern prediction. Although there are some differences among clusters, the variables about the purchasing frequency in a month and the average number of installment repayment are the most predictive variables for the repayment pattern. The accuracy of the presented system leaches 97.5%.

• The Journal of the Korea Contents Association
• /
• v.18 no.8
• /
• pp.525-535
• /
• 2018

If the frequency of a particular class is excessively higher than the frequency of other classes in the classification problem, data imbalance problems occur, which make machine learning distorted. Corporate bankruptcy prediction often suffers from data imbalance problems since the ratio of insolvent companies is generally very low, whereas the ratio of solvent companies is very high. To mitigate these problems, it is required to apply a proper sampling technique. Until now, oversampling techniques which adjust the class distribution of a data set by sampling minor class with replacement have popularly been used. However, they are a risk of overfitting. Under this background, this study proposes ROSE(Random Over Sampling Examples) technique which is proposed by Menardi and Torelli in 2014 for the effective corporate bankruptcy prediction. The ROSE technique creates new learning samples by synthesizing the samples for learning, so it leads to better prediction accuracy of the classifiers while avoiding the risk of overfitting. Specifically, our study proposes to combine the ROSE method with SVM(support vector machine), which is known as the best binary classifier. We applied the proposed method to a real-world bankruptcy prediction case of a Korean major bank, and compared its performance with other sampling techniques. Experimental results showed that ROSE contributed to the improvement of the prediction accuracy of SVM in bankruptcy prediction compared to other techniques, with statistical significance. These results shed a light on the fact that ROSE can be a good alternative for resolving data imbalance problems of the prediction problems in social science area other than bankruptcy prediction.

• International Journal of Fluid Machinery and Systems
• /
• v.6 no.2
• /
• pp.87-93
• /
• 2013

In order to study the effect of the fluid-structure interaction (FSI) on the simulation results, the external characteristics and internal flow features of a diffuser pump were analyzed with a two-way flow solid coupling method. And the static and dynamic structure analysis of the blade was also caculated with the FEA method. The steady flow field is based on Reynolds Averaged N-S equations with standard $k-{\varepsilon}$ turbulent model, the unsteady flow field is based on the large eddy simulation, and the structure response is based on elastic transient structural dynamic equation. The results showed that the effect of FSI on the head prediction based on CFD really exists. At the same radius, the van mises stress on the nodes closed shroud and hub was larger than other nodes. A large deformation region existed near inlet side at the middle of blades. The strength of impeller satisfied the strength requirement with static stress analysis based on the fourth strength theory. The dynamic stress varied periodically with the impeller rotating. It was also found that the fundamental frequency of the dynamic stress is the rotating frequency and its harmonic frequency. The frequency of maximum stress amplitude at node 1626 was 7 times of the rotating frequency. The frequency of maximum stress amplitude at node 2328 was 14 times of the rotating frequency. No matter strength failure or fatigue failure, the root of blades near shroud is the key region to analyse.

• Journal of Korean Society of Water and Wastewater
• /
• v.35 no.1
• /
• pp.83-91
• /
• 2021

The increased turbidity in rivers during flood events has various effects on water environmental management, including drinking water supply systems. Thus, prediction of turbid water is essential for water environmental management. Recently, various advanced machine learning algorithms have been increasingly used in water environmental management. Ensemble machine learning algorithms such as random forest (RF) and gradient boosting decision tree (GBDT) are some of the most popular machine learning algorithms used for water environmental management, along with deep learning algorithms such as recurrent neural networks. In this study GBDT, an ensemble machine learning algorithm, and gated recurrent unit (GRU), a recurrent neural networks algorithm, are used for model development to predict turbidity in a river. The observation frequencies of input data used for the model were 2, 4, 8, 24, 48, 120 and 168 h. The root-mean-square error-observations standard deviation ratio (RSR) of GRU and GBDT ranges between 0.182~0.766 and 0.400~0.683, respectively. Both models show similar prediction accuracy with RSR of 0.682 for GRU and 0.683 for GBDT. The GRU shows better prediction accuracy when the observation frequency is relatively short (i.e., 2, 4, and 8 h) where GBDT shows better prediction accuracy when the observation frequency is relatively long (i.e. 48, 120, 160 h). The results suggest that the characteristics of input data should be considered to develop an appropriate model to predict turbidity.

• The Transactions of the Korean Institute of Electrical Engineers P
• /
• v.61 no.4
• /
• pp.174-179
• /
• 2012

The social and economic level of damages might be highly increased in the case of wide-area black-outages, because of heavy dependence of electricity. Therefore, the development of a wide-area protection relay intelligence techniques is required to prevent massive power outages and minimize the impact strength at failure. The frequency monitoring and prediction for wide-area protection relaying intelligence has been considered as an important technology. In this paper, a network-based frequency monitoring system developed for wide-area protection relay intelligence is presented. In addition, conventional techniques for frequency estimation are compared, and a method for advanced frequency estimation and measurement to improve the precision is proposed. Finally, an integrated monitoring system called K-FNET(Korea-Frequency Monitoring Network) is implemented based on the GPS and various energy monitoring cases are studied.

• Journal of Advanced Navigation Technology
• /
• v.16 no.5
• /
• pp.794-800
• /
• 2012

In this paper, we propose a scheme for improving not only the utilization of frequency bands in the guard channel scheme but also the dropping rate of cognitive radio user in the wireless cognitive radio-based cellular network. The proposed scheme enables cognitive radio users to utilize the guard channel for servicing only handoff calls in normal times, but cognitive radio users must vacate the frequency channel when handoff call appearing. At this time our scheme ensures their seamless services for cognitive radio users, by predicting handoff call's appearance by MMOSPRED (Multi-Media One Step Prediction) method and then reserving the demanded channels for spectrum handoff calls. Our simulations show that our scheme performs better than other schemes; GCS(Guard Channel Scheme) and a scheme without prediction in terms of cognitive users call's dropping rate and resource utilization efficiency.

• The KIPS Transactions:PartA
• /
• v.16A no.5
• /
• pp.347-356
• /
• 2009

Precise branch prediction is a critical factor in the IPC Improvement of modern microprocessor architectures. In addition to the branch prediction accuracy, branch prediction delay have a profound impact on overall system performance as well. However, it tends to be overlooked when the architects design the branch predictor. To tolerate branch prediction delay, this paper proposes Early Start Prediction (ESP) technique. The proposed solution dynamically identifies the start instruction of basic block, called as Basic Block Start Address (BB_SA), and the solution uses BB_SA when predicting the branch direction, instead of branch instruction address itself. The performance of the proposed scheme can be further improved by combining short interval hiding technique between BB_SA and branch instruction. The simulation result shows that the proposed solution hides prediction latency, with providing same level of prediction accuracy compared to the conventional predictors. Furthermore, the combination with short interval hiding technique provides a substantial IPC improvement of up to 10.1%, and the IPC is actually same with ideal branch predictor, regardless of branch predictor configurations, such as clock frequency, delay model, and PHT size.

• Bulletin of the Korean Chemical Society
• /
• v.34 no.4
• /
• pp.1043-1046
• /
• 2013

Prediction of physicochemical properties of organic molecules is an important process in chemistry and chemical engineering. The MSEP approach developed in our lab calculates the molecular surface electrostatic potential (ESP) on van der Waals (vdW) surfaces of molecules. This approach includes geometry optimization and frequency calculation using hybrid density functional theory, B3LYP, at the 6-31G(d) basis set to find minima on the potential energy surface, and is known to give satisfactory QSPR results for various properties of organic molecules. However, this MSEP method is not applicable to screen large database because geometry optimization and frequency calculation require considerable computing time. To develop a fast but yet reliable approach, we have re-examined our previous work on organic molecules using two semi-empirical methods, AM1 and PM3. This new approach can be an efficient protocol in designing new molecules with improved properties.

• Journal of the Semiconductor & Display Technology
• /
• v.18 no.1
• /
• pp.123-128
• /
• 2019

Currently mobile services are on the verge of full commercialization ahead of 5G mobile communication (5G). The first goal could be to preempt the 5G market through realistic media services utilizing VR (Virtual Reality) and AR (Augmented Reality) technologies that users can most easily experience. Basically this movement is based on the advanced development of smart devices and high quality graphics processing computing power of mobile application processors. Accordingly, the importance of mobile GPUs is emerging and the most concern issue becomes a model for predicting the power and performance for smooth operation of high quality mobile contents. In many cases, the performance of mobile GPUs has been introduced in terms of power consumption of mobile GPUs using dynamic voltage and frequency scaling and throttling functions for power consumption and heat management. This paper introduces several studies of mobile GPU performance prediction model with user-friendly methods not like conventional power centric performance prediction models.