• Proceedings of the Korean Radioactive Waste Society Conference
• /
• 2005.06a
• /
• pp.464-473
• /
• 2005

Methodologies of NEWTRIT model, NRC model and AIRDOS-EPA model, which are off-site dose assessment models for regulatory compliance from routine releases of tritium into the environment, were investigated. Using the domestic data, if available, the predictive results of the models were compared. Among them, recently developed NEWTRIT model considers only doses from organically bounded tritium (OBT) due to environmental releases of tritiated water (HTO). A total dose from all exposure pathways predicted from AIRDOS-EPA model was 1.03 and 2.46 times higher than that from NEWTRIT model and NRC model, respectively. From above result, readers should not have an understanding that a predictive dose from NRC model may be underestimated compared with a realistic dose. It is because of that both mathematical models and corresponding parameter values for regulatory compliance are based on the conservative assumptions. For a dose by food consumption predicted from NEWTRIT model, the contribution of OBT was nearly equivalent to that of HTO due to relatively high consumption of grains in Korean. Although a total dose predicted from NEWTRIT model is similar to that from AIRDOS-EPA model, NEWTRIT model may be have a meaning in the understanding of phenomena for the behavior of HTO released into the environment.

• KIPS Transactions on Software and Data Engineering
• /
• v.9 no.5
• /
• pp.169-176
• /
• 2020

Recently, deep neural networks (DNNs) have brought revolutions to many mobile and embedded devices by providing human-level machine intelligence for various applications. However, high inference accuracy of such DNNs comes at high computational costs, and, hence, there have been significant efforts to reduce computational overheads of DNNs either by compressing off-the-shelf models or by designing a new small footprint DNN architecture tailored to resource constrained devices. One notable recent paradigm in designing small footprint DNN models is sharing parameters in several layers. However, in previous approaches, the parameter-sharing techniques have been applied to large deep networks, such as ResNet, that are known to have high redundancy. In this paper, we propose a parameter-sharing method for already parameter-efficient small networks such as ShuffleNetV2. In our approach, small templates are combined with small layer-specific parameters to generate weights. Our experiment results on ImageNet and CIFAR100 datasets show that our approach can reduce the size of parameters by 15%-35% of ShuffleNetV2 while achieving smaller drops in accuracies compared to previous parameter-sharing and pruning approaches. We further show that the proposed approach is efficient in terms of latency and energy consumption on modern embedded devices.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• v.10 no.11
• /
• pp.5476-5496
• /
• 2016

Energy efficiency of resource-constrained wireless sensor networks is critical in applications such as real-time monitoring/surveillance. To improve the energy efficiency and reduce the energy consumption, the time series data can be compressed before transmission. However, most of the compression algorithms for time series data were developed only for single variate scenarios, while in practice there are often multiple sensor nodes in one application and the collected data is actually multivariate time series. In this paper, we propose to compress the time series data by the Lasso (least absolute shrinkage and selection operator) approximation. We show that, our approach can be naturally extended for compressing the multivariate time series data. Our extension is novel since it constructs an optimal projection of the original multivariates where the best energy efficiency can be realized. The two algorithms are named by ULasso (Univariate Lasso) and MLasso (Multivariate Lasso), for which we also provide practical guidance for parameter selection. Finally, empirically evaluation is implemented with several publicly available real-world data sets from different application domains. We quantify the algorithm performance by measuring the approximation error, compression ratio, and computation complexity. The results show that ULasso and MLasso are superior to or at least equivalent to compression performance of LTC and PLAMlis. Particularly, MLasso can significantly reduce the smooth multivariate time series data, without breaking the major trends and important changes of the sensor network system.

• Journal of Radiation Protection and Research
• /
• v.25 no.4
• /
• pp.233-239
• /
• 2000

The sensitivity analysis of input parameters was Performed fer an ingestion dose assessment model (U. S. NRC's Regulatory Guide 1.109 model) from routine releases of radionuclides. In this study, three kinds of typical Korean foodstuffs (rice, leaff vegetables, milk) and two kinds of radionuclides $(^{l37}Cs,\;^{131}I)$ were considered. The values of input parameters were sampled using a Latin hypercube sampling technique based on Monte Carlo approach. Sensitivity indices, which represent the influence or the importance of input parameters for predictive results, were quantitatively expressed by the partial rank correlation coefficients. As the results, the ratio of the interception fraction to the yield of agricultural plants and the human consumption rate were sensitive input parameters for the considered foodstuffs and radionuclides. Additionally, in case of milk, the transfer factor of radionuclides from animal intake to milk and the daily intake rate of feedstuffs were sensitive input parameters. The weathering removal half-life and the delay time from food production to human consumption were relatively sensitive for $^{137}Cs$ and $^{131}I$ depositions, respectively.

• Journal of the Korean Institute of Electrical and Electronic Material Engineers
• /
• v.24 no.1
• /
• pp.81-88
• /
• 2011

A real time monitoring system from a PC has been developed which can be accessed through transmitted data, which incorporates an established low powered transport system equipped with a single chip combined with wireless sensor network technology from a three-axis acceleration sensor. In order to distinguish between static posture and dynamic posture, the extracted parameter from the rapidly transmitted data needs differentiation of movement and activity structures and status for an accurate measurement. When results interpret a static formation, statistics referring to each respective formation, known as the K-mean algorithm is utilized to carry out a determination of detailed positioning, and when results alter towards dynamic activity, fuzzy algorithm (fuzzy categorizer), which is the relationship between speed and ISVM, is used to categorize activity levels into 4 stages. Also, the ISVM is calculated with the instrumented acceleration speed on the running machine according to various speeds and its relationship with kinetic energy goes through correlation analysis. With the evaluation of the proposed system, the accuracy level stands at 100% at a static formation and also a 96.79% accuracy with kinetic energy and we can easily determine the energy consumption through the relationship between ISVM and kinetic energy.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.13 no.10
• /
• pp.4780-4785
• /
• 2012

As many mobile devices such as smart phones and tablets are widely spread, optimized mobile video encoder used during video recording application is needed. In this paper, we implemented H.264/AVC base profile video encoder on a mobile device and empirically optimized control parameters of the encoder. As the experiment, we more than 100 test cases were designed with varying Lagrangian optimization, Hadamard Transform, search range, I-frame period, and reference frames. During the experiment, we measured picture quality, bit-rate, encoding time, motion estimation time, and power consumption. From the result, we can determine optimal values for the H.264 control parameters.

• Fire Science and Engineering
• /
• v.34 no.2
• /
• pp.1-6
• /
• 2020

The energy released by various burning material has a wide range of its magnitude and transient characteristics, the measurement of the heat release rate(HRR) has been considered as one of the most challenging issue among the parameters related to fire. This study compares the measured HRR calculated by the oxygen consumption (OC) method and the carbon dioxide generation (CDG) method using a laboratory-scale fire calorimeter. The feasibility of the CDG method is examined by analyzing the relative error. The relationship between the oxygen depletion factor and CO₂ mass flow rate, which is a key parameter in HRR calculations, showed strong linearity at 6 % for the methane burner fire. The contribution of HRR by CO was less than 7% compared with the of HRR by CO₂ in the CDG calculation method. The linearity of the OC and CDG methods with respect to HRR of the referenced methane burner in a quasi-steady state was less than 1%; this indicates that the CDG method can be utilized as a complementary method in heat release rate measurement.

• Journal of the Korean Society of Fisheries and Ocean Technology
• /
• v.21 no.2
• /
• pp.157-166
• /
• 1985

The Purpose of this study are to investigate the characteristics of the flame behavior of gasoline-methanol blended fuels in spark ignition engine. Ionization probe were installed at the cylinder head and piston in order to measure flame speed. Other parameter such as engine performance, fuel consumption rate and exhaust gas were measured. The results were as follows. 1. In the case of increase methanol contents in blend fuel, flame propagation speed were increased, and thermal efficiency of the engine were increased due to decrease of energy consumption rate. 2. In the case of fixed equivalance ratio, NO sub(X) in exhaust gas were increased in accordance with increase of spark advance, and mean effective pressure were decreased in accordance with increase of methanol contents. 3. CO and HC concentration were decreased in accordance with increase of methanol contents.

• Journal of Microbiology and Biotechnology
• /
• v.16 no.12
• /
• pp.1890-1896
• /
• 2006

Lumostatic operation for cultivation of Haematococcus pluvialis was assessed to test the scale-up strategy of photobioreactors. Lumostatic operation is a method of maintaining a proper light condition based on the specific light uptake rate ($q_e$), by cells. Lumostatic operations were performed in 0.4-, 2-, 10-, and 30-1 scale bubble column photobioreactors and the results were compared with cultures illuminated with constant light intensity. Significant differences were observed in the maximal cell concentrations obtained from 0.4-, 2-, 10-, and 30-1 scale photobioreactors under constant light intensity, yielding the maximal cell concentrations of $2.8{\times}10^5$, $2.2\times10^5$, $1.5\times10^5$, and $1.1\times10^5$ cells/ml, respectively. The maximal cell concentration in a 0.4-1 photobioreactor under lumostatic operation was $4.3\times10^5$ cells/ml. Furthermore, those in 2-, 10-, and 30-1 scale photobioreactors were about the same as that in the 0.4-1 photobioreactor. The results suggest that lumostatic operation with proper $q_e$ is a good strategy for increasing the cell growth of Haematococcus pluvialis compared with a constant supply of light energy. Therefore, lumostatic operation is not only an efficient way to achieve high cell density cultures with minimal power consumption in microalgal cultures but it is also a perfect parameter for the scale-up of photobioreactors.

• Journal of the Korean Society for Precision Engineering
• /
• v.24 no.3 s.192
• /
• pp.108-116
• /
• 2007

To overcome many defects such as the high product cost, large energy consumption, and big space capacity in conventional mechanical machining, the miniaturization of machine tool and micro factory systems has been envisioned recently. The object of this paper is to research the effect of dynamic characteristic parameters in bolted-joint beams, which is widely applied to the joining of mechanical structures in order to identify structural system characteristics and to predict dynamic behavior according to scale-down from macro to micro system as the development of micro/meso-scale machine tool and micro factories. Modal parameters such as the natural frequency, damping ratio, and mode shape from modal testing and dynamic characteristics from finite element analysis are extracted with all 12 test beam models by materials, by size, and by joining condition, and then the results obtained by both methods are compared.