Search

As neural networks are ubiquitous in the real-world environment, we encounter a variety of situations in which neural network-based systems fail. To address these failures, we need to identify their root causes. However, due to the black-box nature of neural networks, it is considered a difficult task to explain/understand internal functions, and cause analysis is usually conducted based on personal experience. To alleviate these problems, we propose a method to compute an element-wise contribution of inputs on current decisions. To this end, we focus on the physical meaning of neuron activations, and formulate an optimization problem that finds partial activations that support current decisions. Then, by accumulating these partial activations in the backward direction, we evaluate the contributions of each element in an input vector to the current decision. Experimental results have shown that the proposed method outperforms conventional methods in terms of cause localization for a range of failure scenarios.

BACKGROUND: Petroleum-contaminated soil from leaking above- and underground storage tanks and spillage during transport of petroleum products is widespread environmental problem in recent years. Application of compost may be the most promising, cost-effective, and eco-friendly technology for soil bioremediation because of its advantages over physical and chemical technology. The objective of this study was to evaluate effect of compost application on degradation of total petroleum hydrocarbon (TPH) in petroleum hydrocarbon-contaminated soil. METHOD AND RESULTS: An arable soil was artificially contaminated by diesel, and compost was applied at the different rate of 0, 10, 30, and 50 Mg/ha. Concentration of TPH in the soil decreased as application rate of compost increased. Degradation efficiency was highest at compost 30 Mg/ha; however, it slightly decreased with compost 50 Mg/ha. Kinetic modeling was performed to estimate the rates of chemical reaction. The correlation coefficient (R2) values for the linear plots using the second-order model were higher than those using the first-oder model. Compost 30 and 50 Mg/ha had the fastest TPH degradation rate in the second-order model. Change of microbial population in soil with compost application was similar to that of TPH. Microbial population in the soil increased as application rate of compost increased. Increasing microbial population in the contaminated soil corresponded to decreased in TPH concentration. CONCLUSION: Conclusively, compost application for soil bioremediation could be an effective response to petroleum hydrocarbon-contaminated soil. The increase in microbial population with compost suggested that compost application at an optimum rate might enhance degradation of TPH in soil.

The use of low-rank coal (physical and combustion characteristics are different from those of high-rank coal) is rapidly expanding both in Korea and globally owing to economic constraints; fully understanding the impact of low-rank coal on combustion and emission processes is thus imperative. Observations from several studies on low-rank coal are reviewed in this paper; these studies were conducted at Pusan clean coal center (PC3) with entrained flow reactors (laminar flow reactor and drop tube furnace). This study provides several contributions to the analysis of the combustion and emission characteristics of low-rank coal. Such contributions include the establishment of a simplified prediction model of burning coal and analysis of the effect of particle size and environmental conditions. A numerical model that incorporates the chemical percolation devolatilization sub-model is also developed, and the effect of volatile matter and in-furnace blending method is investigated. Results obtained from analyzing low rank-coal and the developed models provide insights into the application of low-rank coal in power plants.

Liming Effect on Cadmium Immobilization and Phytoavailability in Paddy Soil Affected by Mining Activity Chang Oh Hong, Yong Gyun Kim, Sang Mong Lee, Hyean Cheal Park, Keun Ki Kim, Hong Joo Son, Jae Hwan Cho and Pil Joo Kim (Department of Life Science and Environmental Biochemistry, Pusan National University, Miryang 627-706, South Korea, Department of Agricultural Economics, Pusan National University, Miryang 627-706, South Korea, Institute of Agriculture and Life Sciences, Gyeongsang National University, 900, Kaswa-dong, Jinju, 660-701, South Korea, Division of Applied Life Science, Graduate School (Brain Korea 21 Program), Gyeongsang National University, 900, Kaswa-dong, Jinju, 660-701, South Korea)

The influence of the volatile matter content on fuel NO emissions from a coal-burning drop-tube furnace (DTF) was analyzed by means of a numerical model for 25 different coals utilized by Korean power plants. The numerical results were fully validated against experimental results obtained using an actual DTF. The results show that the numerical model incorporating the chemical percolation devolatilization submodel is more accurate than that incorporating the two-competing-rates submodel to predict combustion and emission, particularly the devolatilization process and NO. Furthermore, the conversion ratio of fuel N to NO decreases with increasing volatile matter content because the volatile matter releases HCN, which furthers NO reduction. The destruction ratio increases with the amount of HCN produced during the devolatilization process. This paper indicates that HCN is the predominant intermediate species and the volatile matter content has a significant impact on NO emissions from the combustion of coal.

In this study, the effect of particle size on the combustion characteristics of pulverized sub-bituminous coal was experimentally investigated. A laminar-flow-entrained reactor was designed and implemented to realize the desired heating ratio and temperature corresponding to the combustion atmosphere of a pulverized-coal-fueled furnace. The flame length and structure of burning particles according to different sizes were investigated. Coal combustion processes were clearly distinguished by direct visual observation of the flame structure. The onset point of volatile ignition is greatly affected by changes in the particle size, and the burning time of the volatiles is least affected by changes in the particle size. The length and instability of char flame also increase with the increase of the particle size. However, the char consumption rate within the residential time remains nearly constant.

Since the concept of integrated product development environment (IPDE) was introduced from the US Department of Defense, the Korean navy has made efforts to apply the concept to its naval design and production. Recently, the Korean navy has supported a project for naval shipbuilding to achieve the advanced defense acquisition program. This project focuses on the integration of naval shipbuilding information. The Product Model Management System (PMMS) for Korean naval acquisition was a partial result of the project. The architecture of PMMS and the product structure of a naval ship are proposed in this paper. Also, the data acquisition process among acquisition program administration offices and contractors is defined for the construction of three-dimensional naval ship product models in PMMS. By constructing the product model on previous Korean naval ships, the PMMS was confirmed to a suitable innovative acquisition strategy. PMMS is on the leading edge of new product and process innovations in the Korean naval shipbuilding.

The current challenge with which most shipyards are faced is to reduce the design time and the time-to-delivery benefits of new information technologies such as ERP, APS, and SCM system. These technologies should integrate the information that flows in the enterprise wide activities. First of all we should maintain control of ship design and product data such as bill of material, product structure, drawing, 30 modeling and manufacturing information for the integration. The product data management (POM) system copes with these requirements. All the product data is stored in POM database and vaults in a structured manner and all the changes of product data are monitored, controlled and recorded. This study discusses how POM system can be implemented in shipyard. The procedure is surveyed to apply a POM system for the ship design data management. This study also presents several examples in order to make a practical guidance for ship POM implementation. Proposed studies can be useful for shipyards planning to adopt POM systems in enterprise wide manner.

Object tracking in a real time image is one of Interesting subjects in computer vision and many practical application fields past couple of years. But sometimes existing systems cannot find object by recognize background noise as object. This paper proposes a method of object detection and tracking using adaptive background image in real time. To detect object which does not influenced by illumination and remove noise in background image, this system generates adaptive background image by real time background image updating. This system detects object using the difference between background image and input image from camera. After setting up MBR(minimum bounding rectangle) using the internal point of detected object, the system tracks object through this MBR. In addition, this paper evaluates the test result about performance of proposed method as compared with existing tracking algorithm.

In Line Heating method, heating line is determined by kinematics analysis. But the heating line, which is solved by kinematics analysis, doesn`t verify by the point of physical analysis and the choice problem in many heating line doesn`t determine. In this paper, the simulator is developed. When we get the processing information at the kinematics analysis, simulator can estimate the shape of deformed plate that process along the processing information. When we get the initial shape and the object shape, we calculate the processing information first, using kinematics analysis. In a simulator, we estimate deformed shape from the processing information. After this we compare deformed shape and object shape. If the error of deformed shape and object shape is in the proper limits, that information is determined the final processing information. Else we repeat the process changing variables.

A protection algorithm for a wind turbine generator (WTG) operating in a large wind farm is described in this paper. To minimize the outage section, the protection should operate instantaneously for an internal fault or a connected power collection feeder fault, whilst remaining inoperative for an internal fault on a parallel WTG connected to the same feeder or an adjacent feeder fault. In addition, the protection should operate after a delay (i.e. as back-up) for an inter-tie fault or a grid fault. A fault on the parallel WTG or the adjacent feeder is discriminated based on the magnitude of the positive sequence current. To distinguish an internal WTG or connected feeder fault from an inter-tie or grid fault, the phase angle of the negative sequence current is used to determine the type of fault and then the magnitude of the positive sequence current is used to decide if the operating response should be instantaneous or delayed. The performance of the protection is verified under various fault scenarios based on EMTP-RV generated data. The results indicate that the algorithm can successfully distinguish faults that require instantaneous operation, delayed operation or non-operation.

Experimental investigation of the fluid flow and thermal flow pattern around two circular cylinders in cross flow has been carried out. Reynolds number was varied in the range of 100≦Re≦1000 and the distance between the two cylinders in the interval of 2≦L/D≦4. Velocity and turbulence intensity distributions in the isothermal distribution in the whole x−y thermal flow field was obtained by double-exposure holographic interferometry. The influences of Reynolds number and the gap spacing between the two cylinders were investigated. The characteristic flow pattern was found to depend on the distance between the two cylinders. WhenL≦3D, the wake region between the two cylinders became quasi-stationary.

A current differential relay with harmonic based blocking methods for transformers may maloperate for cases with very low harmonic contents in the operating current and its operation is delayed for the internal faults due to the harmonics. Thus, transformer model based protection techniques have been reported that rely on the electro-magnetic equations of a transformer. These methods require the separate values of the winding resistances and the leakage inductances. The former can be obtained experimentally whilst the latter cannot be obtained experimentally. This paper describes an estimation technique of the separate primary and secondary leakage inductances of a Y-Δ transformer. The method uses the line currents of the delta winding, which is available. The proposed technique can obtain the leakage inductances very accurately. The final paper will include the estimation results during the transient states such as magnetic inrush and over-excitation.

In shipbuilding industries, line heating has been used to fabricate plates of compound curvature. Line heating is a labor-intensive, time-consuming, and manual process carried out only by skilled workers. To enhance productivity as well as accuracy, it is imperative to automate such a process. A kinematic analysis has been introduced as a way to suggest a layout of computed heating lines. However, no verification of the resulting heating lines has been made either qualitatively or quantitatively. In this paper, an integrated system that performs the deformation simulation of plates with compound curvature is developed. This system generates optimal process information for the fabrication of curved plates using the line-heating process, by combining kinematic analysis, thermo-elasto-plastic analysis, and object comparison. The integrated system is implemented and verified by large-scale curved models used in shipyards.