• Nuclear Engineering and Technology
• /
• 제46권6호
• /
• pp.753-772
• /
• 2014

The paper presents functional modelling and its application for diagnosis in nuclear power plants. Functional modelling is defined and its relevance for coping with the complexity of diagnosis in large scale systems like nuclear plants is explained. The diagnosis task is analyzed and it is demonstrated that the levels of abstraction in models for diagnosis must reflect plant knowledge about goals and functions which is represented in functional modelling. Multilevel flow modelling (MFM), which is a method for functional modelling, is introduced briefly and illustrated with a cooling system example. The use of MFM for reasoning about causes and consequences is explained in detail and demonstrated using the reasoning tool, the MFMSuite. MFM applications in nuclear power systems are described by two examples: a PWR; and an FBR reactor. The PWR example show how MFM can be used to model and reason about operating modes. The FBR example illustrates how the modelling development effort can be managed by proper strategies including decomposition and reuse.

• KSII Transactions on Internet and Information Systems (TIIS)
• /
• 제15권4호
• /
• pp.1538-1552
• /
• 2021

Machine learning models are vulnerable to adversarial examples generated by adding a deliberately designed perturbation to a benign sample. Particularly, for automatic speech recognition (ASR) system, a benign audio which sounds normal could be decoded as a harmful command due to potential adversarial attacks. In this paper, we focus on the countermeasures against audio adversarial examples. By analyzing the characteristics of ASR systems, we find that frame offsets with silence clip appended at the beginning of an audio can degenerate adversarial perturbations to normal noise. For various scenarios, we exploit frame offsets by different strategies such as defending, detecting and hybrid strategy. Compared with the previous methods, our proposed method can defense audio adversarial example in a simpler, more generic and efficient way. Evaluated on three state-of-the-arts adversarial attacks against different ASR systems respectively, the experimental results demonstrate that the proposed method can effectively improve the robustness of ASR systems.

• ETRI Journal
• /
• 제44권5호
• /
• pp.794-804
• /
• 2022

Transformer models have had a great impact on natural language processing (NLP) in recent years by realizing outstanding and efficient contextualized language models. Recent studies have used transformer-based language models for various NLP tasks, including Persian named entity recognition (NER). However, in complex tasks, for example, NER, it is difficult to determine which contextualized embedding will produce the best representation for the tasks. Considering the lack of comparative studies to investigate the use of different contextualized pretrained models with sequence modeling classifiers, we conducted a comparative study about using different classifiers and embedding models. In this paper, we use different transformer-based language models tuned with different classifiers, and we evaluate these models on the Persian NER task. We perform a comparative analysis to assess the impact of text representation and text classification methods on Persian NER performance. We train and evaluate the models on three different Persian NER datasets, that is, MoNa, Peyma, and Arman. Experimental results demonstrate that XLM-R with a linear layer and conditional random field (CRF) layer exhibited the best performance. This model achieved phrase-based F-measures of 70.04, 86.37, and 79.25 and word-based F scores of 78, 84.02, and 89.73 on the MoNa, Peyma, and Arman datasets, respectively. These results represent state-of-the-art performance on the Persian NER task.

• ETRI Journal
• /
• 제22권2호
• /
• pp.1-10
• /
• 2000

Pattern recognition is one of the most common problems encountered in engineering and scientific disciplines, which involves developing prediction or classification models from historic data or training samples. This paper introduces a new approach, called the Representational Capability (RC) algorithm, to handle pattern recognition problems using radial basis function (RBF) models. The RC algorithm has been developed based on the mathematical properties of the interpolation and design matrices of RBF models. The model development process based on this algorithm not only yields the best model in the sense of balancing its parsimony and generalization ability, but also provides insights into the design process by employing a design parameter (${\delta}$). We discuss the RC algorithm and its use at length via an illustrative example. In addition, RBF classification models are developed for heart disease diagnosis.

• Asia pacific journal of information systems
• /
• 제6권2호
• /
• pp.125-148
• /
• 1996

This paper is primarily focused on the function of model management systems such as higher level representations and buildings of optimization models using them, especially in the area of the telecommunication network models. This research attempts to provide the model builders an intuitive language-namely higher level representation-using five distinctivenesses : Objective, Node, Link, Topological Constraint including five components, and Decision. The paper elaborates all components included in each of distinctivenesses extracted from structural characteristics of typical telecommunication network models. Higher level representations represented with five distinctivenesses should be converted into base level representations which are employed for semantic representations of linear and integer programming problems in knowledge: assisted optimization modeling system(UNIK-OPT). Furthermore, for formulating the network model using higher level representations, the reasoning process is proposed. A system called UNIK-NET is developed to implement the approach proposed in this research, and the system is illustrated with an example of the network model.

• 한국경영과학회지
• /
• 제23권2호
• /
• pp.103-129
• /
• 1998

Corporate level decision making with multiple decision makers in a consistent way is essential in Decision Support System. However, since the decision makers have different interests and knowledge, the models used by them are also different in their level of abstraction. This makes decision makers waste a lot of efforts for an integrated decision making. The purpose of this paper is to propose an integration mechanism so that collaborative decision making models may be used synthetically in multi-abstraction level. Models are classified as multimedia model, mathematical model, qualitative model, causal ＆ directional model, causal model, directional model and relationship model according to the level of abstraction. The proposed integration mechanism consists of model interpretation phase. model transformation phase, and model integration phase. Specifically, the model transformation Phase is divided into (1) model tightening mode which gather information to make a model transformed into upper level model, and (2) model relaxing mode which makes lower level model. In the model integration phase, models of same level are to be integrated schematically. An illustrative M＆A-decision example is given to show the possibility of the methodology.

• 한국정밀공학회지
• /
• 제19권8호
• /
• pp.71-76
• /
• 2002

In this study low cycle fatigue (LCF) behavior of 12Cr steel at high temperature are described. Secondly, comparisons between predicted lives and experimental lives are made for the several sample life prediction models. Two minute hold period in either tension or compression reduce the number of cycles to failure by about a factor of two. Twenty minute hold periods in compression lead to shorter lives than 2 minute hold periods in compression. Experiments showed that life predictions from classical phenomenological models have limitations. More LCF experiments should be pursued to gain understanding of the physical damage mechanisms and to allow the development of physically-based models which can enhance the accuracy of the predictions of components. From a design point-of-view, life prediction has been judged acceptable for these particular loading conditions but extrapolations to thermo-mechanical fatigue loading, for example, require more sophisticated models including physical damage mechanisms.

• Journal of the Korean Data and Information Science Society
• /
• 제28권3호
• /
• pp.685-695
• /
• 2017

The traditional mixture of experts (ME) modeled the gate network using a certain parametric function. However, if the assumed parametric function does not properly reflect the true nature, the prediction strength of ME would become weak. For example, the parametric ME often uses logistic or multinomial logistic models for the network model. However, this could be very misleading if the true nature of the data is quite different from those models. Although, in this case, we may develop more flexible parametric models by extending the model at hand, we will never be free from such misspecification problems. In order to alleviate such weakness of the parametric ME, we propose to use the semi-parametric mixture of experts (SME) in which the gate network is estimated in a non-parametrical way. Based on this, we compared the performance of the SME with those of ME and neural networks via several simulation experiments and real data examples.

• 한국시뮬레이션학회논문지
• /
• 제8권4호
• /
• pp.9-24
• /
• 1999

We describe a heterogeneous simulation framework, so called DEVS-HLA, in which conventional simulation models and the DEVS (Discrete Event System Specification) models are interoperable. DEVS-HLA conceptually consists of three layers: model layer, DEVS BUS layer, and HLA (High Level Architecture) layer. The model layer has a collection of heterogeneous simulation models, such as DEVS, CSIM, SLAM, and so on, to represent various aspects of a complex system. The DEVS BUS layer provides a virtual software bus, DEVS BUS, so that such simulation models can communicate with each other. Finally, the HLA layer is employed as a communication infrastructure, which supports several good features for distributed simulation. The DEVS BUS has been implemented on the HLA/RTI (Run-Time Infrastructure) and a simple example of a flexible manufacturing system has been developed to validate the DEVS-HLA.

• 천문학회보
• /
• 제42권1호
• /
• pp.31.2-31.2
• /
• 2017

The central regions of barred-spiral galaxies contain interesting gaseous structures such as dust lanes and nuclear rings with intense star formation. While our previous studies were useful in understanding the formation of these structures star formation history, they were limited to 2D isothermal galaxies in which the stellar disk and halo are modeled by fixed gravitational potentials. To study the effects of bar growth as well as the vertical dimension, we use the mesh-free hydrodynamic code named GIZMO and run 3D simulations by treating the stellar disk and halo as being live. We find that the new 3D models form the gaseous features similarly to the previous 2D models, although the detailed formation processes are quite different. For example, a ring has a large radius when it first forms and shrinks over time in the previous 2D models. In the 3D live-potential models, however, a ring forms small and grows in size with time. We present the results of the new simulations and discuss them in comparison with the previous 2D results.