• Fire Science and Engineering
• /
• v.32 no.3
• /
• pp.42-50
• /
• 2018

The effect of sub-grid turbulence and combustion models on the mean flame height in a heptane pool fire according to the Fire Dynamics Simulator (FDS) version (5 and 6) based on Large Eddy Simulation (LES) was examined. The heat release rate for the fire simulation was provided through experiments performed under identical conditions and the predictive performance of the mean flame height according to FDS version was evaluated by a comparison with the existing correlation. As a result, the Smagorinsky and Deardorff turbulence models applied to FDS 5 and 6, respectively, had no significant effects on the mean flow field, flame shape and flame height. On the other hand, the difference in pool fire characteristics including the mean flame height was due mainly to the difference in the mixture fraction and Eddy Dissipation Concept (EDC) combustion models applied to FDS 5 and 6, respectively. Finally, compared to FDS 6, FDS 5 provided the predictive result of a significantly longer flame height and more consistent mean flame height than the existing correlation.

• Fire Science and Engineering
• /
• v.15 no.2
• /
• pp.46-52
• /
• 2001

To evaluate the usability of compartment fire models for predicting sprinkler response time, fire experiment was conducted and measured sprinkler response time. The experimental data was compared with zone model "FASTLite"and field model "FDS"and field Model "SMARTFIRE" A Compartment fire conducted in a 2.4 m by 3.6 m by 2.4 m ISO 9705 room and measured H.R.R was approximately 100.3 kW. In test, Sprinkler activation temperature used is $72^{\circ}c$ and responded at 198s. The output of FASTLite, SMARTFIRE and, FDS for this fire scenario were 209s, 183s, and 192s, respectively. As a results, prediction using FDS model approached to that of test very closely and other models showed good approximated results also.

• Journal of the Korean Society of Safety
• /
• v.25 no.3
• /
• pp.22-27
• /
• 2010

A prediction performance of Fire Dynamics Simulator(FDS) developed by NIST for the diffusion flame structure was validated with experimental results of a laminar slot jet diffusion flame. Two mixture fraction combustion models and two finite chemistry combustion models were used in the FDS simulation for the validation of the jet diffusion flame structure. In order to enhance the prediction performance of flame structure, DNS and radiation model was applied to the simulation. The reaction rates of the finite chemistry combustion models were appropriately adjusted to the diffusion flame. The mixture fraction combustion model predicted the diffusion flame structure reasonably. A 1-step finite chemistry combustion model cannot predict the flame structure well, but the simulation results of a 2-step model were in good agreement with those of experiment except $CO_2$ concentration. It was identified that the 2-step model can be used in the investigation of flame suppression limit with further adjustment of reaction rates

• Fire Science and Engineering
• /
• v.32 no.1
• /
• pp.7-15
• /
• 2018

The prediction performance of combustion models in the Fire Dynamics Simulator (FDS) were evaluated by comparing with experiment for compartment propane gas fires. The mixture fraction model in the FDS v5.5.3 and Eddy Dissipation Concept (EDC) model in the FDS v6.6.3 were adopted in the simulations. Four chemical reaction mechanisms, such as 1-step Mixing Controlled, 2-step Mixing Controlled, 3-step Mixing Controlled and 3-step Mixed (Mixing Controlled + finite chemical reactions) reactions, were implemented in the EDC model. The simulation results with each combustion model showed similar level for the temperature inside the compartment. The prediction performance of FDS with each combustion model showed significant differences for the CO concentration while no distinguished differences were identified for the $O_2$ and $CO_2$ concentrations. The EDC 3-step Mixing Controlled largely over-predicted the CO concentration obtained by experiment and the mixture fraction model under-predicted the experiment slightly. The EDC 3-step Mixed showed the best prediction performance for the CO concentration and the EDC 2-step Mixing Controlled also predicted the CO concentration reasonably. The EDC 1-step Mixing Controlled significantly under-predict the experimental CO concentration when the previously suggested CO yield was adopted. The FDS simulation with the EDC 1-step Mixing Controlled showed difficulties in predicting the $CO_2$ concentration when the CO yield was modified to predict the CO concentration reasonably.

• Journal of the Korean Society of Safety
• /
• v.28 no.6
• /
• pp.6-10
• /
• 2013

A numerical reproducibility of the backdraft phenomena in a compartment was investigated. The prediction performance of two combustion models, the mixture fraction and finite chemistry models, were tested for the backdraft phenomena using the FDS code developed by the NIST. The mixture fraction model could not predict the flame propagation in a fuel-air mixture as well as the backdraft phenomena. However, the finite chemistry model predicted the flame propagation in the mixture inside a tube reasonably. In addition, the finite chemistry model predicted well the backdraft phenomena in a compartment qualitatively. The flame propagation inside the compartment, fuel and oxygen distribution and explosive fire ball behavior were well simulated with the finite chemistry model. It showed that the FDS adopted with the finite chemistry model can be an effective simulation tool for the investigation of backdraft in a compartment.

• Journal of the Korea Institute of Information Security & Cryptology
• /
• v.32 no.1
• /
• pp.25-38
• /
• 2022

With the diversification of payment methods and games, related financial accidents are causing serious problems for users and game companies. Recently, game companies have introduced an Fraud Detection System (FDS) for game payment systems to prevent financial incident. However, FDS is ineffective and cannot provide major evidence based on judgment results, as it requires constant change of detection patterns. In this paper, we analyze abnormal transactions among payment log data of real game companies to generate related features. One of the unsupervised learning models, Autoencoder, was used to build a model to detect abnormal transactions, which resulted in over 85% accuracy. Using X-FDS (Explainable FDS) with XAI-SHAP, we could understand that the variables with the highest explanation for anomaly detection were the amount of transaction, transaction medium, and the age of users. Based on X-FDS, we derive an improved detection model with an accuracy of 94% was finally derived by fine-tuning the importance of features that adversely affect the proposed model.

• Fire Science and Engineering
• /
• v.31 no.2
• /
• pp.44-51
• /
• 2017

Accurate predictions of the activation time for smoke detectors using a fire simulation is are required to ensure the reliability of the RSET (Required Safe Egress Time) calculation in the process of PBD (Performance-Based Design). The objective of this study was to enhance the accuracy of input parameters for the numerical models of smoke detector based on the FDS. To this end, a Fire Detector Evaluator (FDE) developed in previous studies was improved. The uniformities of flow and smoke inside the FDE were improved and accurate measurements of the obscuration per meter (OPM) related to detector operation were also performed through a decrease in the forward scattering of smoke particles. The input parameters using the improved FDE showed a significant difference from the previous FDE quantitatively. In particular, a larger difference was found in a photoelectric detector compared to an ionization detector. Considering that the operating conditions of smoke detectors are affected by the detector type, combustibles, smoke particulars, and color, the database (DB) on the input parameters for various detectors and combustibles should be built to improve the reliability of PBD in future studies.

• Proceedings of the Korea Institute of Fire Science and Engineering Conference
• /
• 2011.11a
• /
• pp.139-142
• /
• 2011

오픈 소스인 FDS(fire dynamic simulator)는 건물, 터널내의 화재나 연기, 열기류의 거동을 연구하기 위하여 국내외적으로 광범위하게 이용되고 있다. 소스코드 또한 공개 되어 있어 그 활용범위가 더욱 넓어지고 있다. 대부분의 기본적인 화재, 방재 해석을 위한 모델들을 제공하고 있으나 이를 더욱 발전시키고 새로운 알고리즘의 적용하기 위해서는 이러한 모델의 구조를 잘 이해할 필요가 있다. 본 연구에서는 이러한 FDS모델을 더욱 확장하기 위한 일환으로 현 FDS의 기본적인 구조를 검증모델(verification)을 이용하여 파악하고 이를 향후 소스코드를 확장할 수 있는 근간으로 삼고자 한다.

• Journal of the Korea Academia-Industrial cooperation Society
• /
• v.15 no.8
• /
• pp.4809-4814
• /
• 2014

The aim of the present study was to estimate the prediction performance of a FDS (Fire Dynamic Simulator) to simulate the fire behaviors and suppression characteristics by operating a water-mist. Rosin-Rammler/log-normal distribution function was used to determine the initial droplet distribution of water-mist and the effects of its model constant were considered. In addition, the simulation models were validated by a comparison of the predicted fire suppression characteristics with water-mist injection pressures to the previous experiments, and the thermal flow behaviors and gaseous concentration variations were analyzed. The results showed that water-mists with the same mean diameter were affected by the characteristics of the droplet size distribution, which have different size and velocity distributions at the downstream location. The fire simulations conducted in this study determine the initial droplet size distribution tuned to the base of the spray characteristics measured by previous experiments. The simulation results showed good agreement with the previous measurements for temperature variations and fire suppression characteristics. In addition, it was confirmed that the FDS simulation with a water-mist operation supplies useful details on estimations of the thermal flow fields and gaseous concentration under water mist operation conditions.

• Information Systems Review
• /
• v.22 no.1
• /
• pp.59-72
• /
• 2020

Artificial Intelligence is establishing itself as a familiar tool from an intractable concept. In this trend, financial sector is also looking to improve the problem of existing system which includes Fraud Detection System (FDS). It is being difficult to detect sophisticated cyber financial fraud using original rule-based FDS. This is because diversification of payment environment and increasing number of electronic financial transactions has been emerged. In order to overcome present FDS, this paper suggests 3 types of artificial intelligence models, Generative Adversarial Network (GAN), Deep Neural Network (DNN), and Convolutional Neural Network (CNN). GAN proves how data imbalance problem can be developed while DNN and CNN show how abnormal financial trading patterns can be precisely detected. In conclusion, among the experiments on this paper, WGAN has the highest improvement effects on data imbalance problem. DNN model reflects more effects on fraud classification comparatively.