Earticle

This study explored the possibility of supplementing the existing qualitative analysis method by applying artificial intelligence (AI) technology to the fire investigation field. By applying and analyzing ChatGPT and Circle-to-Search functions to the actual fire case of Nonsan Fire Station, it showed high accuracy and efficiency in tasks such as fire cause prediction, fire pattern classification, and field photo analysis. In particular, automatic analysis of visual evidence through image-based neural network model and automatic summary and classification of reports based on natural language processing suggested the possibility of greatly improving the speed and objectivity of fire investigation. In addition, by constructing the Vision AI prototype model, meaningful results were obtained to identify oil patterns on the floor, and a systematic fire investigation system was prepared by establishing an NSFI shared system. Lastly, the deployment of data collection experts to the fire investigation team was proposed.

This paper studies a technique for estimating fire patterns by analyzing the amount of change in material components. The background of the experiment is to find a method that can help people with less than two years of fire investigation practice find the ignition point. A total of 6 types of test samples were selected, including 4 types of steel plates, a front door collected from a fire site, and a brick building After heating the steel plate at a temperature of 700℃ for 1800s on a sample of 700mm in width and 700mm in length, An experiment was conducted to estimate the Fire pattern at the ignition point by analyzing the component change of the surface in the non-irrigated and watered state, and objective verification was conducted on the field applicability of this study using samples collected from the fire site. For each sample, changes such as Fe component rising and Zn component falling above the melting point were confirmed. By connecting sectors with a high rate of change, the fire pattern was completed and the ignition point was found, The change was greater in the watered state than in the non-watered state. The reason is that the surface is oxidized by water and the change is large. All experimental data are displayed in colors, graphs, and amount of change for easy interpretation, and reliability of data is improved by analyzing in three directions (X, Y, and Z axes) rather than in one direction. Through this material component analysis, it was possible to estimate the fire pattern, and based on this, it was confirmed that the ignition point could be found.

산불은 산림환경을 교란하는 요인 중 하나이며, 최근에는 산불의 발생패턴이 시간적, 공간적 으로 달라지고 있다. 이 연구는 우리나라의 1991년부터 2020년까지 30년간의 산불 발생 자료를 이용하여 산불 발생패턴의 시간적, 공간적 변화를 분석하였다. 30년간 산불 발생 건수는 통계적 으로 유의한 변화를 보이지 않았으나, 계절적으로는 봄철 산불의 최다 발생 시기가 3월로 1개 월 앞당겨졌으며, 여름철 산불 발생 건수가 증가하였다. Hot Spot 공간분석에 따르면, 거주인구 가 많고 야외활동이 활발한 수도권 및 대도시가 발생 건수에 관한 Hot Spot으로 나타났으며, 산 불 발생 면적에 대해서는 강원도, 울진 등 동해안지역이 대형산불이 집중하여 나타나는 지역으 로 분석되었다. 행정 구역별로 10년 단위의 산불 발생 건수를 분석한 결과, 합천군, 광주시, 창 원시가 발생 건수가 점차 증가하는 지역으로 나타났으며, 반대로 청주 상당구, 부산 남구, 태백 시는 산불 발생이 감소하는 지역으로 나타났다.

To carry out missions effectively in battlefield environments, it is essential to rapidly identify enemy threats and respond through precise analysis. Accordingly, technologies that utilize artificial intelligence to identify and classify targets in real time are being actively researched in modern warfare. In this study, we propose a deep learning-based target classification algorithm that simultaneously ensures real-time performance and high classification accuracy. Time-series data are constructed from infrared imagery and augmented to train the model, and the effectiveness of the proposed algorithm is demonstrated through comparative experiments with various CNN- and RNN-based models.

급속한 도시성장에 따라 도시지역에는 다양하고 복잡한 시설물들이 증가되고 있으며, 이에 따른 화재발생 피해에 대한 위험도도 증가되고 있다. 특히 화재사고는 인위적 재해 중 교통사고와 함께 도시지역에서 가장 높은 발생빈도를 나타내고 있다. 이에 따라 소방방재청에서는 효과적인 화재관리를 위하여 국가화재정보시스템을 운영하고 있으며 2007년부터 화재발생정보를 인터넷을 통해 제공하고 있다. 따라서 본 연구에서는 이 시스템에서 제공하는 데이터와 진주시 소방서로부터 자료를 취득하여 진주시 화재데이터베이스를 구축하고, 2007년부터 2013년까지 화재발생 추이에 대한 시계열분석과 Moran's I, Getis-Ord $Gi^*$분석을 통하여 진주시 공간상의 화재발생 밀도변화분석과 시설물별 화재위험도를 산출하였다. 그 결과 화재발생위치의 시계열적 변화와 화재발생 밀집도가 높은 Hot Spot지역을 추출할 수 있었으며, 시설물별 인명피해 및 재산피해 매트릭스를 작성하여 화재위험등급을 산출함으로서 도시지역의 화재발생위험을 예측할 수 있는 방안을 제시하였다.

Diverse and complex facilities have been on the increase in urban areas in accordance with rapid urbanization. Along the lines of the increase in facilities, the risk of fire has increased. In particular, fire accidents as well as traffic accidents accounted for the highest rate in artificial disasters. Therefore, the National Fire Information Systems managed by the National Emergency Management Agency (NEMA) appeared for the effective fire management. The NEMA has provided the public with the Internet services regarding information about fire outbreak since 2007. This study acquired data from both NEMA and the Jinju City Fire Department. It constructed the fire data of Jinju City and calculated the change in spatial density targeting fire, occurred in Jinju city with a view to examining the fire risk of facilities by conducting a time series analysis on the trends of fire outbreak over a span of periods between 2007 and 2013. It also conducted an analysis of Moran's I, Getis-Ord Gi. Therefore, it came to select higher hot spots in terms of fire location and fire density. In addition, it attempted to calculate the levels of fire hazard by drawing up the matrix of personal injury and property damage, depending on facilities to present the methods, which can predict the risk of fire occurrence in urban areas.

Fire-alarm systems are safety equipment that facilitate rapid evacuation and early suppression in case of fire. It is highly desirable that fire-alarm systems have low false-alarm rates and are thus reliable. Until now, researchers have attempted to improve detector performance by applying new technologies such as IoT. To this end, IoT-based fire-detection systems have been developed. However, due to scarcity of large-scale operational data, researchers have barely studied malfunctioning in fire-alarm systems or attempted to reduce false-alarm rates in these systems. In this study, we analyzed false-alarm rates of smoke/temperature detectors and unwanted fire-alarm signal patterns at K institution, where Korea's largest IoT-based fire-detection system operates. After analyzing the fire alarm occurrences at the institution for five years, we inferred that the IoT-based fire-detection system showed lower false-alarm rates compared to the automatic fire-detection equipment. We analyzed the detection pattern by dividing it into two parts: normal operation and unwanted fire alarms. When a specific signal pattern was filtered out, the false-alarm rate was reduced to 66.9% in the smoke detector and to 46.9% in the temperature detector.

Water mist fire suppression system utilizes the fire suppression features such as cooling of fire source, dilution of ambient oxygen, and shielding of radiation heat with the evaporation of microscopic water droplets. The momentum of water mist is relatively low and the infiltration of water mist to the fire source is not effective. In addition to lower penetration force, the evaporated water vapor is liable to decline to limited portion of fire source due to its light weight and sparse density. On the other hand, the cycling water mist system is expected to improve the penetration force of water mist as well as the dilution coverage capability with the stratified spray characteristics. At this paper we present the analyzed fire suppression capability of intermittent water spray pattern by use of FDS which is computational fire dynamics fire model. We expect this analysis shall be supportive to the development of the prototype of water mist nozzle.