Every year, diverse types of safety accidents cause major damage to human life and property. In particular, failure to suppress safety accidents caused by fires during the early stages can lead to large-scale accidents, which in turn can cause more serious damage than other types of accident. Therefore, this paper presents an analysis of the prevailing research trends and future directions for research on preventing safety accidents due to fire. Since fire outbreaks can occur in many types of places, the study was conducted by selecting the places and causes involved in frequent fires, using fire data from Korea. As half of these fires were found to occur in buildings, this paper presents an analysis of the causes of building fires, and then focuses on three themes: fire prevention based on fire and gas detection; fire prevention in electrical appliances; and fire prevention for next-generation electricity. In the gas detection of the first theme, the gas referred to does not denote a specific gas, but rather to the gas used in each place. After an analysis of research trends for each issue related to fire prevention, future research directions are suggested on the basis of the findings. It is necessary to evaluate the risk, select a detection system, and improve its reliability in order to thoroughly prevent fires in the future. In addition, an active emergency response system should be developed by operating a fire prevention control system, and safety training should be developed after classifying the targets of the training targets appropriately.
Among the fire reports caused by seasonal devices registered with the Korea Fire Information Center in 2021, fires caused by heating cables accounted for the largest portion with 350 cases. As a result of analyzing the heating cable fires from 2015 to 2021, we have classified the heating cable fires into four types according to the method of winding the heating wire. First, we hypothesized that the temperature is high when the density is high due to the overlap of the hot wires or when there is a thermal insulating material. We predicted that the temperature would rise through a random game and established a reproducibility test plan. In order to check how heat generation changes depending on the winding method of antifreeze heating cables, we selected 10 manufacturers and checked the temperature characteristics according to the test conditions (Paragraph 11, Paragraph 19.101) of the Technical Regulations for Electrical and Telecommunication Products and Components of Korea (K 10013), tested the four methods mentioned in this thesis and compared and analyzed the results. The experiment results indicate that the temperature of the heater part in antifreeze heating cables was mostly higher than the conditions required by the existing standards in cases 1 to 4. In particular, in the case of No. 5 manufacturer’s sample, the temperature of the heating cable of Case 1 was measured to be the highest at 119.0 °C. In addition, as a result of applying the data engineering reproducibility test results in the framework of the random game λ proposed in this thesis, we have derived the same results as the predicted hypothesis. Case 1 refers to the case where a fire occurs due to the heating cable being wrapped around the water pipe and insulation or taped outside; It is one of the methods that users actually use a lot in the field. Based on experiment, we have concluded that the fire risk is high under the Case 1 condition. Thus, the test conditions in the existing K 10013 Standard need to be strengthened according to the Case 1 condition.
As per the fire statistics survey of 2019, 56.5% (152 cases) of the entire fire accidents (269 cases) caused by heating cables were due to electrical factors. Therefore, in the present work, the electrical factors responsible for heating cable fire have been analyzed, and fire prevention measures have been demonstrated through related reproduction experiments. According to heating cable fire statistics, the fire in anti-freezing appliances (heating cables), except for fires caused by electric cable arcing and other unknown factors, can be classified into four types based on installation configurations. These configurations have been classified and tested according to the Technical Regulations for Electrical and Telecommunications Products and Components (K 10013). The results of a comparative experiment on anti-freezing appliances (heating cable) revealed that the configuration “a type of water pipe with a heating cable wrapped around the water pipe and insulation on the outside” showed the highest temperature among the four installation arrangements. Additionally, the maximum difference between the test temperature (K 10013) and the actual temperature was 40 ℃.
In the Republic of Korea, fire outbreaks caused by electrical devices are one of the most frequent accidents, causing severe damage to human lives and infrastructure. The metropolitan police, The National Institute of Scientific Investigation, and the National Fire Research Institute conduct fire root-cause inspections to determine whether these fires are external or internal infrastructure fires. However, obtaining results is a complex process. In addition, the situation has been hampered by the lack of sufficient digital forensics and relevant programs. Apart from electrical devices, multi-sockets are among the main fire instigators. In this study, we aim to verify the feasibility of utilizing YOLO-based deep-learning object detection models for fire-cause inspection systems for multi-sockets. Particularly, we have created a novel image dataset of multi-socket fire causes with 3300 images categorized into the three classes of socket, both burnt-in and burnt-out. This data was used to train various models, including YOLOv4-csp, YOLOv5n, YOLOR-csp, YOLOv6, and YOLOv7-Tiny. In addition, we have proposed an improved YOLOv5n-SE by adding a squeeze-and-excitation network (SE) into the backbone of the conventional YOLOv5 network and deploying it into a two-stage detector framework with a first stage of socket detection and a second stage of burnt-in/burnt-out classification. From the experiment, the performance of these models was evaluated, revealing that our work outperforms other models, with an accuracy of 91.3% mAP@0.5. Also, the improved YOLOv5-SE model was deployed in a web browser application.