Smart buildings centered on digital technology can be perceived as data-driven buildings. Moreover, these smart buildings can be classified into the following elements: (1) physical building infrastructure management, (2) virtual-based building management, (3) data collection and management, (4) artificial intelligence (AI) and big data, and (5) platform-based integrated building management [1]. Therefore, it is important to analyze the physical and virtual elements in smart buildings, data-based AI technology, and platform-based integrated management of buildings [2]. Therefore, an important element for data-driven smart buildings is optimized building data sharing (such as energy, safety, and physical elements), which can be collected through the IoT sensors [3].

1.1. Core Digital Technology of Smart Building: Internet of Things, Big Data, and AI

Smart Internet of Things (IoT) technology is the most important technology element in data collection, where the IoT refers to objects that are connected to the Internet [4]. In a narrow sense, these “things” are simply connected through the Internet, whereas in a broad sense, the IoT can be defined as a technology that provides intelligent value to users beyond the connection between objects and devices through the Internet. In other words, smart IoT is the smallest unit for collecting data in a building, and this building data can be collected through IoT technology [5]. The essential elements that configure the IoT infrastructure in a building system include IoT-based sensors, 5G/6G network equipment, actuators, servers, and gateways [6]. Through these components, data collection consists of three steps: (1) IoT sensor-based data detection, (2) the transmission of the detected data to the smart gateway, and (3) the storage of the data in the server through the smart gateway. For example, consider an IoT sensor system (including temperature, humidity, motion, power, and light sensors) that is built into user A’s house to collect data. These IoT sensors are termed environmental information sensors, through which various pieces of environmental information can be collected. These data are then stored in the energy platform (the data server) in real time through the smart gateway. Big data are defined as technology that involve collecting a large amount of structured and unstructured data from IoT sensors, followed by extracting values from these unstructured data and analyzing the results [7]. Artificial intelligence (AI) is then used with big data to learn through machine learning algorithms, find meaningful values in the data, and make decisions based on the learned results. Accordingly, big data are essential for learning and necessary for obtaining valuable meaning through AI. Moreover, there is a symbiotic relationship between big data and AI.

1.2. Necessity of Building Safety Management System Based on Digital Technology

The safety environment can change significantly due to the large-scale and complex nature of fires, meaning the demand for problem-solving by advanced technical support increases [8]. In addition, as the scope of disaster accidents expands and disaster patterns evolve, it is necessary to analyze and address the causes of disasters. According to the National Fire Information System statistics of the National Fire Administration of the Republic of Korea, a total of 40,114 fires occurred in 2022, causing 341 deaths, 2321 injuries, and 1.204 trillion KRW of property damage. Compared to the previous year, the number of fires increased by 10.6% (3847 cases), the number of casualties increased by 24.9% (479 people), and property damage increased by 9.5% (KRW 104.9 billion). Among the 341 deaths, 105 (30.8%) were aged 70 years or older, 86 (25.2%) were aged 60–69 years, and 76 (22.3%) were aged 50–59 years. There was a high risk of death from the inhalation of toxic fumes and burns. Deaths occurred due to smoke (flame), making evacuation difficult, or locking exits. The main areas where fires occurred were 14,929 (37.2%) in non-residential facilities, 10,497 (26.1%) in residential facilities, and 4669 (11.6%) in vehicles. Although the overall number of fires in non-residential facilities increased due to increased outdoor activities, the highest number of deaths occurred in residential facilities (216 people) [9]. As the exposure to fire risks and resultant casualties in vulnerable groups (such as the elderly) continues to increase, an AI-based smart fire safety service that can predict the occurrence of disasters to reduce subsequent damage is required [10]. Moreover, the need for smart fire safety services with AI technology based on smart city connections is increasing as the development of optimized smart city infrastructure technology progresses [11]. Since disaster safety R&D is closely related to all fields of science and technology, it is important to connect efficiently with each technology. In addition, the importance of digital twin-based prediction, automation, and simulation technologies that can monitor fire disasters through smart sensors and intelligent recognition systems is increasing. Accordingly, efforts are required to analyze the causes of fires and recognize fire situations due to large-scale fires and irregular occurrences of disasters through this digital twin technology [12].

Hongqiang et al. emphasized that quick real-time judgment ability and a 5G and IoT technology-based effective fire evacuation system are important for saving lives in fire situations ^[13][16]. The authors compared and analyzed various studies that applied IoT technology to safety evacuation systems. Although they did not directly review research that combined safety and energy, IoT data can be used in a variety of ways, such as safety systems in buildings and energy systems. Amandeep et al. proposed an IoT-based cloud-fog-enabled hierarchical evacuation system for large-scale evacuations during emergency situations ^[14][17]. Their system utilized the fog computing paradigm to assess panic health conditions in real time. The evacuation system used the fuzzy K-nearest neighbor (FK-NN) method, and the energy-saving mechanism in the fog layer performed data selection and data reduction to reduce the amount of data transmitted to the upper layer.

Jui-Sheng et al. proposed an intelligent fire management system ^[15][18]. In actual fire situations, firefighters spend a significant amount of time rescuing people. However, due to a lack of real-time information, they have to rely on their judgment, which is based on their experience. The time required to rescue people in the event of a fire can be reduced through an intelligent fire rescue system that combines sensors and communication functions to provide real-time information updates, alarm reports, and evacuation routes. In their paper, the location of the mobile phone (user) is determined through the strength of the signal based on a Bluetooth fire detector and a mobile app. However, as the number of fire detection nodes increases, it becomes more difficult to optimize evacuation routes. Accordingly, their proposal compared the characteristics of various route optimization algorithms. Yuxin et al. proposed a concept termed safe fire suppression time (SFT), which combines the characteristics of fire suppression and evacuation behavior to ensure firefighter safety ^[16][19]. In addition, the available safe firefighting time (ASFT) and the required safe firefighting time (RSFT) were defined, and detailed calculation methods were provided. The authors also demonstrated the calculation and application of safe fire suppression times through the fire safety performance design of tunnels. Yapin et al. developed a BIM-based fire safety management system platform that considered construction sites ^[17][20]. The developed platform included four subsystems, among which the escape route optimization subsystem dynamically optimized the evacuation route by considering the possibility of congestion when people passed through the escape route.

Santiago et al. presented a mobile-based application named Wildfire Analyst™ Pocket Edition, a mobile version of WFA ^[18][21] through which firefighters can monitor the expected progress of the fire in real time. Jaziar et al. presented a disaster fire simulation based on participation in a game simulation ^[19][22]. The authors tested how the fire situation, temperature, and smoke develop through a smartphone app based on a game scenario. They also evaluated user experiences with this application and presented the advantages and disadvantages of rescue activities. Nikos et al. developed a web/mobile AEGIS platform for wildfire information management ^[20][23] in which several tasks can be performed: routing, spatial search for the nearest facilities and fire support infrastructure, accessing meteorological data, and visualizing fire management data (such as water sources, gas stations, and evacuation sites).

Suhyun et al. proposed a building fire information management prototype based on 3D visualization to mitigate fire disasters in buildings ^[21][24]. The proposed system provides fire-related information based on BIM technology, enabling emergency responders to identify the location data of indoor facilities. Based on this scenario-based application, a proposed system was demonstrated that contributes to improving rapid access to relevant information. Dahee et al. proposed a BIM-based fire disaster management process that can track building and dynamic fire data simultaneously [8]. For rapid processing, Smart Fire Rescue Management (SFRM) was developed using Revit™ software. The authors conducted a real-case project and an expert survey to evaluate the feasibility of the system. The authors suggested that the proposed system could contribute to accelerating the disaster management process by improving the efficiency of response and rescue in building fires.

Pietro et al. studied a cyber–physical system for wildfire detection and the early detection of forest fires using a UAV ^[22][25]. Here, IoT-based wildfire detection nodes enable early fire detection, enabling continuous monitoring of environmental conditions. When a fire is detected, a UAV can survey the area automatically based on decision-making techniques to find the location of the fire. The authors simulated and demonstrated real-time fire detection capabilities using a forest fire scenario. Rahmi et al. studied a lightweight and attention-based CNN architecture that enables remote control and detection of wildfires using a UAV ^[23][26]. Their paper presented approaches such as transfer learning, deep CNNs, and lightweight CNN to perform wildfire detection tasks using images acquired by a UAV camera. In the paper, the authors demonstrated the suitability of an EfficientNetB0-based model for forest fire detection using a UAV.

Kadhim et al. studied a method for strategically controlling the clustering and scheduling of IoT sensors to reduce energy consumption while maintaining data transmission effectiveness ^[24][27]. IoT sensors can save energy by reducing unnecessary data transmission and data transmission distance by adjusting the number of nodes that consume the most energy during the data exchange process. Hongyu et al. analyzed energy-saving and emission-reduction technologies for the development of smart cities and sustainable low-carbon cities ^[25][28]. The authors also compared the energy savings of typical data centers and analyzed the impact of green data centers on global carbon neutrality goals.