Intelligent fire-fighting monitoring method and system for high-rise building

Abstract

The invention relates to the technical field of building fire protection, and discloses an intelligent fire protection monitoring method and system for a high-rise building, and the method comprises the following steps: arranging a multi-mode sensor node at a construction site of the high-rise building for collecting environment data; transmitting the environment data to a central monitor through a wireless sensor network; performing denoising processing on the environmental data by using a Kalman filter, and fusing the data of the multi-modal sensor through a dynamically adjusted weighted fusion algorithm to generate preliminary judgment parameters of fire monitoring; and evaluating the fire risk based on the dynamic Bayesian decision network, and generating the risk probability of fire occurrence. According to the technical scheme, the multi-mode sensor network and the wireless sensor module are combined, and all-directional and high-precision real-time collection and transmission of construction site environment data are achieved. The problem of insufficient data acquisition precision in a complex construction environment is solved.

Description

Technical Field

[0001] The present invention relates to the technical field of building fire protection, and specifically to an intelligent fire monitoring method and system for high-rise buildings. Background Art

[0002] The potential fire hazards at the construction sites of high-rise buildings are relatively complex and have significant dynamics. The construction areas are usually unevenly distributed and the environmental conditions are variable. With the acceleration of the urbanization process, the construction projects of super high-rise buildings are gradually increasing. Building fires not only threaten the lives of construction workers but also may cause huge economic losses. During the construction stage, due to factors such as the open structure of the unfinished building, the imperfectness of equipment, and the use of temporary wires and flammable materials, the fire risk is significantly increased. At the same time, there are many environmental interference factors at the construction site, such as dust, noise, high temperature, etc., which further increase the difficulty of fire monitoring.

[0003] In order to cope with the potential risks of fires at the construction sites, some fire monitoring solutions have been proposed in the prior art. Most of these technologies are sensor-based and monitor fire hazards by collecting environmental data (such as temperature, humidity, smoke concentration). However, the application of these technologies is still restricted by various factors. Especially in complex construction environments, their monitoring effects are often not ideal. For example, the monitoring system with a single type of sensor is prone to false alarms or missed alarms due to the lack of comprehensive perception ability for multiple parameters; while the fixed installation and limited coverage of the monitoring equipment make the system unable to adapt to the frequently changing layout of the construction site during actual operation.

[0004] In terms of fire risk assessment, traditional systems generally rely on simple threshold comparison methods and cannot effectively integrate the data information of multiple sensors for comprehensive judgment. This single assessment method is difficult to cope with the dynamic changes of environmental conditions during the construction process. At the same time, the response mode of the fire alarm system is also relatively fixed, usually a single alarm signal, lacking the hierarchical response ability for different fire development stages, and it is difficult to achieve efficient resource scheduling. In addition, there are usually monitoring blind spots in special areas (such as high-altitude operation areas, narrow enclosed spaces) during the construction of high-rise buildings. Fixed monitoring equipment is difficult to provide comprehensive coverage, thus reducing the timeliness and accuracy of fire monitoring. For this reason, those skilled in the art have proposed an intelligent fire monitoring method and system for high-rise buildings to solve the above problems. Summary of the Invention

[0005] Aiming at the deficiencies of the prior art, the present invention provides an intelligent fire monitoring method and system for high-rise buildings, which solves the problem that the fire monitoring system in the prior art cannot adapt to the frequently changing layout of the construction site during actual operation.

[0006] To achieve the above object, the present invention is realized through the following technical solutions: An intelligent fire monitoring method for high-rise buildings, comprising the following steps: Deploy multi-modal sensor nodes at the construction site of high-rise buildings for collecting environmental data; Transmit the environmental data to the central monitoring through a wireless sensor network; Use a Kalman filter to denoise the environmental data, and fuse the multi-modal sensor data through a dynamically adjusted weighted fusion algorithm to generate preliminary judgment parameters for fire monitoring; Evaluate the fire risk based on a dynamic Bayesian decision network to generate the risk probability of a fire occurring; Predict the fire spread path and the location of the fire source through a heat conduction model to generate a dynamic fire situation map; Trigger a multi-level alarm mechanism according to the fire risk assessment result; Optimize the sensor weights according to the feedback of historical monitoring data and dynamically adjust the alarm strategy.

[0007] Preferably, the dynamically adjusted weighted fusion algorithm determines the weights by calculating the contribution degrees of the data of each sensor in real time, and the adjustment basis of the weights includes the sensor type, data accuracy and the influence factor of the current environmental change, and the fusion result is the weighted sum of the monitoring values of each sensor and its weight.

[0008] Preferably, the state update process of the Kalman filter includes calculating the optimal state estimate value through the prior prediction value and the observation value, and the state estimate value is calculated based on the covariance matrix of the sensor measurement value and the noise characteristic.

[0009] Preferably, the input of the dynamic Bayesian decision network is the multi-modal sensor fusion data processed by the Kalman filter, and the dynamic Bayesian decision network calculates the conditional probability of a fire occurring based on the prior fire state probability and the current observation data.

[0010] Preferably, the heat conduction model is based on the initial temperature distribution of the fire source point and the physical boundary conditions of the building area to predict the diffusion behavior of the fire at the construction site, and the fire spread path is dynamically calculated and generated by the heat diffusion coefficient, the convection velocity field and the heat source intensity.

[0011] Preferably, the heat conduction model uses the finite difference method for numerical solution, discretizes the calculation of the heat diffusion behavior, and dynamically updates the heat source intensity according to the environmental temperature data collected by the real-time sensor.

[0012] Preferably, the multi-level alarm mechanism includes: When the probability value of the fire risk assessment exceeds the first preset threshold, activate the on-site audible and visual alarm device and notify the on-site construction personnel; when the predicted fire spread path exceeds the defined high-risk area range, mobilize the drone to the high-risk area to collect real-time image data; When the probability value of the fire risk assessment exceeds the second preset threshold, push the fire situation dynamic map to the project person in charge and the remote management system.

[0013] Preferably, the drone automatically navigates to the high-risk area according to the fire spread path predicted by the heat conduction model, collects real-time image data of the area and transmits it to the central monitoring center for further analysis of the fire development trend.

[0014] Preferably, the dynamic feedback optimization of the sensor weights adjusts the weight distribution according to the false alarm rate and missed alarm rate of the sensor historical data, and improves the accuracy of sensor data fusion by minimizing the sum of false alarms and missed alarms.

[0015] A high-rise building intelligent fire monitoring system, comprising: A multi-modal sensor module for collecting environmental data at the construction site; A wireless sensor module for transmitting the collected environmental data to the central monitoring system; A central monitoring module for denoising the environmental data, data fusion, fire risk assessment and fire situation dynamic map generation; An alarm module for triggering multi-level alarm responses according to the fire risk assessment results; A drone module for entering the high-risk area to collect real-time image data according to the fire spread prediction result and transmitting the image data to the central monitoring system.

[0016] The present invention provides a high-rise building intelligent fire monitoring method and system. It has the following beneficial effects: 1. The present invention adopts the technical scheme of combining a multi-modal sensor network with a wireless sensor module, realizing the all-round and high-precision real-time collection and transmission of the environmental data at the construction site. Compared with the single-sensor monitoring mode in the prior art, which is prone to environmental interference resulting in data loss or false alarms, it solves the problem of insufficient data collection accuracy in complex construction environments.

[0017] 2. Through the data fusion algorithm and fire risk assessment model built in the central monitoring module, the present invention achieves the technical effect of comprehensively processing multi-modal sensor data and generating the fire risk level in real time. Compared with the problem in the prior art that the fire risk is directly judged by single-sensor data, resulting in inaccurate assessment, this solution significantly improves the reliability of fire hazard identification.

[0018] 3. The present invention designs a multi-level alarm mechanism, which gradually triggers on-site alarms, remote notifications, and drone linkages according to the fire risk level, achieving the technical effects of hierarchical alarm and rapid response. Compared with the limitation of the unified alarm mode in the prior art that cannot adjust the alarm intensity according to the fire development stage, this solution effectively solves the problems of false alarms and delayed responses.

[0019] 4. The drone module of the present invention combines the fire spread prediction results and real-time image acquisition technology, achieving the technical effects of flexible monitoring of high-risk areas and dynamic data support. In the prior art, the method relying on fixed monitoring devices is difficult to cover dangerous areas, while the present invention overcomes the problem of monitoring dead spots, significantly improving the coverage and efficiency of fire monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 is a schematic flowchart of the method of the present invention; Figure 2 is a schematic diagram of the system architecture of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0021] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the drawings of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0022] Please refer to the attached Figure 1 , an embodiment of the present invention provides an intelligent fire monitoring method for high-rise buildings, including the following steps: S1. Deploy multi-modal sensor nodes at the construction site of high-rise buildings for collecting environmental data.

[0023] Specifically, the deployment of the multi-modal sensor network is one of the core links. This sensor network provides a basic guarantee for subsequent data collection, processing, fusion, and fire risk assessment. The deployment of sensors not only needs to comprehensively cover high-risk areas, but also needs to reasonably determine the node spacing and distribution density in combination with the dynamic characteristics of the construction scenario to achieve efficient collection of environmental data. The implementation of this step lays a solid foundation for subsequent wireless data transmission and processing.

[0024] The multi-modal sensor network consists of a temperature and humidity sensor, a smoke concentration sensor, an infrared thermal imager, and a carbon dioxide concentration sensor. These sensors are distributed in different areas of the construction site to capture various physical and chemical characteristics generated before, during, and during the spread of a fire.

[0025] The temperature and humidity sensor is used to monitor the changes in temperature and humidity in the environment. When the environmental temperature rises rapidly or the humidity drops significantly, these changes may indicate early signs of a fire.

[0026] The smoke concentration sensor is mainly used to detect the changes in the concentration of suspended particulate matter in the air. Specifically, when smoke appears in the initial stage of a fire, the concentration of particulate matter in the air will rise rapidly. At this time, the smoke concentration sensor will record this change in real time and report it to the central monitoring system.

[0027] The infrared thermal imager is mainly used to detect the distribution of thermal radiation intensity in a region. In one possible implementation, the infrared thermal imager can perform non-blind thermal imaging monitoring on the entire construction site, so as to quickly locate the fire source or high-temperature area.

[0028] The carbon dioxide concentration sensor monitors the CO 2 content in the air to capture specific chemical changes generated during the fire combustion process. For example, during a fire, a large amount of carbon dioxide is released due to the combustion of organic matter, so this sensor is of great significance for monitoring the combustion reaction.

[0029] The layout method of the sensors can be dynamically adjusted according to the different risk levels of the construction area. As an option, in high-risk areas (such as areas near flammable material storage areas and power equipment areas), a high-density layout mode can be adopted. The node spacing is usually 5 - 10 meters to ensure the fineness and real-time nature of the monitoring. In low-risk areas (such as open areas or the periphery of buildings), the node spacing can be relaxed to more than 20 meters, thus saving system resources.

[0030] The layout method also needs to be combined with the structural characteristics of the construction site. For example, in a closed construction area, the sensors can be installed on the indoor walls or ceilings and adjusted according to the construction progress. In an open area, portable sensor nodes can be used to cope with changes in external environments such as wind speed and air flow.

[0031] The sensor nodes are interconnected through a wireless communication module to form a network. This network has self-organizing capabilities and can adjust the communication path according to the real-time working status of the nodes. For example, when a certain node fails, its neighboring nodes can ensure the integrity and timeliness of the data through hop-by-hop relaying transmission.

[0032] To enhance the accuracy and reliability of the monitoring data, the sensor nodes are built-in with basic data preprocessing capabilities. The sensors can perform preliminary data smoothing and outlier filtering locally. For example, when the smoke concentration sensor detects abnormal fluctuations in the concentration value within a short period of time, the node will determine whether the fluctuation is caused by external environmental changes (such as wind-blown dust) through built-in algorithms and decide whether to upload the data.

[0033] The installation of sensor nodes also takes into account the power supply method. The nodes are powered by lithium batteries and combined with solar panels to achieve long-term self-sustaining operation. In fixed construction areas, the nodes can be directly connected to the temporary power system to ensure continuous monitoring.

[0034] The recording of sensor data can be expressed in the following form: D i ={T, H, S, R, C, t} Where: T represents temperature; H represents humidity; S represents smoke concentration; R represents infrared thermal radiation intensity; C represents carbon dioxide concentration; t is the timestamp.

[0035] The layout of sensor nodes can also be dynamically optimized by combining thermodynamic modeling and data analysis results. For example, when the fire risk in a certain area is known to be high and the heat conduction model predicts the fire source propagation path, the sensor density in this area can be temporarily increased.

[0036] S2. Transmit environmental data to the central monitoring through the wireless sensor network.

[0037] Specifically, after the layout of multi-modal sensor nodes is completed, in order to achieve real-time transmission of the collected data, the data of the sensor network needs to be uploaded to the central monitoring. This process is achieved through the wireless sensor network. The accuracy and real-time nature of the transmission are the basis for the efficient operation of the entire monitoring system. The design of the wireless sensor network must be consistent with the aforementioned multi-modal sensor layout scheme to ensure that various environmental data can be quickly and stably collected into the central monitoring, thereby supporting subsequent denoising processing, data fusion, and fire risk assessment.

[0038] The wireless sensor network consists of a multi-hop distributed network architecture, and each sensor node has both data collection and relay functions. Specifically, each node establishes a communication link with its neighboring nodes through a short-range wireless communication protocol (such as LoRa, ZigBee, or Wi-Fi), and gradually converges the data to the regional gateway through the relay method. Finally, the regional gateway transmits the data to the central monitoring through a high-speed communication protocol (such as 5G or Ethernet).

[0039] The wireless transmission distance of sensor nodes is proportional to their power consumption. To balance node energy consumption and transmission reliability, the transmission power of sensor nodes can be dynamically adjusted according to the distance. For example, when the distance between nodes is relatively close, the transmission power can be reduced to extend the battery life; when the nodes are far apart or there are communication obstacles, the transmission power will automatically increase.

[0040] The network topology adopts a dynamic self-organizing mode. When a certain node fails or goes offline, its neighboring nodes can restore the communication link by re-planning the route. For example, when node A fails, node B and node C will re-establish the optimal path through the real-time network protocol to ensure uninterrupted data transmission. This self-healing function significantly improves the robustness of the transmission network.

[0041] The wireless sensor network adopts a hierarchical architecture design. The first layer is the sensor node, which is used to collect and preliminarily transmit data; the second layer is the regional gateway node, which is used to aggregate and compress data; the third layer is the interface of the central monitoring system, which is used to receive and process data. The hierarchical structure can significantly reduce the overall latency of the system and improve the efficiency of data transmission at the same time.

[0042] The data transmission of the wireless sensor network includes various environmental parameters, and the common data structure can be expressed as: D ij ={T, H, S, R, C, t, loc} Where: D ij is the complete data packet collected by node i at time j; T represents the temperature value collected by the node; H represents the humidity value; S represents the smoke concentration; R represents the infrared thermal radiation intensity; C represents the carbon dioxide concentration; t represents the timestamp of the collected data; loc represents the geographical location of the sensor node.

[0043] To further improve the reliability of data transmission, a check code is attached to each data packet in this embodiment. Generally, the check code is generated by using the cyclic redundancy check technology and is used to detect and correct possible data errors during transmission. For example, when the data packet D ij arrives at the central monitoring center, the system will verify the attached check code. If an error code is found, the corresponding node will be automatically requested to re-transmit the data packet.

[0044] The sensor network also has the function of data priority allocation. For example, when the node collects some significantly abnormal data (such as a sharp rise in smoke concentration or a temperature exceeding the critical value), the priority of the data packet can be automatically increased, so that it is preferentially transmitted to the central monitoring center. This mechanism can significantly shorten the transmission time of key data, thus providing support for fire warning.

[0045] The data priority allocation can be calculated by the following formula: P = α·|T - T 0 | + β·S + γ·R Where: P represents the priority of the data packet; T is the currently collected temperature value; T 0 is the reference temperature of the environment; S is the current smoke concentration; R is the current infrared radiation intensity; α, β, γ are weight coefficients, which are dynamically adjusted according to different scenarios.

[0046] To reduce network latency, edge computing technology can be adopted. A preliminary data analysis module is integrated on the regional gateway node to denoise and smooth the collected raw data. For example, when the data of temperature and humidity sensors in a certain area changes little, the gateway node can directly upload the average value, thus reducing the transmission volume of invalid data.

[0047] In addition, the deployment of the wireless sensor network also needs to be combined with the dynamic characteristics of the construction site. For example, in a mobile construction area, the sensor nodes can move with the construction progress, and the regional gateway node can be flexibly deployed through mobile power supply devices. This dynamic adjustment mechanism can ensure that the construction site is always within the coverage of the monitoring network.

[0048] S3. Use a Kalman filter to denoise the environmental data, and fuse the multi-modal sensor data through a dynamically adjusted weighted fusion algorithm to generate preliminary judgment parameters for fire monitoring.

[0049] Specifically, after the multi-modal sensor network is successfully deployed and the data is transmitted through the wireless sensor network, the environmental data enters the central monitoring for processing. Since the sensor data may be interfered by the external environment, such as electromagnetic noise, wind speed changes or instantaneous fluctuations caused by construction activities, directly using the raw data may lead to misjudgment. Therefore, the present invention uses a Kalman filter to denoise the environmental data, and further fuses the multi-modal sensor data through a dynamically adjusted weighted fusion algorithm, so as to generate preliminary judgment parameters for fire monitoring. This process provides accurate and reliable input data for subsequent fire risk assessment.

[0050] The Kalman filter effectively eliminates noise by dynamically predicting and observing and updating the time-series data of each sensor node. Generally, the mathematical model of the Kalman filter includes a system state equation and an observation equation, which are specifically described as follows: System state equation: X t = AX t-1 + BU t + w t Where: X t represents the state vector of the system at the current moment, such as the temperature T, humidity H, smoke concentration S, etc. collected by the sensor; A is the state transition matrix, reflecting the characteristics of the system state changing with time; B is the control input matrix, used to describe the influence of the external input U t on the system state (in this embodiment, it is usually taken as zero, indicating no external interference input); U tis an external input vector, used to represent the external control or interference received by the system (in general, external interference is not considered in the present invention, and here it is a zero vector); w t is process noise, following a normal distribution N(0, Q), where 0 is the mean of the distribution, indicating that the expected value of this random variable is zero, usually used to represent the central position of the noise, and Q is the process noise covariance matrix, used to represent the intensity of the noise within the system.

[0051] Observation equation: Z t = HX t + v t where: Z t represents the observed value of the current sensor; H is the observation matrix, used to map the system state to the observation space; v t is the observation noise, following a normal distribution N(0, R), where R is the observation noise covariance matrix, representing the intensity of the sensor measurement noise.

[0052] The core of the Kalman filter lies in calculating the optimal state estimate value at the current moment through a recursive algorithm. The process includes the following two parts: prediction and update.

[0053] Prediction step: P t|t-1 = AP t-1|t-1 A T + Q where: represents the predicted value (prior estimate) of the state vector at time t, derived based on the state estimate value at time t - 1 and the system model; A is the state transition matrix, representing the dynamic relationship of the system state from time t - 1 to time t, usually defined by the system model; represents the estimated value (posterior estimate) of the state vector at time t - 1, calculated by combining the previous observation data and the system model; P t|t-1 is the predicted covariance matrix, representing the uncertainty of the state prediction value at time t, used to quantify the accuracy of the state estimate; P t-1|t-1 is the posterior covariance matrix, representing the uncertainty of the state estimate value at time t - 1; A T is the transpose of the state transition matrix A, used to maintain the symmetry of matrix operations; Q is the process noise covariance matrix, describing the uncertainty brought by the random process noise during the state transition process of the system.

[0054] Update step: K t = P t|t-1 H T (HP t|t-1 HT +R) -1 P t|t =(I - K t H)P t|t-1 Where: K t Kalman gain, representing the contribution degree of the current observation value to the state estimation. The Kalman gain is used to find the best balance point between the predicted value and the observation value to reduce the estimation error; H is the observation matrix, defining the relationship between the state variables and the actual observation values; H T The transpose of the observation matrix, used for matrix operations; R is the observation noise covariance matrix, describing the uncertainty of the sensor measurement noise; (HP t|t-1 H T +R) -1 The inverse of the innovation covariance, used to quantify the impact of the uncertainty of the observation information on the state update; The posterior state estimate value, the state estimate value updated by combining the predicted state and the observation information, representing the optimal state estimate at the current moment; Z t The observation value vector, the actual observation data from the sensor; Mapping the predicted state value to the observation space through the observation matrix H, representing the predicted observation value; Innovation, representing the difference between the actual observation value and the predicted observation value, used to correct the state estimation; P t|t The posterior covariance matrix, representing the uncertainty of the updated state estimation; I is the identity matrix, with the same dimension as the state vector, used to maintain the linear characteristics of the matrix operations.

[0055] The values of the covariance matrices Q and R can be dynamically adjusted based on historical data. For example, when the noise level at the construction site is high, increasing the value of R can reduce the impact of the observation data on the estimated value, thereby enhancing the robustness of the system.

[0056] After completing the denoising process, the present invention further fuses the multi-modal sensor data through a dynamically adjusted weighted fusion algorithm. This algorithm makes each sensor provide the optimal contribution in different environments by adjusting the weight w i in real time.

[0057] The fusion process can be expressed as: Where: F is the fused comprehensive monitoring parameter, used for subsequent fire risk assessment and decision support, representing the weighted fusion result of the outputs of multiple sensors; F i The processed output data of the i-th sensor, such as temperature, humidity, smoke concentration, etc.; w iThe weight of the i-th sensor, which reflects the importance of the data of this sensor in the comprehensive monitoring parameter F, and the weights satisfy the normalization condition: N is the total number of sensors, representing the types or quantities of sensors in the multi-modal sensor network.

[0058] Generally, the weight w i is calculated by the following formula: where: α i is the dynamic performance index of the i-th sensor and is the basis for comprehensive weight allocation; is the sum of the dynamic performance indexes of all sensors, which is used to normalize the dynamic performance indexes of each sensor to ensure that the sum of the weights is 1.

[0059] When the infrared thermal imager detects an abnormal hot spot, the corresponding weight will increase significantly to enhance the influence of the high-temperature area on the monitoring results. On the contrary, when the historical false alarm rate of a certain sensor is relatively high, its weight will be appropriately reduced.

[0060] The data fusion process can also be combined with time series analysis. For example, when the data of multiple sensors show consistency within a certain time window, the system can enhance the confidence in the data during this time period. This time-weighting mechanism can further improve the accuracy of the fused data.

[0061] S4. Evaluate the fire risk based on the dynamic Bayesian decision network to generate the risk probability of a fire occurring.

[0062] Specifically, environmental data is collected and transmitted by multi-modal sensor nodes, and denoising and fusion processing are completed in the central monitoring to generate preliminary judgment parameters for fire monitoring. However, it is impossible to accurately judge the possibility and risk level of a fire only based on the fused data output by the sensors. Therefore, the present invention evaluates the fire risk based on the dynamic Bayesian decision network, and obtains a scientific risk judgment result by calculating the conditional probability of a fire occurring. This result can not only be directly used for fire warning, but also provides a basic basis for subsequent prediction of fire spread and alarm response.

[0063] The dynamic Bayesian decision network is a technical framework for multi-variable state evaluation through probabilistic reasoning. It consists of nodes and edges, where the nodes represent state variables or sensor observations, and the edges represent the dependence relationships between variables. Generally, the dynamic Bayesian network includes the following basic components: State variable S t : Represents the fire hazard state (such as "no fire", "possible fire", "high-risk fire"), which is the target variable to be evaluated; Observation variable O t: Represents the monitoring data after sensor fusion, including temperature T, humidity H, smoke concentration S, infrared radiation intensity R, carbon dioxide concentration C, etc.; Conditional probability table: Defines the probability relationships between variables and is used to describe the dependence of the hazard state on the observed data.

[0064] The dynamic Bayesian decision network realizes fire risk assessment through the state transition model and the observation model in the time series. Specifically, its mathematical model includes the following two parts: State transition model: P(S t |S t-1 ) = f(S t-1 ) Where: P(S t |S t-1 ) is the conditional probability of the state variable S t at the current moment, which depends on the state variable S t-1 at the previous moment; f(S t-1 ) is the state transition function, which is used to characterize the dynamic characteristics of the fire risk changing over time.

[0065] Observation model: P(O t |S t ) = g(O t , S t ) Where: P(O t |S t ) is the conditional probability of the current observation variable O t , which depends on the state variable S t ; g(O t , S t ) is the observation model, which is used to describe the influence of the fire state on the sensor observation values.

[0066] The inference process of the dynamic Bayesian decision network is realized through the following formula: Where: P(S t |O t , S t-1 ) is the posterior probability of the current fire state S t under the condition of the given observation value O t-1 and the previous state S t ; P(O t ) is the marginal probability of the observation value, which is used as a normalization constant to ensure that the sum of all probabilities is 1.

[0067] Observation model P(O t |S t)It can be obtained through historical data and machine learning methods. For example, by analyzing the changing trends of temperature and humidity data recorded by sensors during a fire, the probability relationship between temperature increase, humidity decrease, and fire status can be established.

[0068] The initial state probability P(S 0 ) of the dynamic Bayesian network can be obtained through expert experience or historical data statistics. For example, in a construction environment, if a certain area has long-term accumulation of flammable materials, the system can initialize the fire hazard status of this area as "medium risk". At the same time, the state transition model P(S t |S t-1 ) can be dynamically adjusted according to the changing rules in the time series. For example, when the smoke concentration rises sharply in a short period of time, the transition probability from "no fire" to "possible fire" will increase significantly.

[0069] The input of the observed variable O t includes the fusion result F of multi-modal sensors, and the specific expression is: O t ={F T , F H , F S , F R , F C} Where: F T is the fused temperature data; F H is the fused humidity data; F S is the fused smoke concentration data; F R is the fused infrared thermal radiation data; F C is the fused carbon dioxide concentration data.

[0070] Through the inference of the dynamic Bayesian network, the posterior probability of the current fire hazard status can be obtained. For example: If P(S t ="high-risk fire") > 0.8, the system will immediately enter the alarm mechanism; If 0.5 < P(S t ="possible fire") ≤ 0.8,, the system will continuously monitor and record the dynamic changes.

[0071] The dynamic Bayesian network can also be combined with a time-weighting mechanism to enhance the stability of fire status inference. For example, in a long time window, if the data of a certain sensor is continuously abnormal, the system will gradually increase the posterior probability of this state, so as to avoid false alarms caused by short-term fluctuations.

[0072] S5. Predict the fire spread path and the location of the fire source through the heat conduction model, and generate a dynamic fire map.

[0073] Specifically, the present invention uses a heat conduction model to predict the fire spread path and the location of the fire source. This model generates a dynamic fire situation map through comprehensive analysis of building structures, heat diffusion characteristics, and real-time environmental data. Fire spread prediction can not only judge in advance the potential threats of a fire to the construction site but also provide a scientific basis for mobilizing resources and formulating emergency plans. This step is closely combined with the aforementioned fire risk assessment results, and through dynamic modeling methods, efficient and accurate fire spread prediction is achieved.

[0074] The mathematical modeling of fire spread is based on the heat conduction equation, which describes the temperature diffusion process at the fire source point and the dynamic distribution of heat within the building area. The two-dimensional form of the heat conduction equation can be expressed as: Where: T(x, y, t) represents the temperature field at time t and position (x, y); α is the thermal diffusivity, indicating the diffusion ability of heat in the medium, which is related to the thermal conductivity of building materials; respectively represent the second-order spatial derivatives of temperature along the x and y directions; v is the convective velocity field, indicating the direction and intensity of heat flow; represents the gradient of the temperature field, describing the direction of heat flow; Q(x, y, t) is the heat source function, indicating the heat release intensity at the fire source point, which varies with time and position.

[0075] Solving the heat conduction model requires setting initial conditions and boundary conditions. Generally: Initial condition T(x, y, 0) = T 0 (x, y), representing the initial temperature distribution at the fire source point, determined by the real-time detection values of sensors.

[0076] Boundary conditions: At the boundary of the internal walls of the building, an adiabatic condition is adopted: indicating that the heat flux is zero at the boundary. At the boundary of the open area, an environmental temperature constraint is adopted: T(x, y, t) = T ambient , where T ambient represents the external environmental temperature.

[0077] The expression of the heat source function Q(x, y, t) can be dynamically adjusted according to the actual combustion characteristics of the fire source. For example, for a certain high-rise building construction site, if the heat released at the initial stage of a fire in the area where flammable materials are concentrated is Q 0 , its time evolution can be expressed as: Q(x, y, t) = Q 0 ·e -βt Where: Q 0 is the initial heat release; e is the natural constant, with an approximate value of 2.718; β is the heat decay coefficient, indicating the decreasing rate of heat release during the combustion process; t is the time variable.

[0078] To achieve the dynamic prediction of the fire spread path, in this embodiment, the finite difference method is used to numerically solve the heat conduction equation.

[0079] Spatial discretization: Temporal discretization: Combining the discretization formulas, we get: Where: T i,j represents the temperature value at the grid point (i, j); T i+1,j represents the temperature of the adjacent point on the right side in the x - direction; T i-1,j represents the temperature of the adjacent point on the left side in the x - direction; T i,j+1 represents the temperature of the adjacent point on the upper side in the y - direction; T i,j-1 represents the temperature of the adjacent point on the lower side in the y - direction; Δx, Δy represent the spatial step sizes (grid intervals) of the grid in the x and y directions; represents the temperature value of the grid point (i, j) at the current time step n; represents the temperature value of the grid point (i, j) at the next time step n + 1; Δt represents the time step; α is the thermal diffusivity, representing the diffusion ability of heat in the medium; v x v y are the components of the convective velocity field, representing the flow velocities of heat in the x and y directions respectively; is the heat source term, representing the heat source intensity at the grid point (i, j) and time step n.

[0080] The calculation results of the heat conduction model can be used to generate a dynamic fire map. Generally, the dynamic map contains the following content: Fire source location: located by the local maximum of the temperature gradient; Fire spread path: determined by the heat diffusion direction and the convective velocity field; Distribution of high - temperature areas: the temperature field range of the construction site is shown by temperature isotherms.

[0081] The dynamic map can be dynamically updated in combination with real - time sensor data. For example, when a new hot spot is detected by an infrared thermal imager in a certain area, the system will automatically adjust the heat source function Q(x, y, t) and recalculate the fire spread path.

[0082] The fire spread prediction can also be optimized in combination with building construction drawings and environmental obstacle information. Specifically, the system will adjust the thermal diffusivity α based on the position of the fire - resistant partition wall in the construction drawing. For example, in the area of the concrete wall, a lower α value is set to reflect its lower heat conduction performance.

[0083] S6. Trigger a multi - level alarm mechanism according to the fire risk assessment results.

[0084] Specifically, the fire monitoring system of the present invention achieves a rapid response to fire incidents through a multi-level alarm mechanism. This mechanism combines the results of fire risk assessment and the dynamic analysis of fire spread, triggering different alarm methods at different levels to maximize the protection of personnel safety and minimize fire losses. The design of the alarm mechanism not only requires timeliness but also needs to be flexibly adjusted according to the severity of the fire process, so as to achieve precise resource scheduling and efficient implementation of emergency response.

[0085] The multi-level alarm mechanism is based on the risk levels of fire hazard status (such as "low risk", "medium risk", "high risk") and adopts a progressive triggering principle. Generally, when the risk probability reaches a preset threshold, different levels of alarm mechanisms are activated in sequence. Specifically, the following alarm methods are included: on-site sound and light alarm, drone scheduling, and remote alarm push.

[0086] The alarm logic in this embodiment is implemented through the following mathematical model: A = f(P, τ, R) Where: A is the alarm level; P is the fire risk probability, calculated by a dynamic Bayesian decision network; τ is the fire spread time threshold, predicted by a heat conduction model; R is the shortest path length from the fire source to the high-risk area.

[0087] The calculation of the alarm level follows the following rules: If P > P th1 , trigger a first-level alarm; if P > P th2 and τ < τ th , trigger a second-level alarm; if P > P th3 and R < R th , trigger a third-level alarm.

[0088] Where: P th1 , P th2 , P th3 represent the risk probability thresholds for the first-level, second-level, and third-level alarms respectively; τ th is the critical value of the fire spread time, such as the predicted spread time from the fire source to the crowded area during construction; R th is the critical distance from the fire source to the high-risk area, such as the shortest path to the flammable material storage point.

[0089] The specific implementation process of the multi-level alarm mechanism is as follows: The first-level alarm is triggered when the probability P in the fire risk assessment exceeds the first preset threshold P th1It is triggered when [condition]. Generally, the first-level alarm is used to alert the staff at the construction site of environmental changes. The system will activate the on-site audible and visual alarm devices, emitting a high-frequency flashing light signal and a high-decibel audio signal. At the same time, the alarm information will be sent to the central monitoring system via the wireless sensor network for recording.

[0090] In some specific scenarios, the first-level alarm can also send vibration or notification signals to the on-site staff through portable wireless terminal devices (such as smart bracelets or mobile phones) to ensure that the personnel can receive them in a timely manner.

[0091] The second-level alarm is triggered when the prediction of the fire spread path indicates that the fire source may approach a high-risk area. Specifically, when the fire spread time τ is less than the set threshold τ th the system will immediately dispatch a drone to enter the high-risk area for on-site investigation. The drone navigates to the vicinity of the fire source through a preset path and collects video or thermal imaging data in real time.

[0092] The logical model of drone dispatch can be expressed as: Path UAV = g(R, v, T env ) where: Path UAV is the flight path of the drone; R is the shortest distance between the fire source and the high-risk area; v is the flight speed of the drone; T env is the environmental temperature, which is used to correct the battery endurance of the drone.

[0093] The drone can correct the position of the fire source in real time according to the thermal imaging data. For example, when it is found that the direction of fire spread deviates from the original predicted path, the drone will adjust its flight route and provide first-hand on-site information to the central monitoring system through video backhaul.

[0094] The third-level alarm is triggered when the fire risk assessment probability P exceeds the third preset threshold P th3 and the shortest path length R between the fire source and the high-risk area is less than the set value R th The third-level alarm is used to notify the remote management system or the emergency command center. The system will push the fire situation dynamic map and the risk assessment report to the mobile terminal devices of the project leader and relevant management departments.

[0095] The push information of the third-level alarm includes the following content: The specific location of the fire source and its real-time changes; The prediction result of the fire spread path; The environmental data of the affected area, such as temperature distribution and smoke concentration.

[0096] The three - level alarm can also be linked to the automatic fire extinguishing system (such as a sprinkler system) at the construction site to initiate preventive fire - fighting measures before the fire source approaches the high - risk area.

[0097] The multi - level alarm mechanism can be further optimized in combination with personnel positioning technology. For example, the system can dynamically adjust the alarm range and intensity based on the real - time positioning data of the personnel at the construction site. In areas with a high density of personnel, the coverage area of the alarm signal will automatically increase; in areas with a low density of personnel, the intensity of the alarm signal will be appropriately reduced to avoid waste of resources.

[0098] S7. Optimize the sensor weights according to the feedback of historical monitoring data and dynamically adjust the alarm strategy.

[0099] Specifically, in order to further improve the accuracy and adaptability of the system, the present invention introduces a dynamic feedback mechanism, which aims to optimize the sensor weight allocation and alarm strategy through real - time monitoring and historical record analysis of sensor data and alarm responses. This process can not only reduce the false alarm rate and missed alarm rate, but also adapt to the dynamic changes at the construction site, improving the stability and efficiency of the system during long - term operation.

[0100] The core of the dynamic feedback mechanism is to use the historical records and current monitoring status of sensor data to adjust the sensor weights and system parameters in real - time to enhance the system performance. Generally, the basis for adjusting the sensor weights includes the historical false alarm rate, missed alarm rate, current data fluctuation characteristics of the sensor, and the reliability coefficient of the sensor type.

[0101] The mathematical model for weight adjustment in the present invention can be expressed as: where: w i is the dynamic weight of the i - th sensor; α i is the performance index of the i - th sensor; is the sum of the dynamic performance indexes of all sensors; N is the total number of sensors.

[0102] The performance index α i has the following calculation formula: α i = k 1 ·(1 - E false,i )+ k 2 ·(1 - E miss,i )+ k 3 ·R i where: E false,i is the false alarm rate of the i - th sensor; E miss,i is the missed alarm rate of the i - th sensor; R i is the real - time stability coefficient of the i - th sensor, which is used to represent the fluctuation range of the current data; k1 , k 2 , k 3 is a weight adjustment parameter used to balance the impacts of false alarm rate, missed alarm rate, and real-time data stability on the overall performance.

[0103] The false alarm rate and missed alarm rate can be obtained through statistical analysis of historical data. For example, in the monitoring records of the past 30 days, if a certain sensor has false alarms for fires multiple times, its false alarm rate E false,i will increase significantly, resulting in a corresponding decrease in its weight w i .

[0104] The dynamic feedback mechanism also includes the optimization of the alarm strategy. The optimization of the alarm strategy is achieved by real-time analysis of the effects of alarm responses. For example, when a certain alarm is confirmed to be a false alarm, the system will automatically record the event and analyze the fire risk assessment results, fire spread prediction, and key data features that triggered the alarm at that time.

[0105] The mathematical model for alarm optimization can be expressed as: A opt = A orig - λ·E false where: A opt is the optimized alarm threshold; A orig is the original alarm threshold; λ is the learning rate used to control the optimization amplitude; E false is the false alarm rate.

[0106] If false alarm events are concentrated in a specific alarm level (such as level 1 alarm), the system can focus on adjusting the alarm trigger conditions for that level. For example, increase the trigger threshold P th1 of the level 1 alarm to reduce alarms triggered by low-risk abnormal data.

[0107] The dynamic feedback mechanism not only acts on a single sensor node but can also perform global optimization for the entire sensor network. In one possible implementation, the system will adjust the layout and sampling frequency among nodes based on the overall false alarm rate and missed alarm rate of the sensor network. For example, in a certain area, if the false alarm rate is consistently high, the system can reduce the sampling frequency of sensors in that area or adjust the node spacing to improve the data fusion effect.

[0108] In addition, the dynamic feedback mechanism also supports targeted optimization for special scenarios at the construction site. For example, in high-temperature weather or strong wind weather, the false alarm rate of certain sensors may increase. The system can analyze the weather data and reduce the weight adjustment parameter k 3 in this scenario to reduce the interference of environmental changes on the monitoring results.

[0109] The dynamic feedback mechanism can be used in combination with time series prediction techniques. The system models the change trend of sensor data over a period of time in the past and predicts the possible range of monitoring errors in the future. For example, when the real-time data fluctuation of a certain sensor exceeds the predicted range, the system will automatically reduce its weight and increase the dependence on the data of other sensors.

[0110] A high-rise building intelligent fire monitoring system described below can be referenced correspondingly with a high-rise building intelligent fire monitoring method described above.

[0111] Please refer to the appendix Figure 2 , a high-rise building intelligent fire monitoring system, comprising: A multimodal sensor module for collecting environmental data at the construction site; A wireless sensor module for transmitting the collected environmental data to the central monitoring system; A central monitoring module for denoising the environmental data, data fusion, fire risk assessment, and generating a fire situation dynamic map; An alarm module for triggering multi-level alarm responses according to the fire risk assessment results; A drone module for entering high-risk areas to collect real-time image data according to the predicted results of the fire spread and transmitting the image data to the central monitoring system.

[0112] Specifically, the multimodal sensor module The multimodal sensor module is responsible for deploying various types of sensors at the construction site for real-time collection of environmental data.

[0113] This module includes but is not limited to the following sensors: Temperature sensor: used to monitor the change of environmental temperature, especially the distribution of high-temperature points; Humidity sensor: used to monitor the air humidity to assist in judging the fire environment; Smoke sensor: used to detect the change of smoke concentration to identify the early signal of fire; Infrared thermal imager: used to capture the thermal radiation characteristics of the fire source and provide the ability to locate high-temperature hot spots; Carbon dioxide concentration sensor: used to detect the combustion gas that may be generated by the fire.

[0114] This module achieves full coverage of the construction site through the collaborative work of multimodal sensors, ensuring the accuracy and sensitivity of monitoring. In addition, the sensor module can adapt to the complex environment of the construction site and automatically adjust its acquisition parameters (such as sampling frequency) during the sensor deployment process to balance energy consumption and data quality Wireless sensor module The wireless sensor module is used to transmit the environmental data collected by the multimodal sensor module to the central monitoring system. This module includes the following sub-components: Sensor nodes: Each sensor node is not only responsible for collecting environmental data but also has the ability to transmit wireless data.

[0115] Area network gateway nodes: Used to converge the data of multiple sensor nodes and transmit the data to the central monitoring system through high-speed communication methods (such as 5G, LoRa, or Wi-Fi).

[0116] Network topology control: Supports dynamic network topology adjustment. When a sensor node fails, its neighboring nodes can restore the communication link through an ad-hoc network to ensure the continuity of data transmission.

[0117] The sensor module supports low-power wireless protocols and can achieve long-term reliable operation without increasing the construction burden. In addition, to ensure the security of data transmission, the module can also use encryption algorithms to encrypt the transmitted data in real time.

[0118] Central monitoring module The central monitoring module is the core of the system and is responsible for comprehensively processing sensor data and intelligently evaluating fire-related risks. Specific functions include: Data denoising: Eliminate random noise and environmental interference in the sensor data through built-in filtering algorithms (such as Kalman filters) to ensure the reliability of the data.

[0119] Multimodal data fusion: Perform weighted fusion on the data of different types of sensors to generate comprehensive monitoring parameters for supporting the judgment of fire risks and the dynamic analysis of fire situations.

[0120] Fire risk assessment: Calculate the conditional probability of a fire occurring based on dynamic Bayesian networks or other intelligent algorithms to generate a fire risk level.

[0121] Fire situation dynamic map generation: Combine the fire source location and the predicted results of the fire spread to draw a fire situation dynamic map in real time, providing intuitive support for fire early warning and resource scheduling at the construction site.

[0122] In addition, the central monitoring module also supports the long-term storage and analysis of data, and can optimize the sensor layout strategy and alarm parameters in combination with historical data to improve the adaptive ability of the system.

[0123] Alarm module The alarm module is used to trigger multi-level alarm responses based on the fire risk assessment results to ensure that fire information can be quickly transmitted to relevant personnel and equipment. Specific functions include: On-site alarm: Remind construction personnel of the fire risk through a sound and light alarm device.

[0124] Remote notification: Push fire alarm information to the construction person in charge, the fire management department or the remote monitoring center.

[0125] Alarm classification: When the fire risk is lower than a certain threshold, only the on-site early warning is triggered; When the fire risk reaches the intermediate threshold, notify the construction person in charge and start the drone reconnaissance; When the fire risk exceeds the highest threshold, fully trigger the remote alarm and the linkage of the automatic fire extinguishing equipment.

[0126] The alarm module can dynamically adjust the alarm strategy. For example, it optimizes the alarm level according to the real-time feedback of the sensor data, reducing the possibility of false alarms and missed alarms.

[0127] Drone module The drone module is used to perform inspection tasks in high fire risk areas and provide real-time fire image data. Specific functions include: Automatic navigation: According to the prediction result of the fire spread, automatically plan the flight path and quickly reach near the fire source.

[0128] Real-time data collection: Use on-board infrared thermal imaging equipment, high-definition cameras, etc. to capture the real-time images and thermal distribution information of the fire scene.

[0129] Reconnaissance of high-risk areas: Enter high-risk areas that construction workers cannot reach and provide important decision-making support data for the central monitoring module.

[0130] Flight path optimization: The drone can adjust the flight path in real time in combination with the dynamic fire map to ensure efficient coverage of high-risk areas.

[0131] In addition, the drone module also supports linkage with other modules. For example, in the initial stage of the fire, the drone can be used as a supplementary tool for fire source positioning; in the stage of fire spread, the drone can be used to evaluate the dynamic changes in high-temperature areas.

[0132] Although the embodiments of the present invention have been shown and described, for those of ordinary skill in the art, it can be understood that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A high-rise building intelligent fire monitoring method, characterized in that: The following steps are involved: Deploy multimodal sensor nodes at high-rise building construction sites to collect environmental data; Transmit environmental data to central monitoring via wireless sensor networks; The Kalman filter is used to denoise the environmental data, and the multi-modal sensor data is fused through a dynamically adjusted weighted fusion algorithm to generate preliminary judgment parameters for fire monitoring; Fire risk is assessed based on a dynamic Bayesian decision network to generate the risk probability of fire occurrence; Predict the fire spread path and fire source location through the heat conduction model, and generate a dynamic fire map; Trigger a multi-level alarm mechanism based on fire risk assessment results; Optimize sensor weights based on historical monitoring data feedback and dynamically adjust alarm strategies.

2. A high-rise building intelligent fire monitoring method according to claim 1, characterized in that: The dynamically adjusted weighted fusion algorithm determines the weight by calculating the contribution of each sensor data in real time. The basis for adjusting the weight includes the sensor type, data accuracy and the influencing factors of its current environmental changes. The fusion result is the weighted sum of each sensor monitoring value and its weight.

3. The intelligent fire monitoring method for high-rise buildings according to claim 1 is characterized in that: The state update process of the Kalman filter includes calculating an optimal state estimate value through a priori prediction values ​​and observation values, and the state estimate value is calculated based on a covariance matrix of sensor measurement values ​​and noise characteristics.

4. The intelligent fire monitoring method for high-rise buildings according to claim 1 is characterized in that: The input of the dynamic Bayesian decision network is the multimodal sensor fusion data processed by the Kalman filter, and the dynamic Bayesian decision network calculates the conditional probability of fire occurrence based on the prior fire state probability and current observation data.

5. The intelligent fire monitoring method for high-rise buildings according to claim 1 is characterized in that: The heat conduction model predicts the diffusion behavior of the fire at the construction site based on the initial temperature distribution of the fire source point and the physical boundary conditions of the building area. The fire propagation path is generated by dynamic calculation of the thermal diffusion coefficient, convection velocity field and heat source intensity.

6. A high-rise building intelligent fire monitoring method according to claim 5, characterized in that: The heat conduction model is numerically solved using the finite difference method, the heat diffusion behavior is discretized and the heat source intensity is dynamically updated according to the ambient temperature data collected by the real-time sensor.

7. The intelligent fire monitoring method for high-rise buildings according to claim 1 is characterized in that: The multi-level alarm mechanism includes: When the probability value of the fire risk assessment exceeds a first preset threshold, the on-site sound and light alarm equipment is activated and the on-site construction personnel are notified; When the fire propagation path is predicted to exceed the designated high-risk area, the drone will be deployed to the high-risk area to collect real-time image data; When the probability value of the fire risk assessment exceeds the second preset threshold, the dynamic fire situation map is pushed to the project leader and the remote management system.

8. The intelligent fire monitoring method for high-rise buildings according to claim 7 is characterized in that: The drone automatically navigates to high-risk areas based on the fire propagation path predicted by the heat conduction model, collects image data of the area in real time and transmits it to central monitoring for further analysis of the fire development trend.

9. The intelligent fire monitoring method for high-rise buildings according to claim 1, characterized in that: The dynamic feedback optimization of the sensor weights adjusts the weight distribution according to the false alarm rate and the missed alarm rate of the sensor historical data, and improves the accuracy of the sensor data fusion by minimizing the sum of the false alarms and the missed alarms.

10. An intelligent fire monitoring system for high-rise buildings, applied to an intelligent fire monitoring method for high-rise buildings as claimed in any one of claims 1 to 9, characterized in that: include: Multimodal sensor modules for collecting environmental data at the construction site; Wireless sensor modules are used to transmit collected environmental data to a central monitoring system; Central monitoring module, used for denoising environmental data, data fusion, fire risk assessment and fire situation dynamic map generation; An alarm module is used to trigger a multi-level alarm response based on the fire risk assessment results; The drone module is used to enter high-risk areas according to the fire spread prediction results to collect real-time image data and transmit the image data to the central monitoring system.

Images