Industrial fire early prevention and control early warning system based on multi-sensor fusion

Abstract

The invention discloses an industrial fire early prevention and control early warning system based on multi-sensor fusion, which belongs to the technical field of industrial safety monitoring and comprises a data acquisition module, a false alarm judgment module, a risk partition module, a fire identification module and a prevention and control optimization module. The data acquisition module acquires fire data through a layered distributed multi-sensor network; the false alarm discrimination module screens effective fire data based on a dynamic false alarm discrimination rule; the risk partitioning module is used for generating a risk thermodynamic diagram and completing sub-region division and risk label distribution; the fire identification module predicts a fire occurrence probability and identifies a fire development stage; the prevention and control optimization module matches a prevention and control strategy based on a fire development stage and a risk label, plans an optimal path for the AGV in combination with a dynamic cost map to execute a prevention and control task, and feeds back an execution result to the data acquisition module to realize dynamic correction; according to the invention, the false alarm rate is reduced, and the reliability and high efficiency of early prevention and control and early warning of industrial fire are improved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of industrial safety monitoring, and particularly relates to an early prevention and warning system for industrial fire based on multi-sensor fusion. BACKGROUND

[0002] There are a large number of flammable and explosive materials, electrical equipment and complex process flows in the industrial production area, and the risk of fire accidents is high. The traditional industrial fire warning system relies on a single type of sensor, and has the problems of limited data collection dimension and high false alarm rate. At the same time, the existing system lacks precise risk zoning capability and cannot develop targeted prevention and control strategies according to the regional fire risk differences. In addition, the fire development stage recognition is lagging, and the AGV path planning does not consider the dynamic risk changes, resulting in untimely prevention and control response and low execution efficiency, which is difficult to effectively curb the spread of early fire. Therefore, an early prevention and warning system for industrial fire with multi-dimensional perception, low false alarm rate, precise zoning and intelligent prevention and control is urgently needed. SUMMARY

[0003] In view of the deficiencies of the prior art, the application provides an early prevention and warning system for industrial fire based on multi-sensor fusion, which comprises a data acquisition module, a false alarm discrimination module, a risk zoning module, a fire identification module and a prevention and control optimization module. The data acquisition module obtains fire data through a hierarchical distributed multi-sensor network. The false alarm discrimination module filters effective fire data based on dynamic false alarm discrimination rules. The risk zoning module generates a risk heat map and completes sub-region division and risk label allocation. The fire identification module predicts the fire occurrence probability and identifies the fire development stage. The prevention and control optimization module matches the prevention and control strategy based on the fire development stage and the risk label, plans the optimal path for the AGV to execute the prevention and control task in combination with the dynamic cost map, and feeds back the execution result to the data acquisition module to realize dynamic correction. The application reduces the false alarm rate and improves the reliability and efficiency of the early prevention and warning of industrial fire.

[0004] To achieve the above-mentioned purpose, the application provides the following technical scheme: The early prevention and warning system for industrial fire based on multi-sensor fusion comprises a data acquisition module, a false alarm discrimination module, a risk zoning module, a fire identification module and a prevention and control optimization module. The data acquisition module obtains the fire data of the industrial area based on a multi-sensor network. The fire data comprises temperature, smoke concentration, flame spectrum and gas concentration data. The false alarm discrimination module uses dynamic false alarm discrimination rules to filter effective fire data based on the fire data. The risk zoning module identifies the fire risk levels of different regions based on the effective fire data and allocates risk labels to each sub-region. The fire identification module identifies the fire development stage based on the effective fire data and the probability of fire occurrence predicted by the sub-regions; The prevention and control optimization module formulates a prevention and control strategy based on the fire development stage, the risk label and the effective fire data, meanwhile, the path planning algorithm of the AGV plans a path based on the sub-regions, executes the prevention and control strategy, and feeds back the execution result to the data acquisition module.

[0005] Specifically, the multi-sensor network adopts a hierarchical distributed architecture, including a bottom sensor node layer, an intermediate wireless transmission layer and a top data processing layer; the bottom sensor node layer is deployed at key positions of the industrial area, each sensor node integrates at least one type of sensor for real-time collection of fire data; the intermediate wireless transmission layer transmits the fire data collected by the bottom sensor node layer to the sink node through ZigBee or LoRa communication protocol; the top data processing layer is used to receive the fire data uploaded by the sink node and perform time synchronization and format standardization processing on the fire data.

[0006] Specifically, the generation method of the dynamic false alarm discrimination rule includes: establishing a historical fire data sample library, the historical fire data sample library includes normal data, false alarm data and real fire data under different working conditions; using a machine learning algorithm to train the historical fire data samples to generate a dynamic false alarm discrimination model, the input of the dynamic false alarm discrimination model is real-time collected fire data, and the output is a data validity determination result.

[0007] Specifically, the dynamic false alarm discrimination rule further includes temperature data rising rate and smoke concentration trend correlation discrimination and flame spectrum signal duration and gas concentration over-standard duration correlation discrimination; The temperature data rising rate and smoke concentration trend correlation discrimination determines that the effective fire data is determined when the temperature rising rate exceeds the preset temperature threshold and the smoke concentration shows exponential growth; The flame spectrum signal duration and gas concentration over-standard duration correlation discrimination determines that the effective fire data is determined when the flame spectrum signal duration exceeds t seconds and the carbon monoxide concentration exceeds the preset lower explosion threshold.

[0008] Specifically, the risk zoning module identifies the fire risk level of different areas based on the effective fire data, and assigns a risk label to each sub-region, including: The risk zoning module receives the effective fire data and gives each effective fire data point a spatial weight to form effective fire data with spatial weight; the spatial weight is calculated based on the geographical position of the sensor; The effective fire data with the space weight is input into a risk heat map generation algorithm, a geographical position of each effective fire data point is taken as a coordinate, the space weight and the sensor reading thereof are taken as weighted values, a risk intensity distribution surface is output through interpolation and smoothing processing, and a risk heat map is formed after visual rendering; According to the preset risk level threshold, the risk heat map is regionally divided, including: automatically identifying a continuous closed region with the risk intensity in the same risk level threshold range, and dividing the continuous risk heat map into a plurality of discrete sub-regions; According to the risk level threshold for dividing the sub-regions, a corresponding risk label is automatically assigned to each generated sub-region, and a dynamic zoning atlas is formed; the risk label includes a first risk, a second risk and a third risk.

[0009] Specifically, the fire identification module identifies a fire development stage based on the effective fire data and the probability of fire occurrence in the sub-region, including: The fire identification module receives the dynamic zoning atlas from the risk zoning module, acquires a geographical boundary of a specified sub-region according to the dynamic zoning atlas, and filters all sensor data located within the geographical boundary from the effective fire data according to the geographical boundary to form a feature data set of the corresponding partition; The feature data set and the risk label are input into a time series prediction model; the time series prediction model is a recurrent neural network model enhanced by an attention mechanism, and different attention weights are dynamically assigned according to the importance of different sensor data in the feature data set for fire prediction; The time series prediction model outputs a probability curve of fire occurrence in a preset time period for the corresponding partition, and identifies a fire development stage according to the slope and key inflection point of the probability curve; the fire development stage includes a smoldering stage, an initial stage of open fire and a spreading stage.

[0010] Specifically, the prevention and control optimization module includes: The prevention and control optimization module receives the probability curve of fire occurrence and the fire development stage from the fire identification module, simultaneously acquires the dynamic zoning atlas from the risk zoning module, and reads the risk label corresponding to the sub-region where the fire occurs from the dynamic zoning atlas; The fire development stage currently identified and the risk label of the corresponding sub-region are taken as a query key to search a preset prevention and control strategy knowledge base, and a corresponding prevention and control action set is matched; the prevention and control strategy knowledge base stores a strategy mapping table, and the strategy mapping table defines a mapping relationship from an input condition combination of the combination of the fire development stage and the risk label to an output predefined prevention and control action set; The matched prevention and control actions are parsed into a structured prevention and control strategy, and elements of the prevention and control strategy include prevention and control actions, resource allocation, and execution sequence; According to the prevention and control strategy and the dynamic partition map, fire extinguishing or isolation task instructions for one or more AGVs are generated; For each AGV that receives the fire extinguishing or isolation task instructions, a path planning algorithm of the AGV is used to calculate an optimal path that avoids a first risk sub-region and leads to a task target position using the dynamic partition map and a real-time position of the AGV as inputs, and the optimal path is bound with the fire extinguishing or isolation task instructions to form an AGV action instruction set and is issued to an AGV executor to drive execution.

[0011] Specifically, the prevention and control strategy in the prevention and control strategy knowledge base includes: For a sub-region in the smoldering stage and with a risk label of a third risk, ventilation and smoke exhaust and early warning notification are started; For a sub-region in the initial stage of open fire and with a risk label of a second risk, an AGV carrying a fire extinguisher is dispatched to start a sprinkler system; For a sub-region in the spreading stage and with a risk label of a first risk, multiple AGVs are dispatched for collaborative operation to perform isolation belt setting and fire extinguishing tasks.

[0012] Specifically, the path planning algorithm of the AGV is specifically: The dynamic partition map is mapped to a dynamic cost map, where each sub-region is regarded as an independent navigation unit; According to the risk label of each sub-region in the dynamic partition map, a passing cost of each sub-region in the dynamic cost map is dynamically calculated; the passing cost exponentially increases with the risk level represented by the risk label; The current position of the AGV and a task target position specified by the prevention and control strategy are obtained at the same time, and the current position and the task target position are used as the starting point and the end point of a path, and a graph search algorithm is used to search on the generated dynamic cost map; the goal of the graph search algorithm is to find a path from the current position to the task target position, so that the sum of the passing costs of all sub-regions traversed by the path, i.e., the cumulative passing cost, is the minimum value among all paths, and then the path with the lowest cumulative passing cost is output as the optimal path for the AGV to execute.

[0013] Specifically, the specific process of feeding the execution result to the data acquisition module includes: During execution of the AGV action instruction set, the AGV acquires environment data after execution through an auxiliary sensor carried by the AGV; Compare the executed environment data with the effective fire data triggering the execution task, and generate a strategy validity evaluation signal; Send the strategy validity evaluation signal to the data acquisition module as a dynamic correction parameter for data acquisition and discrimination.

[0014] Compared with the prior art, the present application has the following advantages: The present application provides an industrial fire early prevention and control early warning system based on multi-sensor fusion, and optimizes the architecture, operation steps and process, which has the advantages of simple process, low investment and operation cost, and low production cost.

[0015] The present application provides an industrial fire early prevention and control early warning system based on multi-sensor fusion, which realizes real-time acquisition and standardization processing of multi-dimensional fire data such as temperature, smoke concentration, flame spectrum and gas concentration through multi-sensor fusion and hierarchical distributed network architecture, solving the problem of single data dimension in traditional systems; combined with dynamic false alarm discrimination rules trained based on historical data, effective fire data is screened through double correlation discrimination, reducing the false alarm rate, and relying on spatial weight and risk heat map, realizing accurate partitioning and dynamic updating of fire risk level.

[0016] The present application provides an industrial fire early prevention and control early warning system based on multi-sensor fusion, which can accurately predict the probability of fire occurrence and identify the development stages of smoldering, initial open fire and spread by means of attention mechanism enhanced recurrent neural network model, and realize targeted matching of prevention and control strategies by combining fire development stage and risk label with strategy mapping mechanism; AGV dynamic optimal path planning avoids the first risk sub-area, improves the execution efficiency, and the execution result feedback forms a dynamic correction closed loop, which optimizes the response speed, resource allocation rationality and execution effect of early fire prevention and control, effectively controls the spread of fire. BRIEF DESCRIPTION OF DRAWINGS

[0017] Figure 1 The present application provides an industrial fire early prevention and control early warning system based on multi-sensor fusion, which can accurately predict the probability of fire occurrence and identify the development stages of smoldering, initial open fire and spread by means of attention mechanism enhanced recurrent neural network model, and realize targeted matching of prevention and control strategies by combining fire development stage and risk label with strategy mapping mechanism; AGV dynamic optimal path planning avoids the first risk sub-area, improves the execution efficiency, and the execution result feedback forms a dynamic correction closed loop, which optimizes the response speed, resource allocation rationality and execution effect of early fire prevention and control, effectively controls the spread of fire. Figure 2 The present application provides an industrial fire early prevention and control early warning system based on multi-sensor fusion, which can accurately predict the probability of fire occurrence and identify the development stages of smoldering, initial open fire and spread by means of attention mechanism enhanced recurrent neural network model, and realize targeted matching of prevention and control strategies by combining fire development stage and risk label with strategy mapping mechanism; AGV dynamic optimal path planning avoids the first risk sub-area, improves the execution efficiency, and the execution result feedback forms a dynamic correction closed loop, which optimizes the response speed, resource allocation rationality and execution effect of early fire prevention and control, effectively controls the spread of fire. Figure 3 The present application provides an industrial fire early prevention and control early warning system based on multi-sensor fusion, which can accurately predict the probability of fire occurrence and identify the development stages of smoldering, initial open fire and spread by means of attention mechanism enhanced recurrent neural network model, and realize targeted matching of prevention and control strategies by combining fire development stage and risk label with strategy mapping mechanism; AGV dynamic optimal path planning avoids the first risk sub-area, improves the execution efficiency, and the execution result feedback forms a dynamic correction closed loop, which optimizes the response speed, resource allocation rationality and execution effect of early fire prevention and control, effectively controls the spread of fire. DETAILED DESCRIPTION

[0018] Example 1: Please refer to Figure 1 The present application provides an embodiment: an industrial fire early prevention and control early warning system based on multi-sensor fusion, which comprises: The data acquisition module, the false alarm discrimination module, the risk partition module, the fire identification module and the prevention and control optimization module; The data acquisition module acquires fire data of the industrial area based on a multi-sensor network; the fire data includes temperature, smoke concentration, flame spectrum and gas concentration data; The multi-sensor network adopts a hierarchical distributed architecture, including a bottom sensor node layer, an intermediate wireless transmission layer and a top data processing layer; the bottom sensor node layer is deployed at key positions of the industrial area, each sensor node integrates at least one type of sensor for real-time acquisition of fire data; the intermediate wireless transmission layer transmits the fire data acquired by the bottom sensor node layer to the sink node through ZigBee or LoRa communication protocol; the top data processing layer is used to receive the fire data uploaded by the sink node and perform time synchronization and format standardization processing on the fire data, wherein ZigBee or LoRa communication protocol is prior art content in the field and is not the inventive scheme of the present application, and will not be described here.

[0019] Further, the key positions of the industrial area refer to areas with high fire risk or areas that need to be monitored, including but not limited to flammable and explosive material storage areas, production operation areas, equipment rooms and other high-incidence fire areas, electrical control rooms, cable trenches, ventilation ducts and other areas prone to hidden fires, as well as warehouse areas, personnel-intensive operation areas and other core areas with large fire impact.

[0020] Further, the types of sensors deployed in the bottom sensor node layer include temperature sensors, smoke sensors, flame spectrum sensors and gas sensors, wherein the temperature sensors are used to acquire ambient temperature and temperature rise rate data, the smoke sensors are used to monitor smoke concentration and change trend, the flame spectrum sensors are used to capture flame characteristic spectrum signals and duration, and the gas sensors are used to detect combustible or toxic gas concentration such as carbon monoxide and determine whether it is over standard.

[0021] The false alarm discrimination module uses dynamic false alarm discrimination rules to filter effective fire data based on the fire data; The risk partition module identifies the fire risk level of different areas based on the effective fire data and assigns a risk label to each sub-area; The fire identification module identifies the fire development stage based on the effective fire data and the probability of fire occurrence in the sub-area; The prevention and control optimization module formulates a prevention and control strategy based on the fire development stage, the risk label and the effective fire data, at the same time, the path planning algorithm of the AGV plans a path based on the sub-area, executes the prevention and control strategy, and feeds back the execution result to the data acquisition module.

[0022] The generation method of the dynamic false alarm discrimination rule comprises: establishing a historical fire data sample library, the historical fire data sample library comprising normal data, false alarm data and real fire data under different working conditions; training the historical fire data sample by using a machine learning algorithm to generate a dynamic false alarm discrimination model, the dynamic false alarm discrimination model inputting real-time collected fire data and outputting a data validity determination result, wherein the machine learning algorithm is prior art content in the field and is not the inventive scheme of the present application, and will not be described here.

[0023] The dynamic false alarm discrimination rule further comprises temperature data rising rate and smoke concentration change trend correlation discrimination and flame spectrum signal duration and gas concentration over-limit duration correlation discrimination. The temperature data rising rate and smoke concentration change trend correlation discrimination determines that it is valid fire data when the temperature rising rate exceeds a preset temperature threshold and the smoke concentration exponentially increases. Further, the preset temperature threshold refers to the abnormal change threshold of the environmental temperature per unit time in an industrial scene, for example, the preset temperature threshold of an ordinary production workshop is 5 ℃ / min, and the preset temperature threshold of a dangerous chemical warehouse area is 3 ℃ / min.

[0024] The flame spectrum signal duration and gas concentration over-limit duration correlation discrimination determines that it is valid fire data when the flame spectrum signal duration exceeds 10 seconds and the carbon monoxide concentration exceeds a preset lower explosion limit threshold. t

[0025] Further, the lower explosion limit threshold refers to the minimum volume percentage concentration of a combustible gas, such as carbon monoxide, that can cause an explosion in air, in the present application, the lower explosion limit threshold of carbon monoxide is set to 12.5%, that is, when the carbon monoxide concentration in air is greater than or equal to 12.5%, it belongs to an explosion danger state.

[0026] Embodiment 2: Please refer to Figure 2 In the present embodiment, the risk zoning module identifies the fire risk level of different regions based on the valid fire data and assigns a risk label to each sub-region, comprising: A1: The risk zoning module receives valid fire data and gives each valid fire data point a spatial weight to form valid fire data with a spatial weight; the spatial weight is calculated based on the geographical position of the sensor and is used to quantify the risk influence degree of the valid fire data point on its surrounding area; Further, the specific steps of A1 comprise: ​(1) The risk zoning module receives valid fire data from the front-end false alarm discrimination module, wherein each valid fire data point carries the accurate geographic location information of its source sensor, thereby constructing a discrete data point set with spatial coordinates inside the risk zoning module; (2) Based on a preset spatial influence radius, the entire discrete data point set is scanned to identify all valid fire data points adjacent to each other within the spatial influence radius, and these points are classified as data point groups with spatial correlation; (3) The system reads the confidence label of each valid fire data point from the false alarm discrimination module, and calculates a basic weight value for each valid fire data point according to the data reliability represented by the confidence label. The higher the confidence, the higher the basic weight value; (4) The spatial context environment of each valid fire data point is examined. For points within the data point group, the basic weight value will be synergistically enhanced because multiple sensors simultaneously abnormal can better represent regional risks. Conversely, for isolated, unrelated valid fire data points within the spatial influence radius, the basic weight value will be systematically reduced to suppress possible local interference; (5) After the above adjustment of the basic weight value, each valid fire data point obtains its final spatial weight.

[0027] A2: Input the valid fire data with spatial weight into the risk heat map generation algorithm, take the geographic location of each valid fire data point as the coordinate, take its spatial weight and sensor reading as the weighted value, output the risk intensity distribution surface through interpolation and smoothing processing, and form the risk heat map after visual rendering, which is used to express the current potential fire risk intensity of each position in the region; Further, the specific steps of A2 include: (1) Obtain valid fire data with spatial weight; (2) Define the physical boundary of the entire industrial area, and cover a regular grid within the physical boundary to discretize the continuous space into multiple calculation units, and the center point of each calculation unit is a target position; (3) Define an influence range for each valid fire data point with spatial weight; the influence range represents the geographic area within which the spatial weight and sensor reading of the data point can have effective influence; (4) Traverse each target position of the calculation unit in the regular grid. For the current target position, the system finds all valid fire data points with spatial weight whose influence range covers it, and these points are regarded as contribution points; (5) For each contribution point, the system calculates its contribution value to the current target position; the contribution value is determined by the product of the sensor reading of the contribution point and its spatial weight, and is positively correlated with the product value, while the contribution value is negatively correlated with the geographical distance from the contribution point to the current target position. The farther the distance, the smaller the influence; (6) The final risk intensity value of the current target position is calculated by weighting and superimposing the contribution values of all contribution points, and a quantitative risk representation value is assigned to each calculation unit; (7) When the risk intensity values of all calculation units in the rule grid are calculated, these discrete numerical points collectively form a continuous risk intensity distribution surface covering the entire industrial area; (8) Map the risk intensity distribution surface to a tangible image, i.e. a risk heat map, which is usually expressed using a color gradient, for example, using cool-toned blue to represent low-risk value areas, and warm-toned yellow, orange, and red to represent medium and high-risk value areas, respectively.

[0028] A3: Regionally divide the risk heat map according to the preset risk level threshold, including: automatically identify the continuous closed area with risk intensity within the same risk level threshold range, and divide the continuous risk heat map into multiple discrete sub-regions; Further, the specific steps of A3 include: (1) Obtain the risk heat map, and call the preset risk level threshold, which divides the risk intensity into continuous intervals of low, medium, and high; (2) Cut the continuous risk intensity distribution surface represented by the risk heat map horizontally along each risk level threshold, and each cut will generate one or more closed contour lines, with any point inside the contour line having a risk intensity greater than or equal to the current cutting threshold; (3) The system processes each closed contour line generated by the cutting, and initially identifies the internal area enclosed by each contour line as a candidate sub-region; (4) Assign a risk label to each candidate sub-region, and the determination rule of the risk label is: observe which risk level threshold the candidate sub-region is cut by, and combine the risk intensity of the internal points to determine which threshold interval they mainly fall into, for example, a region cut by a high-risk threshold and with internal intensity higher than the medium-risk threshold is identified as a first-risk sub-region; (5) Spatial relationship processing and region optimization, including: Check the spatial positions of all candidate sub-regions, and for multiple regions that are not spatially connected but have the same risk label, confirm each as an independent sub-region; For multiple candidate sub-regions that are spatially adjacent and have the same risk label, it is determined whether to merge into a larger and continuous sub-region according to a preset merging rule, such as a distance upper limit and an area lower limit. (6) After all the contour lines are processed and the spatial relationship is clarified, the originally continuous risk heat map is divided into multiple discrete final sub-regions with homogeneous internal risk intensity, and each sub-region has clear geographical boundaries and a unique risk label. All these sub-regions and their attributes together constitute the dynamic zoning atlas required by the system.

[0029] A4: According to the risk level threshold used in the division of the sub-region, a corresponding risk label is automatically assigned to each generated sub-region, forming a dynamic zoning atlas; the risk label includes first risk, second risk and third risk.

[0030] Further, the first risk is high risk, the second risk is medium risk, and the third risk is low risk.

[0031] Further, the specific steps of A4 include: (1) Obtain the sub-region and the corresponding risk level threshold; (2) The system automatically determines the risk level of the sub-region according to the specific risk level threshold used when the boundary of the sub-region is determined, combined with the internal preset mapping rule. For example, a region whose boundary is defined by a high risk threshold is determined to be high risk. (3) The system selects the corresponding text type risk label, such as low risk area, medium risk area or high risk area, from the preset label library according to the determined risk level, and performs binding operation to associate the text type risk label with all geometric properties of the sub-region; the geometric properties include all coordinate point sequences constituting the boundary of the region; (4) After all sub-regions complete the risk label binding, the system integrates these labeled sub-regions, spatial boundary information and their relative position relationship into a unified and structured data object, i.e. dynamic zoning atlas; (5) The newly generated dynamic zoning atlas is standardized and packaged in data format. The dynamic zoning atlas accurately answers the core question of where the risk is high and how the risk level is, which lays the foundation for the spatial situation awareness of the whole system.

[0032] Embodiment 3: Please refer to Figure 3 The fire identification module in this embodiment identifies the fire development stage based on effective fire data and the probability of predicting fire occurrence in the sub-region, including: B1: the fire identification module receives a dynamic partition map from the risk partition module, acquires geographical boundaries of a specified sub-region according to the dynamic partition map, and screens all sensor data located within the geographical boundaries from effective fire data to form a feature data set of the corresponding partition; Further, the specific steps of B1 include: (1) The fire identification module receives the latest dynamic partition map from the front-end risk partition module. The dynamic partition map clearly records all sub-regions divided in the current industrial environment, as well as the accurate geographical boundary coordinates of each sub-region and its corresponding risk label. (2) According to the current analysis requirement or alarm triggering condition, a target sub-region that needs to be focused on is determined from the dynamic partition map. This target sub-region may be a newly appeared high-risk area or a region with rapidly changing risk. Once the target is locked, the system accurately extracts the polygonal geographical boundary that defines the range of the sub-region from the dynamic partition map. The polygonal geographical boundary is composed of a plurality of sequentially connected geographical coordinate points, forming a closed polygonal region in the digital space. (3) Real-time effective fire data stream is acquired. Each effective fire data point carries the unique identity of its source sensor and its accurate geographical position coordinates. (4) For each flowing effective fire data point, a spatial position relationship judgment is made, i.e., whether the geographical position coordinates of the effective fire data point fall within the polygonal geographical boundary of the corresponding target sub-region. If the effective fire data point is determined to be located within the polygonal geographical boundary, it is automatically captured and marked. If the effective fire data point is determined to be located outside the polygonal geographical boundary, it is temporarily ignored in this analysis, obtaining a pure data subset. (5) The pure data subset is aligned and integrated according to time sequence to obtain a feature data set of the specified partition.

[0033] B2: input the feature data set and the risk label into a time series prediction model; the time series prediction model is an attention mechanism enhanced recurrent neural network model, which dynamically allocates different attention weights according to the importance of different sensor data in the feature data set for fire prediction. The recurrent neural network model is a prior art in the field and is not the inventive scheme of the present application, which will not be described here. B3: the time series prediction model outputs a probability curve of fire occurrence in a preset time period for the corresponding partition, and identifies the fire development stage according to the slope and key inflection point of the probability curve. The fire development stage includes the smoldering stage, the initial stage of open fire and the spreading stage.

[0034] Further, the specific identification logic of the fire development stage includes: when the probability curve continuously maintains at a low numerical level, and the slope is close to zero with only a slight fluctuation, it indicates that the risk is in a slow brewing state without significant escalation, and the system identifies it as the smoldering stage; when the probability curve appears the first significant key inflection point, and after the inflection point, the slope of the curve changes to a stable and obvious positive value, which means that the probability starts to continuously and stably increase, which marks the beginning of the stable combustion of open fire, and the system identifies it as the initial stage of open fire; when the probability curve appears the second more steep key inflection point, after the inflection point, the slope of the curve becomes extremely steep, and the probability value almost presents a vertical trend of soaring, which indicates that the fire is breaking through the initial combustion state and entering a rapid spread and difficult to control situation, and the system identifies it as the spreading stage.

[0035] The prevention and control optimization module comprises: C1: The prevention and control optimization module receives the probability curve of fire occurrence and the fire development stage from the fire identification module, obtains the dynamic partition map from the risk partition module, and reads the risk label corresponding to the sub-region where the fire occurs from the dynamic partition map; C2: Taking the currently identified fire development stage and the risk label of the corresponding sub-region as a query key, the preset prevention and control strategy knowledge base is searched to match the corresponding prevention and control action set; the prevention and control strategy knowledge base stores a strategy mapping table, which defines the mapping relationship between the input condition combination of the combination of the fire development stage and the risk label and the output predefined prevention and control action set; It should be understood that the internal core of the prevention and control strategy knowledge base is a strategy mapping table, which can be understood as a large decision dictionary. In this decision dictionary, each legal combination of the fire development stage and the risk label is an independent entry, that is, a unique input condition.

[0036] Further, the specific steps of C2 include: (1) receiving the currently identified fire development stage and the risk label of the corresponding sub-region; (2) combining the received fire development stage and risk label into a standardized query key according to a preset format; (3) accessing the prevention and control strategy knowledge base, comparing the constructed standardized query key with all preset input condition entries in the strategy mapping table, and finding an entry completely consistent with the current query key, for example, when the query key is the combination of the initial stage of open fire and high-risk area, only the entry in the strategy mapping table defined as the initial stage of open fire and high-risk area will be matched, and entries of smoldering stage and high-risk area or other different combinations will not be matched; (4) Once a complete match is found, the matching process is declared successful, and the content corresponding to the matched entry in the policy mapping table, i.e., a predefined set of prevention and control actions, is obtained.

[0037] C3: parsing the matched set of prevention and control actions into a structured prevention and control strategy; the elements of the prevention and control strategy include prevention and control actions, resource allocation, and execution order; Further, the specific steps of C3 include: (1) obtaining the set of prevention and control actions; (2) processing each action entry in the set of prevention and control actions one by one, for each action, the system converts it from an abstract name to a specific, executable task instance, for example, for the action of implementing fire extinguishing, the system parses it into a specific instruction of using a dry powder fire extinguisher for fire fighting, at the same time, the system binds a target location parameter for this task instance, wherein the target location parameter is derived from the accurate coordinates of the sub-region where the fire is located in the dynamic partition map provided by the risk partition module; (3) analyzing each instantiated task to determine the type and quantity of resources required for its execution, for example, the task of using a dry powder fire extinguisher for fire fighting requires the allocation of an AGV loaded with a dry powder fire extinguisher. The system will query the current resource state table, and according to the urgency of the task and the location of the resource, bind a specific, physically existing resource entity to this task, which connects the abstract task with the specific physical resource; (4) analyzing the logical relationships between all tasks that have been instantiated and allocated resources, some tasks must be executed first, for example, establishing an isolation belt may need to be completed before advancing into the fire scene to extinguish the fire to prevent the spread of the fire, according to the preset rules and the internal logic between tasks, a clear and linear execution sequence is planned for all tasks, which clearly defines which task starts first, which task starts later, and whether there is a strict dependency relationship between them; (5) integrating and packaging the three core elements of the instantiated prevention and control actions, the allocated resources, and the determined execution order into a unified data structure to generate the final shaped, structured prevention and control strategy.

[0038] C4: generating fire extinguishing or isolation task instructions for one or more AGVs according to the prevention and control strategy and the dynamic partition map; Further, the specific steps of C4 include: (1) obtaining the structured prevention and control strategy and the dynamic partition map, wherein the structured prevention and control strategy clearly specifies the prevention and control actions to be executed, the type of resource allocated for each action, and the execution order of the actions, and the dynamic partition map depicts the geographical boundaries of all sub-regions and their risk labels; (2) Analyze the serialized prevention and control actions in the prevention and control strategy, and convert each independent and resource-allocated action item into a logical to-be-executed task. For example, the action of using a dry powder fire extinguisher to extinguish a fire in the strategy is initialized as a fire extinguishing task, and then the task is assigned a precise spatial attribute. By cross-referencing the dynamic zoning map, the specific sub-area where the fire occurred is found, and the center coordinates or key point coordinates of the sub-area are obtained from the dynamic zoning map. The coordinates are determined as the target position of the fire extinguishing task; (3) Traverse all the spatialized tasks decomposed from the prevention and control strategy, and select the most suitable AGV execution unit for each task. The matching principle is that the equipment carried by the AGV must meet the task requirements, such as assigning an AGV carrying a fire extinguisher to a fire extinguishing task. Under the premise of meeting the conditions, the idle or schedulable AGV closest to the task target position is preferred to minimize the response time. Once a match is successful, a specific task is formally bound to a specific AGV entity; (4) For each task successfully bound to an AGV, the system generates a highly detailed task instruction for it. This task instruction contains multiple executable commands of mandatory fields, including: a clear instruction type, i.e., a fire extinguishing instruction or an isolation instruction; precise task target position coordinates; specific operation details that need to be executed by the AGV, such as the dosage and duration of fire extinguishing agent spraying, or the number of isolation materials to be placed; and a priority identifier for the task, which is derived from the execution order defined in the prevention and control strategy. High-priority tasks will be prioritized for planning and execution; (5) When all tasks have generated corresponding task instructions, these task instructions are collected and packaged into a task instruction package. The task instruction package clearly indicates the specific responsibilities, destination of action, and operation specifications that each AGV needs to undertake, and finally outputs microscopic and drivable robot instructions.

[0039] C5: For each AGV that has accepted a fire extinguishing or isolation task instruction, use the AGV's path planning algorithm with the dynamic zoning map and the AGV's real-time position as input to calculate the optimal path that avoids the first risk sub-area and leads to the task target position, and bind the optimal path with the fire extinguishing or isolation task instruction to form an AGV action instruction set and issue it to the AGV executor to drive its execution.

[0040] Further, the specific steps of C5 include: (1) When the system has assigned a specific AGV a clear fire extinguishing or isolation task instruction, the AGV's path planning algorithm is activated at this time; (2) Receive the dynamic zoning map from the risk zoning module and convert it into a dynamic cost map specifically for path planning; (3) Obtain the AGV’s current precise position in real time from its own positioning system, and use it as the starting point for path planning. At the same time, the algorithm extracts the target position from the fire extinguishing or isolation task instructions that have been issued, and uses it as the endpoint for path planning. (4) On the constructed dynamic cost map, with the determined starting point and ending point as endpoints, find a path from the starting point to the ending point where the sum of the passage costs of all navigation units traversed by the trajectory is minimized. This sum is called the cumulative passage cost. Since the sub-regions with the first risk are given extremely high passage costs, the path with the lowest cumulative passage cost found by this algorithm is naturally an optimal path that can effectively avoid all sub-regions with the first risk. (5) Bind and encapsulate the calculated optimal path with the fire extinguishing or isolation task instructions initially assigned to the AGV. The optimal path solves the problem of how to arrive safely, while the task instructions define what to do after arrival. The combination of the two forms a self-contained, closed-loop AGV action instruction set. This AGV action instruction set not only contains all waypoint sequences of the path, but also contains the specific operation commands to be executed after arriving at the target point. (6) The AGV action instruction set is sent to the on-board actuator of the corresponding AGV through the communication network. The actuator parses the AGV action instruction set, including: first, controlling the AGV to move autonomously to the task target location along the planned optimal path. After confirming arrival, the actuator triggers the task operation specified in the AGV action instruction set, such as turning on the fire extinguishing device to spray or releasing isolation materials.

[0041] The prevention and control strategies in the knowledge base include: For sub-areas that are in the smoldering stage and are classified as risk level 3, ventilation and smoke extraction and early warning notifications will be activated. For sub-areas in the initial stage of open flame and with a risk label of second risk, dispatch AGVs to carry fire extinguishers and activate the surrounding sprinkler system; For sub-areas that are in the spread stage and are labeled as having the highest risk, multiple AGVs are dispatched to work together to set up firebreaks and carry out key firefighting tasks.

[0042] The path planning algorithm for the AGV is as follows: D1: Map the dynamic partition map to a dynamic cost map, wherein each sub-region is regarded as an independent navigation unit; Furthermore, the process of mapping the dynamic partition map to a dynamic cost map includes: (1) Receive dynamic partitioning map from the risk partitioning module; (2) Establish the corresponding relationship between the navigation unit and the cost structure, including: each independent sub-region in the dynamic zoning atlas is directly defined as a basic navigation unit in the dynamic cost map used for internal path planning, and at the same time, the system internally presets a cost mapping rule, the core content of which is to strongly associate the passing cost of a sub-region with its risk label, and stipulate that the association is an exponential growth relationship, that is, the passing cost will increase by several times or even dozens of times for each increase of one level of risk grade; (3) Perform cost quantization calculation and assignment, including: traversing each navigation unit in the dynamic cost map, that is, each sub-region, for the sub-region currently being processed, read its risk label, and then calculate a specific, quantized passing cost value according to the above-mentioned preset exponential cost mapping rule; (4) After the calculation is completed, assign this calculated specific passing cost value to the navigation unit in the dynamic cost map; (5) When all navigation units have been successfully assigned with quantized passing costs, a brand new dynamic cost map dedicated to path planning is formally generated, which is completely consistent with the original dynamic zoning atlas in terms of spatial range, but its connotation has fundamentally changed: it no longer only indicates where the risk is, but accurately quantizes the risk cost to be borne for traversing different regions.

[0043] D2: dynamically calculating a passing cost for each sub-region in the dynamic cost map according to the risk label of each sub-region in the dynamic zoning atlas; the passing cost increases exponentially with the risk level represented by the risk label; It should be noted that the system internally presets a cost configuration table, the core function of which is to map textually described risk labels, such as low risk, medium risk, and high risk, into basic cost coefficients for mathematical calculation. These coefficients are not the final cost value, but the basic multiplier for calculation, and their numerical values are strictly set to follow the law of exponential growth.

[0044] Further, the specific steps of D2 include: (1) Establish a mapping relationship between risk levels and basic cost coefficients, which is recorded in the cost configuration table and explicitly stipulates that the basic cost coefficient increases exponentially with the risk level, and at the same time, the system sets a universal reference cost value; (2) Obtain the dynamic cost map; each sub-region in the dynamic cost map has been defined as a navigation unit; (3) Traverse each navigation unit in the dynamic cost map, for the navigation unit currently being processed, the system accurately obtains the risk label assigned to the navigation unit by querying the dynamic zoning atlas; (4) For the navigation unit currently being processed, a basic cost coefficient corresponding thereto is found according to the risk label obtained and the cost configuration table, and then a preset reference cost value is taken as a base number, and the basic cost coefficient found is taken as an index to perform an exponent operation, so as to calculate a final passing cost value of the navigation unit; (5) The passing cost value calculated is assigned to the current navigation unit.

[0045] D3: The current position of the AGV and the task target position specified by the prevention and control strategy are obtained at the same time, and the current position and the task target position are taken as the starting point and the ending point of the path. A graph search algorithm is used to search on the generated dynamic cost map. The goal of the graph search algorithm is to find a path from the current position to the task target position, so that the sum of the passing costs of all sub-regions traversed by the path, i.e. the cumulative passing cost, is the minimum value among all possible paths. Then the path with the lowest cumulative passing cost is output as the optimal path for the AGV to execute.

[0046] Further, the specific steps of D3 include: (1) Load the assigned dynamic cost map, and confirm the starting point and the ending point of this navigation. Then, model the assigned dynamic cost map into a graph structure, where the sub-regions are nodes, and the connectivity relationship of adjacent regions is an edge. At the same time, initialize an open set and a closed set, and maintain the known minimum cumulative passing cost and the estimated total cost of reaching the ending point for each node. The initialized open set contains only the starting point node, and the initialized closed set is empty; (2) Select the node with the minimum estimated total cost from the open set as the current node. If the current node is the ending point, then backtrack from the ending point to the starting point. The node sequence traversed by the backtracking path is output as the optimal path with the lowest cumulative passing cost. Otherwise, identify all adjacent nodes of the current node; (3) For each adjacent node, calculate the cumulative passing cost of the new path via the current node, i.e. the known minimum cumulative passing cost of the current node plus the passing cost of the adjacent node itself. If the new cost is lower than the known minimum cumulative passing cost originally recorded by the adjacent node, then update the known minimum cumulative passing cost of the node to the new lower value, and set the predecessor node of the adjacent node to the current node. Then, put the adjacent node into the open set; (4) Move the processed current node into the closed set, and repeat (2)-(3) until the ending point is selected as the current node; (5) When the ending point is selected, backtrack from the ending point to the starting point through the predecessor node pointers of each node. The node sequence traversed by the backtracking path is output as the optimal path with the lowest cumulative passing cost.

[0047] The execution result is fed back to the specific process of the data acquisition module, which comprises: E1: In the process of executing the AGV action instruction set, the AGV collects the executed environment data through the auxiliary sensor carried thereon; E2: Comparing the executed environment data with the effective fire data triggering this execution task to generate a strategy effectiveness evaluation signal; Further, the specific steps of E2 comprise: (1) The system acquires data of two key time points: one is the executed environment data returned by the AGV executor through the auxiliary sensor carried thereon after completing the task; the other is the effective fire data triggering this action originally archived by the system, the system extracts the same set of key fire indicators such as temperature and specific gas concentration from the environment data and the effective fire data triggering this action, and ensures that the types, monitoring positions and statistical caliber of the key fire indicators are completely consistent; (2) For each set of aligned key fire indicators, the system calculates the value in the executed environment data, the change amount and change direction thereof relative to the corresponding value in the effective fire data, such as decrease, increase and maintain unchanged, and comprehensively analyzes the change amount and change direction of all key fire indicators to form a comprehensive situation change portrait; (3) The expected target of the executed prevention and control strategy is retrieved, such as extinguishing the open fire, then the obtained situation change portrait is logically matched with the expected target to judge whether the actual situation change conforms to the expectation, for example, if the target is to extinguish the open fire, the key fire indicators show that the temperature and smoke are greatly decreased, then the conformity is high; (4) Based on the conformity judgment result, a structured strategy effectiveness evaluation signal is generated, the strategy effectiveness evaluation signal contains the overall effectiveness conclusion such as effective, partially effective or ineffective, and can be accompanied by detailed information such as specific unmet indicators, and finally outputs the strategy effectiveness evaluation signal.

[0048] E3: The strategy effectiveness evaluation signal is sent to the data acquisition module as a dynamic correction parameter for data acquisition and discrimination.

[0049] The embodiments of the application are described above in combination with the drawings, but the application is not limited to the above specific embodiments, the above specific embodiments are only illustrative but not restrictive, and those skilled in the art can make changes, modifications, replacements and variations to the above embodiments without departing from the purpose of the application and the protected scope, which are all within the protection of the application.

Claims

1. An industrial fire early prevention and warning system based on multi-sensor fusion, characterized in that, include: Data acquisition module, false alarm detection module, risk zoning module, fire identification module, and prevention and control optimization module; The data acquisition module acquires fire data from the industrial area based on a multi-sensor network; the fire data includes temperature, smoke concentration, flame spectrum, and gas concentration data. The false alarm discrimination module uses dynamic false alarm discrimination rules to filter valid fire data based on fire data. The risk zoning module identifies the fire risk level of different areas based on valid fire data and assigns a risk label to each sub-area; The fire identification module identifies the fire development stage based on valid fire data and the probability of fire occurrence predicted for sub-regions. The prevention and control optimization module formulates prevention and control strategies based on the fire development stage, risk labels, and effective fire data. At the same time, the AGV's path planning algorithm plans paths based on sub-regions, executes the prevention and control strategies, and feeds back the execution results to the data acquisition module.

2. The industrial fire early prevention and warning system based on multi-sensor fusion as described in claim 1, characterized in that, The multi-sensor network adopts a layered distributed architecture, including: a bottom sensor node layer, an intermediate wireless transmission layer, and a top data processing layer. The bottom sensor node layer is deployed in key locations in the industrial area, and each sensor node integrates at least one type of sensor for real-time fire data acquisition. The intermediate wireless transmission layer transmits the fire data acquired by the bottom sensor node layer to the aggregation node via ZigBee or LoRa communication protocols. The top data processing layer receives the fire data uploaded by the aggregation node and performs time synchronization and format standardization processing on the fire data.

3. The industrial fire early prevention and warning system based on multi-sensor fusion as described in claim 1, characterized in that, The method for generating the dynamic false alarm discrimination rule includes: establishing a historical fire data sample library, which includes normal data, false alarm data and real fire data under different operating conditions; training the historical fire data sample library with a machine learning algorithm to generate a dynamic false alarm discrimination model, wherein the input of the dynamic false alarm discrimination model is the real-time collected fire data, and the output is the data validity judgment result.

4. The industrial fire early prevention and warning system based on multi-sensor fusion as described in claim 3, characterized in that, The dynamic false alarm discrimination rules also include the correlation between the rate of increase of temperature data and the trend of smoke concentration change, and the correlation between the duration of flame spectral signal and the duration of gas concentration exceeding the standard. The rate of temperature rise is correlated with the trend of smoke concentration change. When the rate of temperature rise exceeds a preset temperature threshold and the smoke concentration increases exponentially, it is determined to be valid fire data. The duration of the flame spectral signal is correlated with the duration of gas concentration exceeding the limit. When the duration of the flame spectral signal exceeds... t When the carbon monoxide concentration exceeds the preset lower explosion limit threshold, it is considered valid fire data.

5. The industrial fire early prevention and warning system based on multi-sensor fusion as described in claim 1, characterized in that, The risk zoning module identifies the fire hazard level of different areas based on valid fire data and assigns a risk label to each sub-area, including: The risk zoning module receives valid fire data and assigns a spatial weight to each valid fire data point, forming valid fire data with spatial weights; the spatial weights are calculated based on the geographic location of the sensors. Valid fire data with spatial weights are input into the risk heat map generation algorithm. The geographical location of each valid fire data point is used as the coordinate, and its spatial weight and sensor readings are used as the weighting values. Through interpolation and smoothing, the risk intensity distribution surface is output, and a risk heat map is formed after visualization rendering. The risk heatmap is divided into regions based on a preset risk level threshold, including: automatically identifying continuous closed regions with risk intensity within the same risk level threshold range, and dividing the continuous risk heatmap into multiple discrete sub-regions; Based on the risk level threshold used when dividing the sub-regions, a corresponding risk label is automatically assigned to each generated sub-region to form a dynamic partition map; the risk label includes first risk, second risk, and third risk.

6. The industrial fire early prevention and warning system based on multi-sensor fusion as described in claim 5, characterized in that, The fire identification module identifies the fire development stage based on valid fire data and the predicted probability of fire occurrence in sub-regions, including: The fire identification module receives a dynamic partition map from the risk partitioning module, obtains the geographical boundary of a specified sub-region based on the dynamic partition map, and filters out all sensor data located within the geographical boundary from the valid fire data based on the geographical boundary to form the feature dataset of the corresponding partition. The feature dataset and the risk label are input into the time series prediction model; the time series prediction model is an attention-enhanced recurrent neural network model, which dynamically allocates different attention weights according to the importance of different sensor data in the feature dataset to fire prediction; The time series prediction model outputs the probability curve of fire occurrence in the corresponding partition within a preset time period, and identifies the fire development stage based on the slope and key inflection points of the probability curve; the fire development stage includes smoldering stage, initial open flame stage and spread stage.

7. The industrial fire early prevention and warning system based on multi-sensor fusion as described in claim 6, characterized in that, The prevention and control optimization module includes: The prevention and control optimization module receives the probability curve of fire occurrence and the fire development stage from the fire identification module, and simultaneously obtains the dynamic zoning map from the risk zoning module, and reads the risk label corresponding to the sub-region where the fire occurred from the dynamic zoning map. The currently identified fire development stage and the corresponding risk label of the sub-area are used as query keys to retrieve the preset prevention and control strategy knowledge base and obtain the corresponding set of prevention and control actions. The prevention and control strategy knowledge base stores a strategy mapping table, which defines the mapping relationship between the input condition combination of fire development stage and risk label and the output predefined set of prevention and control actions. The matched set of prevention and control actions is parsed into a structured prevention and control strategy; the elements of the prevention and control strategy include prevention and control actions, resource allocation, and execution order. Based on the prevention and control strategy and the dynamic partition map, generate fire extinguishing or isolation task instructions for one or more AGVs; For each AGV that receives a fire extinguishing or isolation task instruction, the AGV path planning algorithm uses the dynamic partition map and the AGV's real-time position as input to calculate the optimal path that avoids the first risk sub-area and leads to the task target location. The optimal path is then bound to the fire extinguishing or isolation task instruction to form an AGV action instruction set, which is then sent to the AGV actuator to drive its execution.

8. The industrial fire early prevention and warning system based on multi-sensor fusion as described in claim 7, characterized in that, The prevention and control strategies in the knowledge base include: For sub-areas that are in the smoldering stage and are classified as risk level 3, ventilation and smoke extraction and early warning notifications will be activated. For sub-areas in the initial stage of open flame and with a risk label of second risk, dispatch AGVs to carry fire extinguishers and activate the sprinkler system; For sub-areas that are in the spread stage and are labeled as the first risk level, multiple AGVs are dispatched to work together to perform tasks such as setting up firebreaks and extinguishing fires.

9. The industrial fire early prevention and warning system based on multi-sensor fusion as described in claim 8, characterized in that, The path planning algorithm for the AGV is as follows: The dynamic partition map is mapped to a dynamic cost map, wherein each sub-region is regarded as an independent navigation unit; Based on the risk label of each sub-region in the dynamic zoning map, the passage cost is dynamically calculated for each sub-region in the dynamic cost map; the passage cost increases exponentially with the risk level represented by the risk label. Simultaneously, the current position of the AGV and the target position specified by the control strategy are obtained. Using the current position and the target position as the start and end points of the path, a graph search algorithm is used to search on the generated dynamic cost map. The goal of the graph search algorithm is to find a path from the current position to the target position such that the sum of the travel costs of all sub-regions traversed by the path, i.e., the cumulative travel cost, is the minimum among all paths. The path with the lowest cumulative travel cost is then output as the optimal path for the AGV to execute.

10. The industrial fire early prevention and warning system based on multi-sensor fusion as described in claim 9, characterized in that, The specific process of feeding back the execution results to the data acquisition module includes: During the execution of AGV action command sets, the AGV collects environmental data after execution through its onboard auxiliary sensors; The environmental data after execution is compared with the valid fire data that triggered the execution task to generate a strategy effectiveness evaluation signal; The effectiveness evaluation signal of the strategy is sent to the data acquisition module as a dynamic correction parameter for data acquisition and discrimination.

Images