Multi-source data fire-fighting facility risk grade assessment and early warning decision-making method

Abstract

The invention discloses a multi-source data fire-fighting facility risk level assessment and early warning decision-making method, and particularly relates to the technical field of intelligent fire-fighting alarm devices and public safety, and the method comprises the steps: S1, collecting the state of a fire-fighting facility and environment dynamic data in real time, S2, calculating the static health degree index of the facility through an uncertainty reasoning model and a topological relation, and S3, calculating the risk level of the fire-fighting facility according to the static health degree index. The method comprises the steps of S1, generating an environment dynamic risk degree index through an adaptive evaluation model, S4, fusing the two indexes through a nonlinear coupling function to generate a comprehensive risk level and a visual map, and S5, matching, verifying and deducing a recommendation decision scheme with a utility score based on risk information. And a grading early warning signal is generated so as to drive a corresponding alarm device to execute early warning. According to the method, the defects of static facility assessment and dynamic environment risk separation are overcome, accurate risk quantification and prospective early warning are realized, and the intelligent level and execution reliability of alarm response are remarkably improved through an alarm decision closed loop and a self-optimization mechanism.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of intelligent fire alarm devices and public safety technology, more specifically, the present application relates to a fire facility risk level evaluation and early warning decision method of multi-source data. BACKGROUND

[0002] With the rapid development of Internet of Things, big data and artificial intelligence technology, intelligent fire protection has become an important direction to improve the level of public safety. The current fire alarm device and monitoring system is usually based on Internet of Things technology, which can collect real-time operation state data and environmental parameters of various fire facilities. At the same time, by integrating building information model, video monitoring and patrol records and other multi-source information, it provides a data basis for realizing digital management of fire safety.

[0003] However, the existing alarm device and system mainly have three limitations in application: 1. The alarm evaluation dimension is single, and mostly stays in the binary judgment of facility normal / failure or the independent threshold triggering alarm of environmental parameters, which cannot quantitatively evaluate the overall health degree of facility system and the topological dependence relationship, cannot identify the systemic risk caused by the implicit failure of key equipment, and leads to inaccurate or missed alarm signals; 2. The risk model has poor adaptability. The risk evaluation in the alarm system mostly depends on fixed rules or a large amount of historical data, which is difficult to dynamically adapt to the real risk characteristics of different scenes and time periods, has weak generalization ability, and affects the timeliness and reliability of the alarm; 3. The early warning and decision support are disconnected. The existing alarm device mostly only provides risk alarm signals, cannot automatically generate reliable and executable emergency plans combined with real-time facility availability status, leads to response delay and insufficient decision-making scientificity after alarm, and cannot realize the closed loop from alarm to intelligent disposal.

[0004] In view of the above situation, the present application provides a fire facility risk level evaluation and early warning decision method of multi-source data, aiming at improving the intelligent level and decision support ability of the fire alarm device. SUMMARY

[0005] In order to overcome the above-mentioned defects of the prior art, the present application provides a fire facility risk level evaluation and early warning decision method of multi-source data to solve the problems proposed in the background technology.

[0006] In order to achieve the above-mentioned purpose, the present application provides the following technical scheme: a fire facility risk level evaluation and early warning decision method of multi-source data, comprising the following steps: S1, real-time collection of state data and environmental dynamic data of fire facilities to form multi-source heterogeneous data; S2, based on the state data of the fire-fighting facilities, using an uncertainty reasoning model and facility topology relationship analysis, a facility static health degree index is calculated; S3, based on environmental dynamic data, real-time evaluation is performed through an adaptive evaluation model to obtain an environmental dynamic risk degree index; S4, the facility static health degree index and the environmental dynamic risk degree index are nonlinearly coupled to calculate a comprehensive risk level and a visualized risk map; S5, the comprehensive risk level, risk position and key failure facility information are input into a risk-preplan matching engine, at least one recommended decision scheme with an effectiveness score is matched and deduced from a structured preplan library, and a hierarchical early warning signal output and an alarm device linkage are executed.

[0007] Preferably, in step S1, the state data of the fire-fighting facilities includes operating state parameters, failure records, maintenance history and installation topology relationship; The environmental dynamic data includes real-time sensing data, building information model data, unit patrol data and public government data.

[0008] Preferably, step S2 specifically includes: Multiple state data of a single fire-fighting facility are input as evidence, and an uncertainty reasoning model is used to calculate the confidence distribution of the facility belonging to different health states; According to the physical connection and functional dependence topology relationship of the fire-fighting facility system, the confidence distributions of the associated facilities are weighted and aggregated, and finally a quantitative facility static health degree index is output.

[0009] Preferably, the adaptive evaluation model in step S3 is used to adaptively adjust the evaluation weights of different risk factors according to the characteristics of the environmental dynamic data collected at the current time, to real-time generate and apply risk evaluation rules, and to output an environmental dynamic risk degree index.

[0010] Preferably, the nonlinear coupling calculation in step S4 is realized by a coupling function ; Wherein, is a comprehensive risk value, is a facility static health degree index, is an environmental dynamic risk degree index, is a correction coefficient preset according to the importance of the facility, which is positively correlated with the criticality of the facility in the fire-fighting system.

[0011] Preferably, the risk map generated in step S4 is a visualized heat map based on a building information model or an electronic map, wherein a high-risk area and a key failure facility identifier causing the risk to rise are synchronously associated and displayed.

[0012] Preferably, the structured preplan library in step S5 is composed of a plurality of atomic operation instructions, each of which is marked with a triggering precondition, a required facility resource state and an expected execution effect. The matching process of the risk-preplan matching engine includes: preliminary screening according to the comprehensive risk level and the position, and checking whether the preconditions of the atomic operation instructions in the candidate preplan match the current facility health state, and filtering out the preplans that cannot be executed due to the dependence on facility failure.

[0013] Preferably, the deduction process in step S5 is: for the candidate preplan that passes the verification, simulating the execution of the atomic operation instruction sequence contained therein, estimating the execution time, resource consumption and influence on personnel safety, so as to calculate the utility score of the preplan.

[0014] Preferably, the method further comprises step S6: collecting actual execution feedback data of the early warning decision scheme, and using the feedback data and corresponding historical evaluation data to continuously optimize the non-linear coupling calculation function and the preplan utility score model through the reinforcement learning model.

[0015] Technical effects and advantages of the present application: 1、The present application introduces facility static health index and environment dynamic risk index, and uses non-linear coupling function for deep fusion calculation, so that the model can quantitatively reflect the core logic that the decline of facility reliability will amplify the environmental risk, so that when the environmental parameters have not reached the traditional alarm threshold, but the key facilities have hidden failures, the alarm system can generate risk warning signals in advance and accurately, significantly improving the accuracy and foresight of the alarm, effectively solving the problem of missing potential high-risk scenes of traditional alarm devices; 2、The present application can adapt to different scenes and time periods flexibly by dynamically adjusting the risk factor weight through the adaptive evaluation model, without relying on a large amount of preset historical data, improving the adaptability and real-time performance of the alarm model. At the same time, by constructing a structured atomic operation preplan library and driving the risk-preplan matching engine to perform real-time verification and deduction, invalid preplans that cannot be executed due to the dependence on facility failure can be automatically filtered out, and optimized decision schemes with utility scores are simulated and deduced, realizing the intelligent alarm decision-making leap from single passive alarm to active provision of executable and verifiable decision support, greatly shortening the alarm response decision-making time and greatly improving the success rate of preplan execution; 3. This invention introduces a reinforcement learning-based feedback optimization mechanism to continuously collect actual execution feedback data of early warning decisions. This data is then used to iteratively optimize the core coupled calculation function and the contingency plan utility scoring model. This allows the alarm system to continuously absorb actual operational experience and adaptively adjust evaluation parameters and decision preferences. As a result, the risk assessment model becomes increasingly aligned with real-world scenarios, ensuring that the alarm system maintains high reliability, advanced technology, and low maintenance costs in complex and ever-changing environments. This achieves continuous self-optimization of the alarm and decision-making closed loop. Attached Figure Description

[0016] Fig. 1 This is a flowchart illustrating the steps of the method of the present invention.

[0017] Fig. 2 This is a flowchart illustrating the overall operation of the method of the present invention. Detailed Implementation

[0018] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention.

[0019] This invention provides a method for risk level assessment and early warning decision-making of fire protection facilities based on multi-source data. This method can be executed by a fire alarm decision-making system, and its overall process is as follows: Figs. 1-2 As shown, the main steps include: S1. Real-time collection of status data of fire protection facilities and dynamic environmental data to form multi-source heterogeneous data; This step forms the foundational data input layer of the alarm system. It collects the following two types of data in real-time or near real-time through fire safety IoT gateways, system log APIs, manual data entry platforms, and government data interfaces: Status data of fire protection facilities: including but not limited to: Operating status parameters: such as the start / stop status of the water pump, current, water pressure, communication status of the alarm host, and the operating voltage and signal value of the detector; Fault records and maintenance history: Equipment fault codes, occurrence time, historical maintenance records, and the date and results of the most recent maintenance extracted from the fire control system log; Installation topology: The logical relationships of facilities extracted from building information models or fire protection design drawings, such as the connection relationships between water pumps, valves and sprinkler heads in fire water supply networks, and the hierarchical relationships between controllers, loops and detectors in fire alarm systems; Environmental dynamic data: including but not limited to: Real-time sensor data: Temperature, humidity, smoke concentration, and carbon monoxide concentration acquired through IoT sensors, and flame and smoke feature recognition results obtained through video analysis; Building Information Modeling (BIM) data: A 3D model of a building, including information such as spatial layout, fire compartments, evacuation routes, and the location of important equipment; Unit inspection data: Daily inspection results reported through the mobile inspection APP, such as records of potential hazards such as blocked safety exits and insufficient fire extinguisher pressure; Public government data: Information such as building usage, peak periods of population density, and location of surrounding fire hydrants obtained through the government data platform interface.

[0020] S2. Based on the status data of fire protection facilities, the static health index of the facilities is calculated by using an uncertainty reasoning model and facility topology analysis. This step aims to quantitatively assess the reliability of the fire protection system itself. Its core principle is to move beyond simple fault / normal judgments and adopt a health concept, which specifically includes: S2.1 Unit-level health confidence calculation: For a single fire protection facility unit (such as a smoke detector), its multiple state data (such as signal value S, self-test error count E, and installation environment dust coefficient D) are used as evidence input into the uncertainty reasoning model. In a preferred embodiment, an evidence reasoning algorithm is adopted. First, a confidence allocation rule corresponding to the evaluation level of health, sub-health, and failure is set for each piece of evidence. For example, when the signal value S is within the standard range, its confidence level for the health level is 0.8. Then, the Dempster combination rule is used to fuse the confidence levels of all evidence to finally obtain the confidence distribution of the facility unit belonging to each health level, such as: health = 0.7, sub-health = 0.2, failure = 0.1, which means that the detector has a 70% confidence level of being healthy. S2.2 System-level health aggregation: Based on the facility topology obtained in step S1, the confidence levels of functionally related units are aggregated. For example, if the sprinkler system of a fire compartment is set to consist of one water pump, five valves, and several sprinkler heads, then the failure confidence of the water pump will significantly reduce the overall health of the sprinkler system. A weighted average method can be used for aggregation, with the weights set according to the criticality of the facility in the subsystem. Finally, a normalized and scalarized static health index of the facility is output. Its range is [0,1], and the closer it is to 1, the healthier and more reliable it is overall.

[0021] S3. Based on dynamic environmental data, an adaptive assessment model is used to conduct real-time assessment to obtain the dynamic environmental risk index. This step is used to assess the immediate likelihood of a fire caused by the external environment. The core of the adaptive assessment model lies in its dynamically adjustable weights, which do not rely on a fixed historical data model. The specific process is as follows: S3.1 Feature extraction and normalization: Extract features from real-time sensing data (such as temperature T and smoke concentration C) (e.g., calculate the rate of temperature rise dT / dt) and normalize all feature values. S3.2 Dynamic Weight Evaluation and Adaptive Evaluation Model (In one embodiment, a lightweight optimization algorithm based on real-time data feedback, such as a simplified particle swarm optimization algorithm, can be used to dynamically adjust the rule weights). Based on the feature combination at the current moment, the weights of each risk factor are automatically calculated. For example, in a library / archives scenario, the weight of smoke particle features is increased; in a power distribution room scenario, the weight of the temperature change rate is increased. Then, based on the current weights and feature values, an environmental dynamic risk index is calculated. The value range is also [0,1]. The larger the value, the higher the environmental risk of fire.

[0022] S4. Nonlinearly couple the facility static health index with the environmental dynamic risk index to generate a comprehensive risk level and a visualized risk map. This step is one of the key innovations of this invention, achieving a deep integration of inherent reliability and external threats. The specific process is as follows: S4.1 Coupled calculation, using a nonlinear coupling function for calculation: ,in, For the comprehensive risk value, The static health index of the facility obtained in step S2. This refers to the environmental dynamic risk index obtained in step S3. This is a facility importance correction factor, the value of which is positively correlated with the criticality of the facility in the fire protection system, and is preset by expert knowledge or system importance analysis; For example, the core fire pump It can be set to 0.8, for ordinary emergency lighting. It can be set to 0.2. The physical meaning of this function is: when the facility's health... During descent, Increased, leading to the same environmental risks This is significantly amplified, which aligns with the engineering logic that the less reliable the equipment, the more serious the consequences could be even from a small fire source threat. S4.2 Risk map generation: The calculated comprehensive risk value is then used to generate the risk map. The risk level is mapped to four levels: red, orange, yellow, and blue. At the same time, a visual risk map is generated by combining the building information model. This map renders the risk distribution on the building's plan or 3D model in the form of a heat map, and displays the identification of key fault facilities that cause high risk (such as flashing a sprinkler pump icon with extremely low health). This provides intuitive spatial and status information for subsequent alarms and decisions.

[0023] S5. Input the comprehensive risk level, risk location and key fault facility information into the risk-contingency plan matching engine, match and deduce at least one recommended decision-making scheme with an effectiveness score from the structured contingency plan library, and execute the graded early warning signal output and alarm device linkage accordingly. This step enables an intelligent leap from risk assessment to alarm decision support, and its workflow is as follows: S5.1, Structured Contingency Plan Library: The contingency plan library does not store text files, but rather consists of a large number of atomic operation instructions. Each instruction describes a basic action and is annotated with metadata, for example: Command ID: ACT 001 Description: Start exhaust fan No. 1. Triggering conditions: Overall risk level ≥ Orange and fire confirmation status = Confirmed; Required resource status: Health status of exhaust fan No. 1 > 0.6; Expected results: Visibility in the target area is expected to increase by 50% within 60 seconds; S5.2 Matching and Verification: The risk-contingency plan matching engine receives the output of step S4. First, it matches the contingency plan template according to the risk level and location. Then, the key step is to verify whether the required resource status of all atomic operation instructions in the template matches the actual facility health status obtained from step S2. The relevant data is matched. For example, if the plan requires the start of spray pump No. 1, but the system detects that the health of the pump is only 0.2 (high confidence of failure), then the plan will be filtered out to ensure the feasibility of the recommended plan. S5.3, Deduction and Scoring: For the contingency plan that has passed the verification, the engine simulates its execution process, estimates the execution time, resource consumption, potential impact on personnel evacuation, etc., and calculates a utility score through a utility function. S5.4 Early Warning and Linkage Decision Output: Ultimately, the system outputs 1-3 recommended decision schemes with effectiveness scores and success rates, and triggers tiered early warnings. The early warning decision output module simultaneously performs the following operations: Based on the comprehensive risk level, it drives the corresponding area's audible and visual alarm devices to issue corresponding level audiovisual alarms; it updates the risk map on the graphical alarm platform and highlights the decision suggestions; and it sends early warning instructions containing risk location, key fault information, and recommended decision schemes to the responsible person via SMS, application push, etc. Example information is as follows: The dynamic risk of the warehouse on the east side of the third floor has increased (level: orange), and the health of the area's sprinkler system is insufficient. Recommended decisions: 1. Prioritize sending personnel to verify on-site (estimated success rate 95%); 2. Linkage activation of nearby fireproof roller shutters (estimated success rate 90%). S6. Collect actual implementation feedback data of the early warning decision-making plan, and use the feedback data and corresponding historical evaluation data to continuously optimize the nonlinear coupling calculation function and the plan utility scoring model through reinforcement learning model. This step enables the alarm decision-making system to self-evolve. For example, when multiple feedbacks show that the recommended activation plan for a certain wind turbine under a certain risk mode fails repeatedly due to insufficient on-site personnel, the reinforcement learning model will gradually reduce the utility score of such plans under that mode, or adjust the parameters in the coupling function. This will make future assessments and alarm decisions more aligned with actual circumstances.

[0024] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for risk level assessment and early warning decision-making of fire protection facilities based on multi-source data, characterized by: Includes the following steps: S1. Real-time collection of status data of fire protection facilities and dynamic environmental data to form multi-source heterogeneous data; S2. Based on the status data of fire protection facilities, the static health index of the facilities is calculated by using an uncertainty reasoning model and facility topology analysis. S3. Based on dynamic environmental data, an adaptive assessment model is used to conduct real-time assessment to obtain the dynamic environmental risk index. S4. Nonlinearly couple the facility static health index with the environmental dynamic risk index to generate a comprehensive risk level and a visualized risk map. S5. Input the comprehensive risk level, risk location and key fault facility information into the risk-contingency plan matching engine, match and deduce at least one recommended decision-making scheme with an effectiveness score from the structured contingency plan library, and execute the output of graded early warning signals and linkage with alarm devices.

2. The method for assessing and issuing early warning decisions based on multi-source data for fire protection facilities, as described in claim 1, is characterized in that: In step S1, the status data of the fire protection facilities includes operating status parameters, fault records, maintenance history, and installation topology. Environmental dynamic data includes real-time sensor data, building information modeling data, unit inspection data, and public administration data.

3. The method for risk level assessment and early warning decision-making of fire protection facilities based on multi-source data according to claim 1, characterized in that: Step S2 specifically includes: Using multiple state data of a single fire protection facility as evidence input, the confidence distribution of the facility belonging to different health states is calculated using an uncertainty reasoning model; Based on the physical connections and functional dependencies of the fire protection system, the confidence distribution of related facilities is weighted and aggregated to finally output a quantitative static health index of the facilities.

4. The method for assessing and making early warning decisions on the risk level of fire protection facilities based on multi-source data according to claim 1, characterized in that: The adaptive assessment model in step S3 is used to adaptively adjust the assessment weights of different risk factors based on the characteristics of the environmental dynamic data collected at the current moment, generate and apply risk assessment rules in real time, and output the environmental dynamic risk index.

5. The method for risk level assessment and early warning decision-making of fire protection facilities based on multi-source data according to claim 1, characterized in that: The nonlinear coupling calculation in step S4 is performed using the coupling function. accomplish; in, For the comprehensive risk value, The static health index of the facility. This is an environmental dynamic risk index. This is a correction factor preset based on the importance of the facility, and its value is positively correlated with the criticality of the facility in the fire protection system.

6. The method for risk level assessment and early warning decision-making of fire protection facilities based on multi-source data according to claim 1, characterized in that: The risk map generated in step S4 is a visual heat map based on building information model or electronic map, in which high-risk areas and key fault facilities that cause the risk to increase are displayed synchronously.

7. The method for assessing and making early warning decisions on the risk level of fire protection facilities based on multi-source data according to claim 1, characterized in that: The structured contingency plan library in step S5 consists of multiple atomic operation instructions, each of which is marked with triggering preconditions, required facility resource status, and expected execution effect; The risk-contingency matching engine's matching process includes: initial screening based on comprehensive risk level and location, then verifying whether the preconditions of atomic operation instructions in candidate contingency plans match the current facility health status, and filtering out contingency plans that cannot be executed due to facility failure.

8. The method for risk level assessment and early warning decision-making of fire protection facilities based on multi-source data according to claim 1, characterized in that: The deduction process in step S5 is as follows: for the candidate plan that has passed the verification, simulate the execution of its atomic operation instruction sequence, estimate its execution time, resource consumption and impact on personnel safety, and thus calculate the utility score of the plan.

9. The method for risk level assessment and early warning decision-making of fire protection facilities based on multi-source data according to claim 1, characterized in that: The method also includes step S6: collecting actual implementation feedback data of the early warning decision-making scheme, and using the feedback data and corresponding historical evaluation data to continuously optimize the nonlinear coupling calculation function and the scheme utility scoring model through a reinforcement learning model.

Images