Coal mine goaf fire situation early warning method based on multi-source data fusion

Abstract

The invention relates to the technical field of coal mine safety monitoring, and discloses a coal mine goaf fire situation early warning method based on multi-source data fusion, and the method comprises the steps: firstly obtaining oxygen, carbon monoxide and other multi-source environmental parameters deployed at a monitoring section, and constructing evidence vectors for different fire states according to a preset mapping relation; then sequential fusion is carried out on the evidence vectors, conflict coefficients among the evidences are calculated in the fusion process, and according to the conflict coefficients, a direct fusion strategy, a fusion correction strategy or a check blocking strategy is adopted in a self-adaptive mode to obtain a comprehensive confidence degree result; and finally, determining a current fire development stage based on a maximum confidence criterion and validity verification, and executing corresponding grading early warning measures. According to the method, high-conflict data caused by sensor faults or environmental interference can be effectively processed, the fusion paradox problem of a traditional evidence theory is solved, and the anti-interference capability and accuracy of goaf fire situation monitoring are remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of coal mine safety monitoring technology, specifically to a method for early warning of coal mine goaf fires based on multi-source data fusion. Background Technology

[0002] Spontaneous combustion of residual coal in coal mine goaf areas is one of the major hazards threatening safe production in mines. Early and accurate monitoring and warning of such fires are of great significance for protecting the lives of miners and the safety of equipment and property. Currently, coal mine goaf fire monitoring mainly relies on various sensors deployed at monitoring sections to collect environmental parameters such as oxygen, carbon monoxide, and temperature to determine fire risk.

[0003] However, due to the uneven distribution of air leakage flow fields and the complex environment within goaf areas, single-category environmental parameters often only reflect local environmental characteristics and are insufficient to comprehensively characterize the overall evolution of fires. Furthermore, different types of sensor data possess different dimensions and physical properties. Existing monitoring methods primarily focus on independent threshold judgments for each parameter, making it difficult to achieve effective complementarity and unified quantification between heterogeneous data, resulting in low reliability of monitoring results under complex operating conditions.

[0004] To address the limitations of single data sources, multi-source data fusion technology has been increasingly applied in fire early warning systems. Among these, the Data Synthesis (DS) evidence theory has gained attention due to its advantages in handling uncertain information. However, existing evidence fusion methods typically employ fixed combination rules for calculation. When sensor malfunctions or strong environmental interference leads to high conflicts between collected data, traditional combination rules can easily produce fusion results that contradict the facts. Furthermore, existing fusion algorithms lack dynamic identification and adaptive processing mechanisms for highly conflicting evidence, resulting in insufficient anti-interference capabilities when facing abnormal data and a tendency to make incorrect decisions.

[0005] Furthermore, in the final situation assessment stage, existing technologies often directly output results based on the maximum probability value, lacking secondary verification of the validity and uncertainty of the decision-making results. In critical states where fire characteristics are not yet obvious or evidence is insufficient, blind decision-making can easily lead to false alarms. At the same time, traditional early warning systems mostly employ a single alarm trigger mode, lacking a tiered response mechanism that matches the different development stages of a fire from its inception to its full extent, thus failing to achieve refined management from early detection to emergency response. Summary of the Invention

[0006] To address the shortcomings of existing technologies, this invention provides a method for early warning of coal mine goaf fires based on multi-source data fusion, which solves the problems of low reliability of single sensor data and decision-making errors caused by high-conflict evidence during multi-source data fusion in existing coal mine goaf fire monitoring.

[0007] The first aspect of this invention provides a method for early warning of coal mine goaf fires based on multi-source data fusion, comprising the following steps: Multi-source environmental parameters deployed at monitoring sections in goaf areas are acquired using a multi-source sensing module. These multi-source environmental parameters include oxygen concentration, carbon monoxide concentration, acetylene concentration, methane concentration, and ambient temperature. The data transmission module transmits the data collected by the multi-source sensing module to the fusion computing module. Using the fusion computing module, based on the preset parameter range and fire state mapping relationship, the multi-source environmental parameters are converted into basic probability allocation functions for different fire states, and multiple independent evidence vectors are constructed. The multiple independent evidence vectors are sequentially fused using a fusion calculation module. During the fusion process, the conflict coefficient between the evidence is calculated. Based on the magnitude of the conflict coefficient, strategies such as direct fusion, modified fusion, or blocking verification are adopted to obtain a comprehensive confidence result that characterizes the current fire situation. The situation warning module uses the maximum confidence criterion to determine the comprehensive confidence result, thereby identifying the current fire development stage in the goaf area, and then implements corresponding graded warning measures based on the determination result.

[0008] Preferably, the acquisition of multi-source environmental parameters deployed at the monitoring section of the goaf is achieved by using a sensor array deployed in the oxidation zone and return air corner of the goaf. The sensor array includes electrochemical oxygen concentration sensors, electrochemical carbon monoxide concentration sensors, gas chromatography or laser spectroscopy acetylene concentration sensors, infrared absorption or catalytic combustion methane concentration sensors, and thermocouple or fiber optic temperature sensors. By deploying multiple types of sensors in specific key areas, characteristic signals of the early stages of a fire can be captured from different physical dimensions.

[0009] Preferably, after acquiring the multi-source environmental parameters, the collected data is preprocessed. Specifically, this includes denoising and normalizing the multi-source environmental parameters to construct a monitoring vector containing values ​​for oxygen concentration, carbon monoxide concentration, acetylene concentration, methane concentration, and ambient temperature. Each component in this monitoring vector corresponds to the real-time value of the aforementioned multi-source environmental parameters. This step aims to eliminate sensor noise interference and dimensional differences, providing standardized input data for subsequent data fusion.

[0010] Preferably, the fire state mapping relationship is constructed based on a predefined fire situation identification framework, which includes four mutually exclusive state elements: normal state, nascent stage, development stage, and open flame stage. By establishing clear state divisions, the system can discretize continuously changing monitoring data into specific situation descriptions.

[0011] Preferably, the process of converting multi-source environmental parameters into basic probability assignment functions involves: pre-setting membership models for each parameter within different numerical ranges for each fire state; for each environmental parameter input, by consulting the membership models, converting the physical environmental parameter into a basic probability assignment function defined on the identification framework. Further, by converting the monitored values ​​of each type of environmental parameter through the pre-set membership models, five independent evidence vectors are generated, corresponding to oxygen concentration, carbon monoxide concentration, acetylene concentration, methane concentration, and ambient temperature, respectively. This process achieves a semantic mapping from physical quantities to confidence information.

[0012] Preferably, the sequential fusion process of evidence vectors adopts an iterative mechanism: initialize the fusion sequence, take the first set of evidence vectors as the initial cumulative fusion result; iteratively introduce subsequent evidence vectors, and calculate the conflict coefficient between the current cumulative fusion result and the newly introduced evidence vector at each introduction; determine and execute the corresponding fusion strategy according to the magnitude of the conflict coefficient.

[0013] Preferably, this invention proposes a hierarchical conflict handling mechanism, specifically including: when the conflict coefficient is less than or equal to a first preset threshold, it is determined to be a low-conflict state, at which point the evidence from each sensor has high consistency, and the Dempster combination rule is used for direct fusion to strengthen the consistent evidence; when the conflict coefficient is greater than the first preset threshold and less than or equal to a second preset threshold, it is determined to be a moderate-conflict state, at which point there is some divergence among the sensors, and weighted average evidence is used for correction before fusion, thereby reducing the impact of local anomalies on the overall judgment; when the conflict coefficient is greater than the second preset threshold, it is determined to be a high-conflict state, indicating that there is a serious deviation between the current sensor data and the accumulated results, triggering a blocking verification mechanism, temporarily removing the newly introduced evidence vector and maintaining the fusion result of the previous moment. This mechanism effectively solves the problem that traditional evidence theory is prone to paradoxes or misjudgments when dealing with high-conflict evidence, and improves the robustness of the system in complex environments.

[0014] Preferably, the process of determining the target state based on the maximum confidence criterion includes: identifying the target state with the highest basic probability allocation value from the comprehensive confidence results; verifying the validity of this target state; and only when the validity verification passes is the target state determined as the current stage of fire development. The specific verification conditions are: the basic probability allocation value of the target state is greater than the sum of the basic probability allocation value of the second most probable state and a preset discrimination threshold, and the uncertainty probability value in the comprehensive confidence results is less than a preset uncertainty tolerance threshold. This dual verification mechanism ensures that the final judgment result not only has the highest probability but also sufficient discrimination and certainty, avoiding ambiguous false alarms.

[0015] Preferably, the system executes tiered early warning measures based on the judgment results: when the condition is judged to be normal, the regular monitoring mode is maintained; when the condition is judged to be in the nascent stage, a blue early warning response is triggered, the sensor sampling frequency is increased, and attention information is pushed to the on-duty personnel; when the condition is judged to be in the development stage, a yellow early warning response is triggered, the audible and visual alarm is activated, and a manual inspection instruction is generated; when the condition is judged to be in the open flame stage, a red early warning response is triggered, and the system performs coordinated control to cut off the power supply to non-intrinsically safe equipment, activate the area fire extinguishing device, and issue an emergency evacuation instruction for personnel.

[0016] A second aspect of the present invention provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the method described in the first aspect above.

[0017] A third aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of the method described in the first aspect above.

[0018] This invention provides a method for early warning of coal mine goaf fires based on multi-source data fusion. It has the following beneficial effects: 1. This invention establishes a mapping relationship between multi-source environmental parameters and fire status, transforming heterogeneous physical monitoring values ​​such as oxygen and carbon monoxide into a unified probability allocation function based on evidence theory; it shields the differences in dimensions and properties of data from different sensors, solves the problems of strong one-sidedness and low reliability of single-indicator monitoring, and can comprehensively reflect the complex fire evolution situation in the goaf from multiple dimensions.

[0019] 2. This invention proposes an adaptive sequential fusion strategy based on the conflict coefficient, which can dynamically select three processing methods—direct fusion, weighted correction, or blocking verification—according to the degree of conflict between evidence. This overcomes the shortcomings of traditional evidence theory, which is prone to paradoxical results when faced with sensor failures or highly conflicting data, ensuring the stability of the fusion results under complex operating conditions and improving the system's anti-interference capability.

[0020] 3. This invention introduces a dual validity verification mechanism in the decision-making stage and constructs a hierarchical early warning and response system; by constraining the discrimination threshold and uncertainty tolerance threshold, it avoids misjudgment when fire characteristics are not obvious or evidence is insufficient; at the same time, it automatically triggers differentiated measures from push notification of information of concern to linkage control based on the judgment result, so as to achieve precise control over the entire process of coal mine goaf fire from its inception to its disaster. Attached Figure Description

[0021] Figure 1 This is a schematic diagram of the structure of the present invention; Figure 2 This is a schematic diagram of the process of the present invention.

[0022] Among them, 101 is the multi-source sensing module; 102 is the data transmission module; 103 is the fusion computing module; and 104 is the situation warning module. Detailed Implementation

[0023] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0024] See attached document Figure 1 The present invention provides a coal mine goaf fire situation early warning system based on multi-source data fusion. The system includes: a multi-source sensing module 101, a data transmission module 102, a fusion computing module 103, and a situation early warning module 104.

[0025] The multi-source sensing module 101 is used to collect environmental parameters inside the goaf in real time. The multi-source sensing module 101 is deployed at preset monitoring sections in the heat dissipation zone, oxidation zone, and asphyxiation zone of the goaf. Specifically, the multi-source sensing module 101 includes an oxygen concentration sensor, a carbon monoxide concentration sensor, an acetylene concentration sensor, a methane concentration sensor, and a temperature sensor. Each sensor independently collects its corresponding physical quantity data, forming a multi-dimensional monitoring vector.

[0026] The data transmission module 102 is connected to the multi-source sensing module 101. The data transmission module 102 converts the raw analog signals collected by the multi-source sensing module 101 into digital signals and transmits the data to the ground monitoring center via a mining fiber optic industrial ring network or a wireless sensor network. The data transmission module 102 verifies the transmitted data to ensure its integrity and real-time performance.

[0027] The fusion computing module 103 is the core processing unit of the system, deployed in the ground monitoring host or edge computing server. The fusion computing module 103 receives raw monitoring data uploaded by the data transmission module 102. The fusion computing module 103 has a pre-built computational program based on DS evidence theory. This program includes a basic probability allocation unit, a conflict detection unit, and an adaptive fusion unit. The fusion computing module 103 performs data cleaning, probability mapping, conflict coefficient calculation, and multi-source evidence fusion operations, outputting the comprehensive confidence level of the current goaf fire status.

[0028] The situation warning module 104 is connected to the fusion computing module 103. The situation warning module 104 determines the current fire development stage based on the comprehensive confidence level output by the fusion computing module 103. The situation warning module 104 stores a tiered warning strategy and generates corresponding control commands based on the determined normal state, nascent stage, development stage, or open flame stage. The situation warning module 104 connects to and controls the underground audible and visual alarms, the ground display terminal, and the area fire extinguishing system.

[0029] See attached document Figure 2 This invention provides a method for early warning of coal mine goaf fires based on multi-source data fusion, comprising the following steps: S1, acquire multi-source environmental parameters deployed at the monitoring section of the goaf area, the multi-source environmental parameters including oxygen concentration, carbon monoxide concentration, acetylene concentration, methane concentration and ambient temperature; S2, based on the preset parameter range and fire state mapping relationship, the multi-source environmental parameters are converted into basic probability allocation functions for different fire states, and multiple independent evidence vectors are constructed. S3, sequentially fuse multiple sets of evidence vectors, calculate the conflict coefficient between evidence during the fusion process, and adopt strategies such as direct fusion, modified fusion or blocking verification according to the magnitude of the conflict coefficient to obtain a comprehensive confidence result that represents the current fire situation; S4. Based on the maximum confidence criterion, the comprehensive confidence result is judged to determine the current fire development stage in the goaf area, and corresponding graded early warning measures are implemented according to the judgment result.

[0030] The following section will elaborate on the specific technical implementation details, calculation formulas, and parameter definitions in steps S1 to S4, based on the above process.

[0031] In step S1, multi-source environmental parameters deployed at the monitoring section of the goaf are acquired. This step specifically includes continuous acquisition, signal conversion, and vectorization processing of gas composition and thermal parameters inside the goaf.

[0032] The acquisition of multi-source environmental parameters relies on sensor arrays deployed in specific areas of the goaf. Goaf areas are typically divided into heat dissipation zones, oxidation zones, and asphyxiation zones. Sensors are mainly deployed at key monitoring sections such as the oxidation zone and return air corners. For five physical quantities—oxygen concentration, carbon monoxide concentration, acetylene concentration, methane concentration, and ambient temperature—corresponding sensing elements are used for detection. For oxygen and carbon monoxide concentrations, electrochemical sensors are used, utilizing the current signal generated by the oxidation-reduction reaction of gases in an electrolytic cell to invert the gas concentration. For methane concentration, infrared absorption or catalytic combustion sensors are used. For acetylene concentration, gas chromatography or laser spectroscopy sensors are used. For ambient temperature, thermocouples or fiber optic temperature sensors are used. Each sensor converts the detected analog electrical signals into digital signals via analog-to-digital converters and transmits them to the data processing unit through a mine communication interface. The sensor selection, signal conditioning circuitry, and implementation of the underlying communication protocol are well-known technologies to those skilled in the art and will not be elaborated upon here.

[0033] After receiving the real-time values ​​of the above five types of parameters, the data processing unit performs noise reduction and normalization processing to construct the monitoring vector for the current moment. The monitoring vector is set as follows: This vector contains five components, represented as: ; in This indicates the oxygen concentration value. This indicates the carbon monoxide concentration value. This indicates the acetylene concentration value. This indicates the methane concentration value. Represents ambient temperature value, monitoring vector It serves as the direct input source for subsequent DS evidence theory fusion algorithms.

[0034] In step S2, based on the preset parameter range and fire state mapping relationship, the multi-source environmental parameters are converted into basic probability allocation functions for different fire states. Before executing this step, the system predefines an identification framework for the fire situation in the goaf. The identification framework is the set of all possible outcomes in the DS evidence theory, and the elements within the set must satisfy mutual exclusivity.

[0035] Based on the evolution of spontaneous combustion of coal in coal mine goaf areas, a fire situation identification framework is defined as follows: It contains four mutually exclusive fire development state elements, and its mathematical expression is: ; in, This indicates a normal state, where there is no obvious oxidation of the residual coal in the goaf, and all parameters are within the background range. At this time, the gas concentration and ambient temperature in the monitoring environment are within the fluctuation range of the mine background values, and there is no risk of spontaneous combustion fire.

[0036] This indicates the nascent stage, at which point the remaining coal begins to slowly oxidize, and some gas parameters show slight anomalies, but no obvious heat accumulation effect has yet formed; this stage is the best early intervention window for fire warning.

[0037] This indicates the development stage. In this state, the oxidation reaction intensity of coal is significantly enhanced, and the rate of heat release from oxidation exceeds the rate of heat dissipation, leading to the continuous accumulation of heat inside the goaf. At this time, the concentration of characteristic gases such as carbon monoxide shows an exponential increase or large fluctuations, and high-temperature cracking gases such as acetylene may begin to appear in trace amounts. The ambient temperature monitoring value is significantly higher than the normal ground temperature, and the risk of fire is high. If no suppression measures are taken, it is easy to transform into an open flame state. This indicates the open flame stage, at which point the coal has undergone intense oxidation and combustion, exhibiting open flame characteristics. Not only is the carbon monoxide concentration extremely high, but the concentrations of hydrocarbon gases such as acetylene and ethylene also rise sharply. The ambient temperature exceeds the coal's ignition point or the sensor's upper limit, accompanied by smoke production, requiring the immediate activation of emergency fire extinguishing and disaster avoidance procedures.

[0038] Based on the above identification framework In this embodiment, each monitoring parameter is pre-built. For each state in different numerical ranges The membership model was developed by collecting historical monitoring data from the mine and statistically analyzing the distribution characteristics of each parameter under known fire conditions. For each sensor input... By consulting the pre-defined membership model, it is transformed into a definition within the recognition framework. The basic probability assignment function on is denoted as . .function Satisfying the basic axioms of the DS evidence theory: ; ; In the formula, To represent the empty set, Representation of recognition framework any subset, Indicates the first Evidence provided by individual sensors The level of trust. In this embodiment, to reduce computational complexity and focus on single-state decisions, Mainly taking single-element sets and the complete series .

[0039] Specifically, the system targets vectors The five components generate five independent evidence vectors, namely: evidence vectors corresponding to oxygen concentrations. Evidence vector corresponding to carbon monoxide concentration Evidence vector corresponding to acetylene concentration Evidence vector corresponding to methane concentration and the corresponding evidence vector of ambient temperature Each set of evidence vectors contains probability assignment values ​​for the normal state, nascent stage, development stage, open flame stage, and uncertain state. For example, its specific form is expressed as: ; Through the above transformation, the multi-source heterogeneous environmental parameters at the physical level are uniformly mapped into isomorphic probabilistic evidence at the decision level, laying the data foundation for subsequent conflict detection and fusion computing.

[0040] In step S2, establishing a specific mapping relationship between multi-source environmental parameters and the basic probability allocation function of fire status is crucial for achieving situational awareness. This step discretizes the continuously changing physical monitoring values ​​using a pre-set expert knowledge base or historical database and assigns them confidence levels for each fire status. For the feature vector... Each component in The system generates corresponding independent evidence vectors through piecewise function calculation. The specific parameter settings are shown in the table below.

[0041] Table 1:

[0042] Table 2:

[0043] Table 3:

[0044] Table 4:

[0045] Table 5:

[0046] After the above processing, the physical monitoring vector at any sampling time is... It is transformed into five sets of normalized basic probability assignment vectors Each set of vectors quantifies the independent judgment result of the corresponding sensor on the current fire situation in the goaf.

[0047] In step S3, multiple sets of evidence vectors are sequentially fused, and dynamic conflict detection and adaptive processing are performed during the fusion process. This step aims to address the uncertainties and contradictions of multi-source heterogeneous data in the spatiotemporal dimensions, ensuring the stability of the final decision. This embodiment adopts a serial fusion architecture, incorporating the evidence vectors of oxygen, carbon monoxide, acetylene, methane, and temperature into the fusion calculation system according to the preset priority or physical order of the sensors.

[0048] Step S3 specifically includes the following sub-steps: S31, Initialize the fusion sequence. Define the cumulative fusion result as... At the start of the fusion process, the evidence vector corresponding to the oxygen concentration is... Assign the value to the initial cumulative fusion result, i.e. Subsequently, the system enters an iterative calculation process, sequentially introducing subsequent evidence vectors. ,in The values ​​range from 2 to 5, respectively, corresponding to to .

[0049] S32, calculate the conflict coefficient between the current accumulated evidence and the newly introduced evidence. Let the current accumulated fusion result be... The newly introduced evidence is The system calculates the conflict coefficient between the two based on the following formula. : ; In the formula, and They represent the recognition framework respectively. Subset elements in This represents the empty set. This indicates the current accumulated fusion evidence. In China, support The basic probability allocation value; Indicates newly introduced evidence In China, support The basic probability assignment value. This formula represents the sum of the probability products of all focal element combinations with empty intersections. The range of values ​​is Its numerical value directly reflects the degree of contradiction between the two sources of evidence regarding the assessment of the fire situation. The closer it is to 1, the greater the difference in judgment between the sensors.

[0050] S33, based on the calculated conflict coefficient The fusion processing logic is divided into three branches, and the system automatically determines and executes the corresponding strategies: Branch 1: When At this point, the system determines that the current evidence is in a low-conflict state, and the perceptions of the fire situation by each sensor are basically consistent. Then, the system uses the Dempster combination rule for direct fusion to calculate and generate new fused evidence. The calculation formula is as follows: ; In the formula, For identification framework The non-empty subset of represents the target proposition supported after fusion; Indicates support after fusion The basic probability distribution value. Coefficient. As a normalization factor, it is used to eliminate the probability of the empty set and redistribute the probability quality of conflicting parts to other propositions, thereby enhancing the consensus part of multi-source evidence.

[0051] Branch 2: When At this point, the evidence is determined to be in a state of moderate conflict. Directly applying the Dempster rule at this stage might lead to illogical results, so the system executes a corrective fusion strategy. The corrective process involves pre-setting weighting coefficients based on the historical accuracy of the sensors or the environmental signal-to-noise ratio. Let... The weight is , The weight is Constructing a revised weighted average evidence : ; In the formula, Indicates support after weighted average The basic probability allocation value. Then, using this weighted average evidence... Instead of the original evidence, it is substituted into the Dempster combination rule formula in the above branch one for self-fusion, thereby reducing the negative impact of conflict on decision-making while retaining the main characteristics of the evidence.

[0052] Branch 3: When At this point, the system determines that the current evidence is in a state of high conflict. This is highly likely due to sensor malfunction, data transmission errors, or extreme anomalies in the local environment. Instead of performing mathematical fusion calculations, the system triggers a blocking verification mechanism. Specifically, this involves: maintaining the fusion result from the previous moment unchanged and temporarily removing any newly introduced evidence. It then sends a fault self-check command for the corresponding sensor to the ground control center. Once the sensor has completed calibration or the data has returned to normal, it is reintegrated into the fusion sequence.

[0053] S34, determine whether the fusion of all parameters is complete. If the current... Then let The result is equal to the fusion result output in step S33, and will Increment the value by 1, return to step S32; if This indicates that the five parameters—oxygen, carbon monoxide, acetylene, methane, and temperature—have all been processed. At this point, the final cumulative fusion result is marked as... This result is the confidence vector representing the overall fire situation in the goaf at the current moment.

[0054] Through the sequential fusion and hierarchical conflict handling described above, this embodiment can effectively filter environmental noise interference and achieve accurate quantitative assessment of fire hazards in goaf areas while ensuring data consistency.

[0055] In step S4, the comprehensive confidence result is determined based on the maximum confidence criterion. The core of this step lies in transforming the probability vector output by the DS evidence theory into a deterministic single-state description, thereby establishing a clear correspondence between the mathematical calculation results and the physical fire scenario, providing a decision-making basis for subsequent engineering control.

[0056] Specifically, the system receives the final cumulative fusion result output in step S3. The result is based on a recognition framework. The probability distribution vectors correspond to the probability allocation values ​​for the normal state, nascent stage, development stage, open flame stage, and uncertain state, respectively. The judgment logic first performs a maximum value search operation, traversing the recognition framework. Given all elements in a subset, find the target state with the largest basic probability assignment value. Let the target state be... It satisfies the following mathematical relationship: ; In the formula, This represents the selected target state, which belongs to the recognition framework. Element; This indicates the operation of finding the maximum value. These represent the probability values ​​for supporting the normal state, the nascent stage, the development stage, and the open flame stage in the final fusion result, respectively.

[0057] To ensure the reliability of decisions and avoid misjudgments caused by data fluctuations, after determining... Afterwards, the system will perform a validity check. The validation process requires both of the following conditions to be met simultaneously: Condition one: ; Condition two: ; In the formula, This represents the probability value corresponding to the state that ranks second in probability value among all elements of the single subset; This represents a preset discrimination threshold, used to ensure that there is a significant difference between the most probable state and the second most probable state; This represents the probability value of uncertainty in the final fusion result; This represents the preset uncertainty tolerance threshold. The system can only lock when the above conditions are met. The current final judgment result will be determined; otherwise, the system will determine the current state as uncertain and maintain the judgment result of the previous moment or prompt manual intervention.

[0058] After confirming the validity of the judgment result, the system proceeds according to... The specific physical state mapping is performed as follows: If correspond The system determines that the current goaf is in a normal state, indicating that all environmental parameters within the area are within safe ranges, and no significant exothermic oxidation reaction has occurred in the coal; if correspond The system determines that the current goaf is in its nascent stage, indicating that localized coal oxidation has begun, with abnormal fluctuations in gas concentration, but no sustained heat accumulation has yet occurred; if correspond The system determines that the current goaf is in the development stage, characterized by an accelerated oxidation reaction rate, heat accumulation leading to a significant increase in ambient temperature, and the continuous release of characteristic hydrocarbon gases; if correspond The system determined that the current goaf is in the open flame stage, indicating that a violent combustion reaction has occurred in the area, posing an extremely high safety risk.

[0059] In step S4, the system executes a tiered early warning response mechanism based on the determined fire development stage. This mechanism, through a preset logical control strategy, translates the decision-making level's state judgment into specific physical operations at the execution level, thereby achieving targeted control over different hazard levels.

[0060] When the judgment result is During this period, the system maintains the normal monitoring mode. The multi-source sensing module 101 maintains the default sampling frequency, for example, setting the sampling period to once every 10 minutes. The data transmission module 102 only uploads the current real-time monitoring values ​​to the database for archiving and storage, without providing audible or visual alerts on the monitoring terminal. The system primarily performs background data trend analysis and updates the historical sample library for subsequent correction of the baseline parameters of the BPA mapping model.

[0061] When the judgment result is At this time, the system triggers a blue alert response. The control center sends a command to the downhole multi-source sensing module 101 to adjust the sensor's operating mode, increasing the sampling frequency from the default period to a higher frequency period, such as setting it to once per minute, to capture minute dynamic changes in environmental parameters. Simultaneously, a notification window pops up on the display interface of the ground monitoring host, and a notification is sent to on-duty personnel via SMS or instant messaging software. The system initiates a trend prediction subroutine, focusing on monitoring the rate of change in carbon monoxide and oxygen concentrations, but at this time, it does not trigger on-site physical alarm devices or activate fire extinguishing equipment.

[0062] When the judgment result is At this time, the system triggers a yellow alert response. The system outputs control signals to the audible and visual alarms on the surface and underground, emitting intermittent alarm sounds and flashing yellow lights to indicate a significant fire hazard in the area. Simultaneously, the system generates an inspection instruction, requiring the underground safety officer to manually check the relevant monitoring sections to confirm whether the sensors are experiencing false alarms due to water or dust accumulation. At the control level, the system sends a preparatory signal to the control valves of the nitrogen injection system, checks the pipeline pressure status, prepares for the immediate activation of fire prevention and extinguishing measures, and calculates and recommends the optimal pressure equalization ventilation scheme on the display terminal based on real-time airflow data to suppress air leakage and oxygen supply in the goaf.

[0063] When the judgment result is Upon activation, the system triggers a red alert response. This response level corresponds to the highest priority emergency linkage control. The system immediately outputs a cut-off command to the power supply switch of the underground power system, cutting off the power supply to non-intrinsically safe electrical equipment in the goaf return airway and related work areas to prevent electrical sparks from igniting gas. Simultaneously, the system sends a start signal to the area fire extinguishing device, automatically opening the electric valves of the nitrogen injection pipeline or grouting pipeline to inject inert gas or grout into the goaf to reduce oxygen concentration and extinguish open flames. In addition, the system issues an emergency evacuation command through the underground broadcast communication system and displays the optimal disaster avoidance route on the monitoring screen to assist underground personnel in quickly evacuating to a safe area.

[0064] Through the aforementioned tiered response mechanism, this embodiment achieves closed-loop management from monitoring and perception to execution and control, ensuring that appropriate response measures are taken at different stages of fire evolution.

[0065] This embodiment selects a set of actual monitoring data from the oxidation zone of the goaf in a fully mechanized longwall mining face of a coal mine as input to verify the system's processing logic in a complex data environment.

[0066] Suppose that at some point during system operation, the multi-source sensing module 101 collects a set of environmental parameters. The specific values ​​of these parameters are: oxygen concentration 18.5%, carbon monoxide concentration 35 ppm, acetylene concentration 0.2 ppm, methane concentration 0.8%, and ambient temperature 65 degrees Celsius.

[0067] The system first performs a data mapping operation. Based on the mapping relationship between the parameter range and the fire state defined in step S2 above, the above physical quantities are converted into a basic probability allocation function defined on the identification framework.

[0068] For an oxygen concentration of 18.5%, it falls within the range supporting the development stage, and the generated evidence vector assigns the majority of its confidence to this stage, while allocating smaller probability values ​​to other states. Similarly, for a carbon monoxide concentration of 35 ppm, it also falls within the range supporting the development stage, and the generated evidence vector primarily supports this stage. For an acetylene concentration of 0.2 ppm, due to its strong affinity for high-temperature open flames, it falls within the range primarily supporting the open flame stage. Therefore, this evidence vector assigns most of its confidence to the open flame stage, a significant difference from the evidence for oxygen and carbon monoxide. For a methane concentration of 0.8%, the generated evidence vector is relatively flat, with confidence primarily allocated to the nascent and development stages. For an ambient temperature of 65 degrees Celsius, it falls within the range supporting the development stage, and the generated evidence vector primarily supports this stage.

[0069] The system then proceeds to a sequential fusion calculation process. In the first round of fusion, evidence for oxygen and carbon monoxide is processed. Since both have a high probability of supporting the development stage, the conflict coefficients calculated by the system are less than the preset low-conflict threshold. The system determines this to be a low-conflict state and directly applies the standard Dempster combination rule for fusion. The intermediate results after fusion show a significant increase in confidence for the development stage, demonstrating the consistency-enhancing effect of multi-source evidence.

[0070] In the second round of fusion, evidence from acetylene data was introduced. At this point, the accumulated fusion results strongly supported the development stage, while the newly introduced acetylene evidence strongly supported the open flame stage. The system calculated the conflict coefficient between the two, obtaining a value in the moderate conflict range, indicating a significant contradiction in the evidence. The system triggered a revised fusion strategy, calculating a weighted average of the evidence based on preset sensor weights. Subsequently, this weighted average evidence was used for fusion calculation, yielding a revised result incorporating acetylene information. In this result, although the development stage still dominated, the probability of the open flame stage increased compared to the previous round, accurately reflecting the potential deterioration trend.

[0071] The system continued to fuse the evidence for methane and temperature sequentially. Since the temperature evidence again strongly supported the development stage, the confidence level for the development stage was further increased in subsequent fusion steps, while also correcting some of the uncertainty introduced by the acetylene evidence. Finally, after five rounds of fusion calculations, a comprehensive confidence vector was output. In this example, the final results are as follows: 0.78 for the development stage, 0.15 for the open flame stage, 0.05 for the nascent stage, 0.01 for the normal state, and 0.01 for uncertainty.

[0072] Based on the maximum confidence criterion, the system identifies the development stage as having the maximum confidence level and satisfying the difference threshold condition between it and the second most probable state. Therefore, the system determines that the goaf is currently in the development stage.

[0073] Based on this judgment, the situation warning module 104 automatically matches and executes the yellow warning response logic. The system immediately activates the audible and visual alarms in the relevant underground area and sends a manual verification instruction to the ground monitoring center. Simultaneously, the system calculates the current equalization pressure ventilation parameters and displays suggested values ​​for adjusting the ventilation window opening on the control screen to suppress accelerated oxidation in the goaf. This implementation process demonstrates that, through conflict processing and fusion of multi-source information, this invention can accurately assess the fire situation when a single sensor shows a sudden change in data, without ignoring its warning function or triggering the highest-level false alarm due to overreaction to a single indicator.

Claims

1. A coal mine goaf fire situation early warning method based on multi-source data fusion, characterized in that, The method comprises the following steps: obtaining multi-source environmental parameters of a monitoring section of a goaf by using a multi-source sensing module (101), wherein the multi-source environmental parameters include oxygen concentration, carbon monoxide concentration, acetylene concentration, methane concentration and environmental temperature; transmitting the data collected by the multi-source sensing module (101) to a fusion computing module (103) by using a data transmission module (102); converting the multi-source environmental parameters into basic probability assignment functions for different fire states by using the fusion computing module (103) according to a preset parameter interval and fire state mapping relationship, and constructing multiple independent evidence vectors; performing sequential fusion on the multiple independent evidence vectors by using the fusion computing module (103), calculating a conflict coefficient between the evidences in the fusion process, adopting a direct fusion, modified fusion or blocking verification strategy according to the size of the conflict coefficient, and obtaining a comprehensive confidence result representing a current fire situation; determining a fire development stage of the goaf based on the maximum confidence criterion by using a situation early warning module (104), and executing corresponding graded early warning measures according to the determination result.

2. The coal mine goaf fire situation early warning method based on multi-source data fusion according to claim 1, characterized in that, The multi-source environmental parameters of the monitoring section of the goaf are specifically obtained by using a sensor array arranged in an oxidation zone and a return air corner of the goaf. The sensor array includes an electrochemical oxygen concentration sensor, an electrochemical carbon monoxide concentration sensor, a gas chromatograph or laser spectroscopy acetylene concentration sensor, an infrared absorption or catalytic combustion methane concentration sensor, and a thermocouple or fiber optic grating temperature sensor. After the multi-source environmental parameters of the monitoring section of the goaf are obtained, the collected multi-source environmental parameters are further subjected to denoising and normalization processing, a monitoring vector containing oxygen concentration, carbon monoxide concentration, acetylene concentration, methane concentration and environmental temperature values is constructed, and each component in the monitoring vector corresponds to the real-time value of the above multi-source environmental parameters.

3. The coal mine goaf fire situation early warning method based on multi-source data fusion according to claim 1, characterized in that, The fire state mapping relationship includes a predefined identification framework of fire situation, and the identification framework includes four mutually exclusive state elements of normal state, budding stage, development stage and open fire stage.

4. The coal mine goaf fire situation early warning method based on multi-source data fusion according to claim 1, characterized in that, The conversion of the multi-source environmental parameters into basic probability assignment functions for different fire states specifically includes:

5. The coal mine goaf fire situation early warning method based on multi-source data fusion according to claim 4, characterized in that, presetting a membership degree model of each parameter in different value intervals for each fire state; for each environmental parameter input, the environmental parameter is converted into a basic probability assignment function defined on the identification framework by consulting the membership degree model. The construction of multiple independent evidence vectors specifically includes: converting the monitoring values of each type of environmental parameter into basic probability assignment functions defined on the identification framework through the preset membership degree model, generating multiple independent evidence vectors, and the multiple independent evidence vectors are five groups of evidence vectors corresponding to oxygen concentration, carbon monoxide concentration, acetylene concentration, methane concentration and environmental temperature.

6. The coal mine goaf fire situation early warning method based on multi-source data fusion according to claim 4, characterized in that, The process of performing sequential fusion on the multiple independent evidence vectors specifically includes:

7. The coal mine goaf fire situation early warning method based on multi-source data fusion according to claim 6, characterized in that, initializing the fusion sequence, and taking the first group of evidence vectors as the initial cumulative fusion result; ​ Iteratively introducing a subsequent evidence vector, at each time of introduction, calculating a conflict coefficient between the current cumulative fusion result and the newly introduced evidence vector; According to the size of the conflict coefficient, a corresponding fusion strategy is determined and executed.

8. The coal mine goaf fire situation early warning method based on multi-source data fusion according to claim 7, characterized in that, The process of determining and executing the corresponding fusion strategy specifically includes: When the conflict coefficient is less than or equal to a first preset threshold, it is determined to be a low conflict state, and a direct fusion is performed using the Dempster combination rule; When the conflict coefficient is greater than the first preset threshold and less than or equal to a second preset threshold, it is determined to be a moderate conflict state, and fusion is performed after the weighted average evidence is modified; When the conflict coefficient is greater than the second preset threshold, it is determined to be a high conflict state, and a blocking verification mechanism is triggered to temporarily exclude the current newly introduced evidence vector and maintain the fusion result at the last time. 9.The coal mine goaf fire situation early warning method based on multi-source data fusion according to claim 1, characterized in that, The process of determining based on the maximum confidence criterion specifically includes: In the comprehensive confidence result, a target state with the maximum basic probability assignment value is found; The target state is subjected to an effectiveness check, and when the effectiveness check passes, the target state is determined as the current fire development stage.

10. The coal mine goaf fire situation early warning method based on multi-source data fusion according to claim 9, characterized in that, The effectiveness check of the target state includes the following conditions: The basic probability assignment value of the target state is greater than the sum of the basic probability assignment value of the second largest probability state and a preset distinction threshold, and the uncertainty probability value in the comprehensive confidence result is less than a preset uncertainty tolerance threshold.

11. The coal mine goaf fire situation early warning method based on multi-source data fusion according to claim 1, characterized in that, The process of executing the corresponding hierarchical early warning measures includes: When it is determined to be a normal state, the system maintains a normal monitoring mode; When it is determined to be a budding stage, the system triggers a blue early warning response, increases the sensor sampling frequency and pushes attention information to the on-duty personnel; When it is determined to be a development stage, the system triggers a yellow early warning response, starts the sound and light alarm and generates an artificial patrol instruction; When it is determined to be an open fire stage, the system triggers a red early warning response, performs a linkage control of cutting off the power supply of non-essential safety type equipment, starting the regional fire extinguishing device and issuing an emergency evacuation instruction to personnel.

12. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that, The processor executes the computer program to realize the steps of the method of any one of claims 1-11.

13. A computer readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to realize the steps of the method of any one of claims 1-11.

Images