Mine safety production data fusion analysis method based on multi-source perception

Abstract

The invention relates to the technical field of mine safety production analysis, and discloses a mine safety production data fusion analysis method based on multi-source perception, which comprises the following steps: step 1, carrying out real-time data acquisition on physical parameters related to safety production in a mine environment through various types of sensors arranged in the mine environment, the collected data are stored as independent data streams with timestamps, and the data of different sensors have time inconsistency according to the difference set by the sampling frequency and the clock; and step 2, carrying out time delay compensation processing on the multi-source data with the timestamps acquired in the step 1. An optimization model is established in combination with a variational method to perform time delay compensation and time synchronization processing on sensor data, the problem of time inconsistency caused by different sampling frequencies and clocks of sensors is eliminated, fusion processing of multi-source sensor data on a unified time reference is realized, and an accurate data basis with consistent time sequence is obtained.

Description

Technical Field

[0001] The present invention relates to the technical field of mine safety production analysis, and in particular to a mine safety production data fusion analysis method based on multi-source perception. Background Art

[0002] With the expansion of mining scale and the advancement of deep resource development, the risks and challenges faced by mine safety production are increasing. In order to ensure the safety and efficiency of the mine operating environment, modern mine management systems generally rely on multi-source sensing technology, which monitors the parameters of the mine environment in real time through a distributed sensor network, and combines data fusion and analysis technology to achieve safety risk assessment and emergency decision support. However, due to the particularity of the mine environment, the complexity of multi-source sensors, and the technical bottleneck of data processing, there are many defects in the existing mine monitoring and analysis technology, which urgently need to be optimized. Among them, the defects are as follows:

[0003] The differences in sensor sampling frequencies and clock asynchrony in mining environments result in inconsistent timestamps in data from multi-source sensors. Time inconsistency makes it impossible to align data during fusion processing, which in turn affects the accuracy and reliability of data analysis.

[0004] Affected by the complex environment of the mine, the measurement data of each sensor usually contains different degrees of noise, and the noise characteristics vary depending on the performance of the sensor hardware. Traditional data fusion methods cannot effectively evaluate and suppress the impact of noise on the fusion results, especially when there is high noise interference in the sensor, which can easily lead to deviation or even misjudgment of the fusion data.

[0005] In the sensor network that has been running in mines for a long time, some sensors may lose or experience abnormalities in collected data due to equipment failure, signal interference, and power shortage, which affects the integrity of the data and interferes with subsequent analysis and safety decision-making processes.

[0006] The sources and distribution of various sensor data are heterogeneous, and the data distribution of different sensors is different. The existing methods cannot effectively optimize the sensor distribution differences during the fusion process, which affects the consistency and accuracy of the fusion results.

[0007] The existing mine monitoring system has a lag in the safety status assessment and early warning information generated after data processing, and is unable to respond to potential risks in a timely manner.

[0008] Therefore, those skilled in the art provide a mine safety production data fusion analysis method based on multi-source perception to solve the above-mentioned problems. Summary of the invention

[0009] In view of the deficiencies in the prior art, the present invention provides a mine safety production data fusion analysis method based on multi-source perception to solve the problems raised in the above background technology.

[0010] To achieve the above objectives, the present invention is implemented through the following technical solutions: a mine safety production data fusion analysis method based on multi-source perception, comprising:

[0011] Step 1: Real-time data collection of physical parameters related to safe production in the mining environment is performed by setting up various types of sensors in the mining environment. The collected data is stored as an independent data stream with a timestamp. The data of different sensors will have time inconsistency due to differences in sampling frequency and clock settings;

[0012] Step 2: Perform time delay compensation on the multi-source data with time stamps collected in step 1. By correcting the time tags of different sensor data, the time inconsistency problem caused by different sampling frequencies and clocks is eliminated. In the process, the time axis of each sensor data is aligned;

[0013] Step 3: Perform noise processing on the time-synchronized data in step 2. By evaluating the environmental noise characteristics in the data of each sensor, the data of different sensors are classified and weighted, and the data weight is adjusted according to the noise level of each sensor. In the process, the multi-source data with corrected time is optimized to generate preliminary fusion data with high signal-to-noise ratio.

[0014] Step 4: For the preliminary fusion data processed with noise in step 3, interpolation and completion processing is performed using the physical model in the mining environment in response to missing and abnormal sensor data, laying a data foundation for subsequent fusion analysis;

[0015] Step 5: Based on the complete and optimized multi-source data obtained in step 4, combined with the decision-making model of mine safety production, data fusion analysis is performed to support safety production decision-making. The data after fusion analysis is used to generate mine safety status assessment results and form early warning information in combination with the risk assessment model.

[0016] Preferably, in step 2, the delay compensation process is implemented by establishing an optimization model, and the objective function of the model is:

[0017]

[0018] Among them, J[τ i ] represents the delay compensation optimization target for sensor i, x i (t+τ i ) is the data of sensor i after delay compensation, t is time, τ i is the delay compensation parameter of sensor i, represents the target fusion data at time t, and T is the sampling time range.

[0019] Preferably, in step 2, during the delay compensation process, a Kalman filter method is used to optimize the synchronization data, and the state prediction formula of the Kalman filter is:

[0020]

[0021] in, represents the predicted state at time t, A(t) is the state transfer matrix, represents the estimated state at time t-1;

[0022] The measurement update formula is:

[0023]

[0024] in, represents the final estimated state at time y, represents the predicted state at time t, z i (t) represents the actual observation data at time t, H i represents the observation model of sensor i, is the mapping value of the predicted state in the observation space, K i (t) represents the degree of correction of the estimated state by sensor i at time t, which is defined as:

[0025]

[0026] Among them, K i (t) represents the degree of correction of the estimated state by sensor i at time t, P(t|t-1) represents the distribution of prediction error at time t, is the observation matrix H i The transposed matrix, R i (t) is the observation noise covariance of sensor i,

[0027] H i P(t|t-1) is the distribution of prediction error in the observation space,

[0028] Represents the weight of the observed data on the correction of the predicted state.

[0029] Preferably, in step 3, noise processing is implemented by weighted averaging method, and the preliminary data after fusion is:

[0030]

[0031] in, represents the final estimated value of the fusion of multi-sensor data at time t, n represents the number of sensors, is the noise variance of sensor i, x i (t) is the data of sensor i at time t.

[0032] Preferably, in step 3, the noise processing utilizes the information geometry optimization method to optimize the fusion strategy by minimizing the relative entropy divergence between sensor data. The relative entropy divergence is defined as:

[0033]

[0034] Among them, D KL (P(x i )||P(x j )) is the relative entropy divergence, which is used to measure the probability distribution P(x i ) and P(x j ),

[0035] P(x i ) is the probability distribution of sensor i, P(x j ) is the probability distribution of sensor j, log is the natural logarithm, and x represents the independent variable of integration.

[0036] Preferably, in step 4, when the physical model is used to supplement the missing sensor data, the gas diffusion model, heat conduction model and aerodynamic model in the mining environment are combined, wherein the interpolation formula of the gas diffusion model is:

[0037]

[0038] Among them, C(t) represents the gas concentration, Q is the gas release rate, D is the diffusion coefficient, t is the time, r is the diffusion distance, exp is the exponential function, and π is a mathematical constant.

[0039] Preferably, in step 4, the interpolation is optimized by a variational method, and the variational objective function is:

[0040]

[0041] Among them, J[x i (t)] is the objective function, x i (t) is the actual data of sensor i at time t, f(x j (t)) is the interpolation function based on the data of sensor i, T is the time range, and t is the time variable.

[0042] Preferably, in step 5, the decision model is based on a Markov decision process and is solved by a dynamic programming method, and the state value function is defined as:

[0043]

[0044] Where s is the current state, a is the action, R(s, a) is the reward value of state s and action a, γ is the discount factor, P(s′|s, a) is the state transition probability, V(s) is the value function of state s, s′ is the next state, and V(s′) is the value function of state s′.

[0045] Preferably, the result calculated by the state value function is used to generate early warning information in real time. If excessive gas concentration, abnormal equipment vibration and abnormal temperature are detected in the mining environment, an early warning signal is triggered and an emergency response mechanism is started.

[0046] Preferably, the data fused and analyzed in step 5 is used to generate a dynamic assessment report on mine safety production, which includes real-time environmental monitoring results, potential risk points, and emergency response recommendations for different risks.

[0047] The present invention provides a mine safety production data fusion analysis method based on multi-source perception. It has the following beneficial effects:

[0048] 1. The present invention establishes an optimization model by combining the variational method to perform delay compensation and time synchronization processing on sensor data, eliminates the time inconsistency problem caused by different sensor sampling frequencies and clocks, realizes the fusion processing of multi-source sensor data on a unified time base, and obtains an accurate and time-consistent data basis.

[0049] 2. The present invention evaluates the noise in the sensor data and adjusts the weights by combining the weighted average method with the information geometry optimization strategy, dynamically optimizes the fusion weights according to the noise variance of the sensor, effectively suppresses the influence of high-noise data sources, obtains preliminary fusion data with a high signal-to-noise ratio, and improves the accuracy and credibility of the data.

[0050] 3. The present invention interpolates and completes the sensor data through a physical model, and optimizes the interpolation function using the variational method, effectively completes the missing and abnormal data, achieves data continuity and integrity, obtains data input for subsequent analysis, and eliminates the impact of missing data on the fusion results. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] Figure 1 It is a flow chart of the present invention. DETAILED DESCRIPTION

[0052] In order to make the technical personnel in the technical field understand the scheme of the present invention, the technical scheme in the embodiment of the present invention will be clearly and completely described below in combination with the drawings in the embodiment of the present invention. Obviously, the described embodiment is a partial embodiment of the present invention, not a complete embodiment. Based on the embodiment of the present invention, other embodiments obtained by ordinary technicians in the field without creative work should fall within the scope of protection of the present invention.

[0053] The present invention is described in detail below in conjunction with the accompanying drawings:

[0054] Example:

[0055] Please see attached Figure 1 The embodiment of the present invention provides a method for fusion analysis of mine safety production data based on multi-source perception, including:

[0056] Step 1: Real-time data collection of physical parameters related to safe production in the mining environment is performed by setting up various types of sensors in the mining environment. The collected data is stored as an independent data stream with a timestamp. The data of different sensors will have time inconsistency due to differences in sampling frequency and clock settings;

[0057] Step 2: Perform time delay compensation on the multi-source data with time stamps collected in step 1. By correcting the time tags of different sensor data, the time inconsistency problem caused by different sampling frequencies and clocks is eliminated. In the process, the time axis of each sensor data is aligned;

[0058] Step 3: Perform noise processing on the time-synchronized data in step 2. By evaluating the environmental noise characteristics in the data of each sensor, the data of different sensors are classified and weighted, and the data weight is adjusted according to the noise level of each sensor. In the process, the multi-source data with corrected time is optimized to generate preliminary fusion data with high signal-to-noise ratio.

[0059] Step 4: For the preliminary fusion data processed with noise in step 3, interpolation and completion processing is performed using the physical model in the mining environment in response to missing and abnormal sensor data, laying a data foundation for subsequent fusion analysis;

[0060] Step 5: Based on the complete and optimized multi-source data obtained in step 4, combined with the decision-making model of mine safety production, data fusion analysis is performed to support safety production decision-making. The data after fusion analysis is used to generate mine safety status assessment results and form early warning information in combination with the risk assessment model.

[0061] Benefits of Step 1: Realize multi-dimensional real-time monitoring of the mining environment and provide a comprehensive and authentic source of original data for subsequent data analysis and processing.

[0062] Benefits of step 2: Eliminate time deviations between sensors, ensure synchronization and time consistency of multi-source data, and improve the accuracy of subsequent data processing.

[0063] Benefits of step 3: effectively suppress the impact of environmental noise on data quality, improve data accuracy and reliability, and provide input data for fusion analysis.

[0064] Benefits of step 4: Solve the problem of missing or abnormal data, improve data integrity, and ensure the continuity and reliability of subsequent fusion analysis.

[0065] Benefits of Step 5: By integrating analysis and decision-making models, the risks of the mining environment can be assessed and early warning information can be generated in a timely manner, significantly improving the real-time and effectiveness of mine safety management.

[0066] Summary: The mine safety production data fusion analysis method based on multi-source perception provided by the present invention solves the problems of time inconsistency, noise interference, data missing and distribution differences in mine monitoring through a complete process design from data collection to decision support.

[0067] Through time synchronization, noise processing and data completion of multi-source sensors, the quality and integrity of input data are ensured; combined with fusion analysis and dynamic decision support, the monitoring accuracy and risk assessment capabilities of the mining environment are significantly improved, providing scientific and efficient management support for mine safety production, and ultimately achieving comprehensive intelligent monitoring and early warning management.

[0068] In step 2, the delay compensation process is implemented by establishing an optimization model, and the objective function of the model is:

[0069]

[0070] Among them, J[τ i ] represents the delay compensation optimization target for sensor i, x i (t+τ i ) is the data of sensor i after delay compensation, t is time, τ i is the delay compensation parameter of sensor i, represents the target fusion data at time t, and T is the sampling time range.

[0071] By constructing the objective function and calculating the delay compensation parameter τ i Optimization and dynamic adjustment of data timestamps of different sensors can align the time axis of each data stream, eliminate the time inconsistency problem caused by different sensor sampling frequencies and clocks, and provide a time benchmark for subsequent data fusion processing.

[0072] After delay compensation, the data from each sensor is fused based on a unified time reference, avoiding fusion deviation caused by time inconsistency and significantly improving the accuracy and consistency of the fused data.

[0073] Delay compensation makes the time series of sensor data accurate, ensuring that when sudden security incidents occur, the system can monitor and handle risks in a timely manner based on the real time sequence, enhancing real-time performance and reliability.

[0074] Time alignment processing ensures the temporal consistency of sensor data, provides input data for subsequent noise processing, missing data completion, and decision analysis, and improves the efficiency and effectiveness of the data processing process.

[0075] Summary: By establishing an optimization model and dynamically adjusting the compensation parameter τ i , which solves the problem of inconsistent timestamps of multi-source data in mine monitoring, significantly improves the accuracy and real-time performance of data fusion, provides basic support for subsequent data analysis, and ultimately achieves efficient monitoring and intelligent management of mine safety production.

[0076] In step 2, during the delay compensation process, the Kalman filter method is used to optimize the synchronization data. The state prediction formula of the Kalman filter is:

[0077]

[0078] in, represents the predicted state at time t, A(t) is the state transfer matrix, represents the estimated state at time t-1;

[0079] The measurement update formula is:

[0080]

[0081] in, represents the final estimated state at time y, represents the predicted state at time t, z i (t) represents the actual observation data at time t, H i represents the observation model of sensor i, is the mapping value of the predicted state in the observation space, K i (t) represents the degree of correction of the estimated state by sensor i at time t, which is defined as:

[0082]

[0083] Among them, K i(t) represents the degree of correction of the estimated state by sensor i at time t, P(t|t-1) represents the distribution of prediction error at time t, is the observation matrix H i The transposed matrix, R i (t) is the observation noise covariance of sensor i,

[0084] H i P(t|t-1) is the distribution of prediction error in the observation space,

[0085] Represents the weight of the observed data on the correction of the predicted state.

[0086] Through the state prediction formula, the Kalman filter can dynamically predict and correct the delay-compensated data based on the sensor's historical state and current observation values, effectively eliminating the errors caused by time inconsistency and significantly improving the accuracy of delay compensation.

[0087] During the measurement update process, the Kalman filter adjusts the degree of correction of the observed data to the predicted state through the Kalman gain, thereby minimizing the uncertainty introduced by sensor noise and effectively improving the reliability and consistency of the data.

[0088] By introducing the observation noise covariance R i (t) is dynamically adjusted. Kalman filtering can dynamically weight the fusion data according to the noise characteristics of different sensors and optimize the synchronization quality of multi-source data.

[0089] Kalman filtering can adjust the state transfer matrix and observation noise model in real time according to the dynamic changes in the mining environment to ensure the applicability of delay compensation in complex environments.

[0090] The recursive algorithm of Kalman filtering can quickly process sensor data, ensure the efficiency and real-time delay compensation process, and provide rapid response capabilities for subsequent data processing and decision support.

[0091] Summary: The introduction of Kalman filtering method in time delay compensation solves the problem of inconsistent time axes of multi-source sensor data and significantly improves the accuracy and real-time performance of data fusion.

[0092] In step 3, noise processing is achieved by weighted averaging, and the preliminary data after fusion is:

[0093]

[0094] in, represents the final estimated value of the fusion of multi-sensor data at time t, n represents the number of sensors, is the noise variance of sensor i, x i(t) is the data of sensor i at time t.

[0095] In step 3, noise processing uses the information geometry optimization method to optimize the fusion strategy by minimizing the relative entropy divergence between sensor data. The relative entropy divergence is defined as:

[0096]

[0097] Among them, D KL (P(x i )||P(x j )) is the relative entropy divergence, which is used to measure the probability distribution P(x i ) and P(x j ),

[0098] P(x i ) is the probability distribution of sensor i, P(x j ) is the probability distribution of sensor j, log is the natural logarithm, and x represents the independent variable of integration.

[0099] Weighted average noise processing formula and benefits: through noise variance The inverse of is used as the weight to reduce the impact of noisy sensor data on the final fusion result.

[0100] The noise characteristics of different sensors change dynamically over time. The weighted average method can adaptively adjust the weights according to the current noise variance of the sensor to ensure the applicability of the fusion strategy in a dynamic environment.

[0101] Through the weighted average method, multi-sensor observation data are integrated to weaken the impact of high-noise data, and ultimately generate preliminary fusion data with a high signal-to-noise ratio to provide input for subsequent analysis.

[0102] The relative entropy divergence formula and benefits of information geometry optimization: The relative entropy divergence can quantify the difference in data distribution between sensors. By minimizing the relative entropy divergence, the data fusion strategy is optimized to make the distribution of the fused data consistent.

[0103] By optimizing the relative entropy divergence, the deviation caused by the differences in distribution characteristics of different sensors can be effectively eliminated, ensuring the rationality and consistency of the data fusion results.

[0104] The information geometry optimization method can adapt to the dynamic changes in the distribution of different sensor data in complex environments, ensuring that the fusion results have high robustness and reliability.

[0105] Summary: Noise processing is an important part of the data fusion analysis of the present invention. By combining the weighted average method with the information geometry optimization method, the problems of noise interference and distribution difference in sensor data can be effectively solved.

[0106] The weighted average method can dynamically adjust the weights of sensors, suppress the influence of high-noise sensors, and generate preliminary fusion data with a high signal-to-noise ratio; information geometry optimization improves the distribution consistency and reliability of multi-source data fusion by minimizing the relative entropy divergence.

[0107] In step 4, when the physical model is used to supplement the missing sensor data, the gas diffusion model, heat conduction model and aerodynamic model in the mining environment are combined. The interpolation formula of the gas diffusion model is:

[0108]

[0109] Among them, C(t) represents the gas concentration, Q is the gas release rate, D is the diffusion coefficient, t is the time, r is the diffusion distance, exp is the exponential function, and π is a mathematical constant.

[0110] In step 4, the interpolation is optimized by the variational method, and the variational objective function is:

[0111]

[0112] Among them, J[x i (t)] is the objective function, x i (t) is the actual data of sensor i at time t, f(x j (t)) is the interpolation function based on the data of sensor i, T is the time range, and t is the time variable.

[0113] Gas diffusion model interpolation formula and benefits: The gas diffusion model can simulate the dynamic distribution of gas in the mining environment by introducing the diffusion coefficient D, release rate Q and time variable t, providing a physical basis for the interpolation of missing data.

[0114] Compared with the traditional simple interpolation method, the gas diffusion model is based on the physical laws of mine gas diffusion and can generate reasonable and practical concentration data, which is especially applicable in gas leakage scenarios.

[0115] Variational interpolation formula and benefits: By constructing the optimization objective function through variational method, the interpolation function f(x j The parameters (t)) are used to minimize the error between the interpolated data and the actual data and improve the accuracy of the interpolation result.

[0116] Summary: In step 4, the gas diffusion model is used to interpolate the missing data, and the interpolation function is optimized by the variational method, which significantly improves the accuracy and adaptability of data completion.

[0117] The gas diffusion model describes the dynamic changes of mine gas distribution by introducing physical quantities, providing a scientific basis for interpolation; the variational method minimizes the interpolation error, realizes dynamic optimization of the interpolation function, and improves the reliability of the interpolation results.

[0118] In step 5, the decision model is based on the Markov decision process and is solved by the dynamic programming method. The state value function is defined as:

[0119]

[0120] Where s is the current state, a is the action, R(s, a) is the reward value of state s and action a, γ is the discount factor, P(s′|s, a) is the state transition probability, V(s) is the value function of state s, s′ is the next state, and V(s′) is the value function of state s′.

[0121] The results calculated by the state value function are used to generate early warning information in real time. If the gas concentration in the mining environment exceeds the standard, the equipment vibrates abnormally, and the temperature is abnormal, the early warning signal is triggered and the emergency response mechanism is activated.

[0122] The data fused and analyzed in step 5 is used to generate a dynamic assessment report on mine safety production, which includes real-time environmental monitoring results, potential risk points, and emergency response recommendations for different risks.

[0123] Benefits of state value function and its application: Through the Markov decision process, the optimal decision-making strategy can be dynamically calculated based on real-time monitoring data, weighing current actions and future potential risks to ensure the rationality and scientificity of each decision.

[0124] Based on the optimal decision-making results calculated by the value function, the system can evaluate the risk level in real time, generate detailed early warning information, and realize real-time monitoring and efficient management of the mining operating environment.

[0125] After the early warning is triggered, the system quickly initiates the emergency response mechanism based on the guidance information provided by the decision-making model, and provides specific response measures for different types of risks, significantly improving the response efficiency of mine safety management.

[0126] Benefits of generating and applying dynamic assessment reports: Dynamic assessment reports can comprehensively present the real-time safety status of mines by integrating multi-source sensor data and the results of decision-making models, providing managers with global safety situation awareness.

[0127] The report contains detailed markings of potential risk points and their risk levels, which can help managers quickly locate problem areas and reduce risk investigation time.

[0128] For different types of risks, the report provides emergency response recommendations based on decision-making models, providing reliable support for managers to formulate scientific emergency plans.

[0129] In summary, in step 5, the generation of the decision model based on the Markov decision process and the dynamic evaluation report is the key link in realizing intelligent mine safety management in the present invention.

[0130] Through the dynamic programming method of the state value function, the system can optimize the decision-making strategy in real time, quickly identify potential risks and trigger the early warning mechanism, effectively improving the real-time and scientific nature of mine safety management; combined with the dynamic assessment report generated by fusion analysis, the system provides managers with comprehensive environmental monitoring information, risk assessment results and emergency response recommendations, providing decision-making support for mine safety production.

[0131] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A mine safety production data fusion analysis method based on multi-source perception, characterized in that: include: Step 1: Real-time data collection of physical parameters related to safe production in the mining environment is performed by setting up various types of sensors in the mining environment. The collected data is stored as an independent data stream with a timestamp. The data of different sensors will have time inconsistency due to differences in sampling frequency and clock settings; Step 2: Perform time delay compensation on the multi-source data with time stamps collected in step 1. By correcting the time tags of different sensor data, the time inconsistency problem caused by different sampling frequencies and clocks is eliminated. In the process, the time axis of each sensor data is aligned; Step 3: Perform noise processing on the time-synchronized data in step 2. By evaluating the environmental noise characteristics in the data of each sensor, the data of different sensors are classified and weighted, and the data weight is adjusted according to the noise level of each sensor. In the process, the multi-source data with corrected time is optimized to generate preliminary fusion data with high signal-to-noise ratio. Step 4: For the preliminary fusion data processed with noise in step 3, interpolation and completion processing is performed using the physical model in the mining environment in response to missing and abnormal sensor data, laying a data foundation for subsequent fusion analysis; Step 5: Based on the complete and optimized multi-source data obtained in step 4, combined with the decision-making model of mine safety production, data fusion analysis is performed to support safety production decision-making. The data after fusion analysis is used to generate mine safety status assessment results and form early warning information in combination with the risk assessment model.

2. The method for fusion analysis of mine safety production data based on multi-source perception according to claim 1 is characterized in that: In step 2, the delay compensation process is implemented by establishing an optimization model, and the objective function of the model is: Among them, J[τ i ] represents the delay compensation optimization target for sensor i, x i (t+τ i ) is the data of sensor i after delay compensation, t is time, τ i is the delay compensation parameter of sensor i, represents the target fusion data at time t, and T is the sampling time range.

3. The method for fusion analysis of mine safety production data based on multi-source perception according to claim 2 is characterized in that: In step 2, during the delay compensation process, the Kalman filter method is used to optimize the synchronization data, and the state prediction formula of the Kalman filter is: in, represents the predicted state at time t, A(t) is the state transfer matrix, represents the estimated state at time t-1; The measurement update formula is: in, represents the final estimated state at time y, represents the predicted state at time t, z i (t) represents the actual observation data at time t, H i represents the observation model of sensor i, is the mapping value of the predicted state in the observation space, K i (t) represents the degree of correction of the estimated state by sensor i at time t, which is defined as: Among them, K i (t) represents the degree of correction of the estimated state by sensor i at time t, P(t|t-1) represents the distribution of prediction error at time t, is the observation matrix H i The transposed matrix, R i (t) is the observation noise covariance of sensor i, H i P(t|t-1) is the distribution of prediction error in the observation space, Represents the weight of the observed data on the correction of the predicted state.

4. The method for fusion analysis of mine safety production data based on multi-source perception according to claim 1 is characterized in that: In step 3, noise processing is achieved by weighted averaging, and the preliminary data after fusion is: in, represents the final estimated value of the fusion of multi-sensor data at time t, n represents the number of sensors, is the noise variance of sensor i, x i (t) is the data of sensor i at time t.

5. The method for fusion analysis of mine safety production data based on multi-source perception according to claim 4 is characterized in that: In step 3, noise processing uses the information geometry optimization method to optimize the fusion strategy by minimizing the relative entropy divergence between sensor data. The relative entropy divergence is defined as: Among them, D KL (P(x i )||P(x j )) is the relative entropy divergence, which is used to measure the probability distribution P(x i ) and P(x j ), P(x i ) is the probability distribution of sensor i, P(x j ) is the probability distribution of sensor j, log is the natural logarithm, and x represents the independent variable of integration.

6. The method for fusion analysis of mine safety production data based on multi-source perception according to claim 1 is characterized in that: In step 4, when the physical model is used to supplement the missing sensor data, the gas diffusion model, heat conduction model and aerodynamic model in the mining environment are combined, wherein the interpolation formula of the gas diffusion model is: Among them, C(t) represents the gas concentration, Q is the gas release rate, D is the diffusion coefficient, t is the time, r is the diffusion distance, exp is the exponential function, and π is a mathematical constant.

7. The method for fusion analysis of mine safety production data based on multi-source perception according to claim 6 is characterized in that: In step 4, interpolation is optimized by variational method, and the variational objective function is: Among them, J[x i (t)] is the objective function, x i (t) is the actual data of sensor i at time t, f(x j (t)) is the interpolation function based on the data of sensor i, T is the time range, and t is the time variable.

8. The method for fusion analysis of mine safety production data based on multi-source perception according to claim 1 is characterized in that: In step 5, the decision model is based on the Markov decision process and is solved by the dynamic programming method. The state value function is defined as: Where s is the current state, a is the action, R(s, a) is the reward value of state s and action a, γ is the discount factor, P(s′|s, a) is the state transition probability, V(s) is the value function of state s, s′ is the next state, and V(s′) is the value function of state s′.

9. The method for fusion analysis of mine safety production data based on multi-source perception according to claim 8 is characterized in that: The result calculated by the state value function is used to generate early warning information in real time. If excessive gas concentration, abnormal equipment vibration and abnormal temperature are detected in the mining environment, an early warning signal is triggered and an emergency response mechanism is started.

10. The method for fusion analysis of mine safety production data based on multi-source perception according to claim 9, characterized in that: The data fused and analyzed in step 5 is used to generate a dynamic assessment report on mine safety production, which includes real-time environmental monitoring results, potential risk points, and emergency response recommendations for different risks.

Images