Deep mine disaster big data analysis and prevention and control decision system

Abstract

The invention relates to the technical field of mine disaster monitoring and prevention and control, in particular to a deep mine disaster big data analysis and prevention and control decision-making system. Comprising a data pollution quantification unit used for collecting a multi-source heterogeneous data stream and reference prediction data and carrying out noise pollution quantitative analysis to obtain a data pollution index; the toughness entropy evaluation unit is used for collecting physical risk factors, resolving the basic safety entropy, correcting the basic safety entropy according to the data pollution index to obtain the toughness entropy of the mine system, and judging the toughness entropy of the mine system to obtain a risk level; the anti-fragility decision-making unit is used for dynamically adjusting a risk aversion coefficient according to the risk level, and performing utility function solution on a preset decision to be selected to obtain an optimal decision; and the closed-loop scheduling control unit is used for matching a preset scheduling instruction set according to the optimal decision and the risk level, and generating and executing a hierarchical control instruction. The problem that a traditional risk assessment model fails due to serious pollution of deep mine monitoring data is solved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of mine disaster monitoring and prevention, in particular to a deep mine disaster big data analysis and prevention and control decision system. BACKGROUND

[0002] The existing data processing scheme generally adopts a batch processing architecture to realize periodic scanning, decompression and warehousing of data through task scheduling, and completes index calculation in the database; this architecture is highly dependent on disk I / O and contains multiple intermediate data storage, with a long process; at the same time, manual adaptation processing is required for heterogeneous file formats, resulting in high response delay and large repeated calculation of the system; This mode can still be applied in offline statistical scenarios, but its inherent defects make it difficult to support high-frequency index update requirements and rapid fault positioning; therefore, improving the timeliness and accuracy of real-time data analysis and output is a key technical problem to be solved at present.

[0003] The above information disclosed in the above background section is only used to strengthen the understanding of the background of the present disclosure, and therefore it can include information that does not constitute prior art known to those of ordinary skill in the art. SUMMARY

[0004] To solve the above technical problems, the present application discloses a deep mine disaster big data analysis and prevention and control decision system, in particular, the technical scheme of the present application comprises: The data pollution quantization unit is used to collect multi-source heterogeneous data streams and baseline prediction data, perform noise pollution quantization analysis, and obtain a data pollution index; The resilience entropy evaluation unit is used to collect physical risk factors, solve the basic safety entropy, and modify the basic safety entropy according to the data pollution index to obtain the mine system resilience entropy, and perform discrimination processing on the mine system resilience entropy to obtain the risk level; The anti-fragile decision unit is used to dynamically adjust the risk aversion coefficient according to the risk level, and perform utility function calculation on the preset selected decision to obtain the optimal decision; The closed-loop scheduling control unit is used to match the preset scheduling instruction set according to the optimal decision and the risk level, generate and execute the hierarchical control instruction.

[0005] Preferably, the noise pollution quantization analysis of the data pollution quantization unit comprises the following steps: Obtain the real-time observation value of each sensor; Obtain the theoretical reading output by the baseline prediction model trained based on historical data; Calculate the relative deviation of the real-time observation value and the theoretical reading to obtain the relative deviation of each type of data; The data pollution index is obtained by weighting and summing the relative deviations of various types of data according to preset normalization weights.

[0006] Preferably, the resiliency entropy evaluation unit includes the following steps for calculating the basic safety entropy: Obtaining factor observation values of key physical risk factors; Obtaining preset factor safety benchmark values corresponding to the factor observation values; Obtaining preset factor critical danger values corresponding to the factor observation values; Combining the factor observation values and the factor safety benchmark values, and combining the factor critical danger values, to perform normalization processing and calculate the risk degree of each factor; The basic safety entropy is obtained by weighting and summing the risk degrees of various factors according to preset factor weights.

[0007] Preferably, the resiliency entropy evaluation unit includes the following steps for modifying the basic safety entropy: According to the preset pollution sensitivity coefficient and the preset pollution influence index, and combining the data pollution index, a nonlinear correction factor is constructed; The basic safety entropy is multiplied by the nonlinear correction factor to obtain the mine system resiliency entropy.

[0008] Preferably, the resiliency entropy evaluation unit includes the following steps for discriminating the mine system resiliency entropy: The mine system resiliency entropy is compared with a preset first warning threshold; The mine system resiliency entropy is compared with a preset second danger threshold; When the mine system resiliency entropy is less than the first warning threshold, the risk level is determined as a normal state; When the mine system resiliency entropy is greater than or equal to the first warning threshold and less than the second danger threshold, the risk level is determined as a first warning; When the mine system resiliency entropy is greater than or equal to the second danger threshold, the risk level is determined as a second warning.

[0009] Preferably, the anti-fragile decision unit includes the following steps for calculating the utility function: Obtaining expected economic benefits of a preset candidate decision; Calculating a system collapse probability based on the mine system resiliency entropy; Subtracting the product of the risk aversion coefficient and the system collapse probability from the expected economic benefits to obtain a decision utility of the candidate decision; Determining the candidate decision corresponding to the maximum decision utility as the optimal decision.

[0010] Preferably, the anti-fragile decision unit includes the following steps for dynamically adjusting the risk aversion coefficient: When the risk level is a normal state, the risk aversion coefficient is set as a preset low coefficient value; When the risk level is a first-level early warning, the risk aversion coefficient is switched to a preset high coefficient value; When the risk level is a second-level early warning, the risk aversion coefficient is switched to a preset critical coefficient value; wherein the critical coefficient value is greater than the high coefficient value. Wherein, the high coefficient value ensures that the product of the risk aversion coefficient and the system collapse probability is greater than the expected economic benefit.

[0011] Preferably, the closed-loop scheduling control unit matches the preset scheduling instruction set, including the following steps: The preset scheduling instruction set includes production optimal mode instructions, first-level safety mode instructions, and second-level safety mode instructions. When the optimal decision is a production optimal solution and the risk level is a normal state, the hierarchical control instruction is determined as the production optimal mode instruction. When the optimal decision is a safety solution and the risk level is a first-level early warning, the hierarchical control instruction is switched to the first-level safety mode instruction. When the optimal decision is a safety solution and the risk level is a second-level early warning, the hierarchical control instruction is switched to the second-level safety mode instruction.

[0012] Preferably, the closed-loop scheduling control unit is further used for: After executing the first-level safety mode instruction or the second-level safety mode instruction, the mine system resilience entropy is continuously monitored. When the mine system resilience entropy falls below the first-level early warning threshold, the anti-fragile decision unit restores the risk aversion coefficient to the low coefficient value. The hierarchical control instruction is switched back to the production optimal mode instruction.

[0013] Compared with the prior art, the present application has the following beneficial effects: 1. The present application solves the problem of traditional risk assessment model failure caused by serious pollution of deep mine monitoring data. The system calculates the relative deviation between the observed value and the theoretical reading in real time through the unique data pollution quantization unit, and obtains an accurate data pollution index by weighted summation. This overcomes the defect of blindly trusting data in the prior art, realizes real-time quantitative perception of the overall pollution degree of data flow, and provides a solid and quantifiable data quality foundation for subsequent reliable risk assessment and decision-making.

[0014] 2、The present application penetrates the dirty data fog, and realizes reliable evaluation of mine risk. The resilience entropy evaluation unit in the system not only calculates the basic safety entropy based on physical risk factors, but also innovatively constructs a nonlinear correction factor according to the data pollution index to correct the basic safety entropy. The corrected resilience entropy couples physical risk and data uncertainty risk, so that even when physical indicators are distorted due to data pollution, it can be significantly amplified, thereby making a conservative and reliable estimate of the true risk.

[0015] 3、The present application realizes anti-fragile decision-making in an uncertain environment. Based on the evaluated risk level, the system dynamically adjusts the risk aversion coefficient in the anti-fragile decision-making unit. When the risk increases, the system automatically switches to a high coefficient value, ensuring that the risk penalty term dominates in the utility function calculation, so that the optimal decision switches from the production optimal solution to the safety solution. This design prioritizes avoiding the occurrence of worst-case scenarios such as system collapse, achieving the anti-fragile goal at the decision-making level.

[0016] 4、The present application constructs a complete dynamic resilience closed loop from risk perception to field control and then to state recovery. The closed-loop scheduling control unit matches and executes hierarchical control instructions based on the optimal decision and risk level. After executing the safety instructions, the system will continuously monitor the resilience entropy, and when it falls below the safety threshold, it can automatically switch the decision mode and control instructions back to the production optimal state, achieving dynamic balance between safety and efficiency in the time scale, and ensuring that the system can recover to high-efficiency production after the risk is removed. BRIEF DESCRIPTION OF DRAWINGS

[0017] The present application will be further explained in conjunction with the accompanying drawings and examples: Figure 1 is a system structure diagram of the present application. DETAILED DESCRIPTION

[0018] To make the purpose, technical scheme and advantages of the present application clearer, the present application will be further explained in conjunction with specific examples.

[0019] Example 1: Please refer to Figure 1 Deep mine disaster big data analysis and prevention and control decision-making system, including: The data pollution quantification unit is used to collect multi-source heterogeneous data streams and benchmark prediction data, perform noise pollution quantification analysis, and obtain the data pollution index; The resilience entropy evaluation unit is used to collect physical risk factors, calculate the basic safety entropy, and correct the basic safety entropy according to the data pollution index, to obtain the mine system resilience entropy, and to process the mine system resilience entropy to obtain the risk level; The anti-fragile decision unit is used for dynamically adjusting a risk-averse coefficient according to the risk level, and performing utility function calculation on the preset to-be-selected decision to obtain an optimal decision. The closed-loop scheduling control unit is used for matching a preset scheduling instruction set to generate and execute a hierarchical control instruction according to the optimal decision and the risk level. The embodiment provides a deep mine disaster big data analysis and prevention and control decision system, which comprises a data pollution quantization unit, a resilience entropy evaluation unit, an anti-fragile decision unit and a closed-loop scheduling control unit. The data pollution quantization unit comprises: The purpose of the unit is to quantitatively monitor the overall reliability of the data stream collected by the monitoring system in real time, that is, the pollution degree. Under high noise, such as microseismic noise interference, the sensor data may be distorted. The unit compares the deviation of the real-time observation value and the benchmark prediction value to construct a dimensionless index. In the embodiment, the data pollution quantization unit is used for collecting multi-source heterogeneous data streams, such as real-time readings of microseismic, ground stress, gas and other sensors and benchmark prediction data, such as theoretical readings output by a time series model trained based on historical data. Through weighted relative deviation calculation of the two, noise pollution quantization analysis is performed, and finally a comprehensive data pollution index is obtained. ; The higher, the more unreliable the current data stream is. The resilience entropy evaluation unit comprises: The purpose of the unit is to evaluate the real safety distance of the mine system. The core innovation lies in that it not only evaluates the known physical risk, but also evaluates the uncertainty risk caused by data pollution. In the embodiment, the resilience entropy evaluation unit is used for collecting physical risk factors, such as observation values of ground stress and microseismic energy. The basic safety entropy reflecting the danger degree of the physical index is calculated through normalization and weighted summation . The unit performs nonlinear correction on the basic safety entropy based on the data pollution index input by the data pollution quantization unit to obtain the final mine system resilience entropy . The physical risk and the data pollution risk are comprehensively considered. Finally, the unit discriminates the mine system resilience entropy and the preset two-level threshold value , and outputs discrete risk levels, such as a normal state, a first-level warning and a second-level warning. The anti-fragile decision unit comprises: The purpose of the unit is to solve the conflict between mine safety and benefit, and realize anti-fragile decision, that is, when the risk increases, the decision system can actively deviate from the worst case rather than pursuing the maximum benefit. In this embodiment, the antifragile decision-making unit is used to dynamically adjust the risk aversion coefficient in the decision-making model based on the risk level output by the resilience entropy assessment unit. When the risk level is normal, The risk level is relatively low, and decision-making is biased towards efficiency; when the risk level is at the warning level, Switch to the maximum value; simultaneously, the unit makes preset decisions on candidate solutions, such as producing the optimal solution. and security solution Perform utility function The solution shows that this function incorporates expected economic benefits. Japanese Classics probability of collapse after punishment Ultimately, the unit output makes Maximizing the optimal decision ; Closed-loop scheduling control unit: The purpose of this unit is to transform the optimal decision output by the upstream decision-making unit into specific operational instructions at the mine site and to achieve closed-loop control. In this embodiment, the closed-loop scheduling control unit is used to make the optimal decision based on the output of the antifragile decision unit. The risk level output by the resilience entropy assessment unit, within a preset set of scheduling instructions, such as... The corresponding intelligent tunneling mode, The corresponding speed reduction mode, Match the corresponding shutdown mode to generate and execute specific hierarchical control instructions; This embodiment, through the collaborative work of the four units mentioned above, constructs a complete technical closed loop, from data quality perception to comprehensive risk assessment, and then to antifragile decision-making and closed-loop control execution. It solves the significant hidden danger of decision-making errors in deep mines due to blindly trusting dirty data under high noise interference. This system can penetrate the data fog and quantify the real system risks brought about by data uncertainty, namely resilience entropy. Based on this risk, it automatically switches its decision-making logic between benefits and safety, i.e., it dynamically adjusts... This ensures that when facing conflict, the mining system always prioritizes maintaining long-term survival resilience.

[0020] Example 2: The data pollution quantification unit performs noise pollution quantification analysis, which includes the following steps: Acquire real-time observations from each sensor; Obtain the theoretical readings output by the benchmark prediction model trained on historical data; The relative deviation between real-time observations and theoretical readings is calculated to obtain the relative deviation of various types of data. The data pollution index is obtained by weighting and summing the relative deviations of various types of data according to the preset normalization weight; The embodiment refines the specific steps of noise pollution quantization analysis of the data pollution quantization unit based on embodiment 1; the motivation of the analysis is to quantify the overall distortion degree of multi-source heterogeneous data with different physical dimensions under noise interference, and a weighted relative deviation model is defined ; The specific steps include: Obtaining real-time observation values of various sensors : Real-time observation value, the dimension of which depends on the first type of sensor, such as may be a ground stress value with a dimension of MPa; the value is collected in real time by a deployed distributed sensor network; Obtaining theoretical readings output by a benchmark prediction model trained based on historical data : Benchmark prediction value, i.e. theoretical reading, with the same dimension as ; the value is output by a pre-trained benchmark prediction model, for example, a time series model trained based on historical data or a mechanism model based on geomechanical parameters, for predicting the theoretical reading of the sensor at the next moment; Calculating the relative deviation of the real-time observation value and the theoretical reading to obtain the relative deviation of various types of data ; According to the preset normalization weight , the relative deviation of various types of data is weighted and summed to obtain the data pollution index ; The above steps 3 and 4 are mathematically combined into one formula, i.e. the calculation model of the data pollution index : Wherein, : data pollution index, dimensionless; the value is calculated by this unit and delivered to the resilience entropy evaluation unit as a key input for correcting the risk assessment result; The higher the value is, the more unreliable the data is; : total number of sensors, integer, indicating the total number of sensors included in the quantitative evaluation in the system; : sensor number, integer; : the first Normalized weights for class-based data, dimensionless, and The method for determining this parameter is based on the importance of different data types to disaster prediction, such as microseismic data and geostress data. It should be higher than production data; its value can be determined through regression analysis of historical disaster data or by domain experts. : No. The relative deviation of the data is dimensionless. : Real-time observations, as defined in step 1; The baseline prediction value has been defined in step 2; : No. Data smoothing coefficient, dimensions and Same; the method for determining this parameter is: according to The typical order of magnitude is set to a very small positive number, such as The average value is 10 MPa, then Can be set to MPa is used to ensure calculation stability and prevent the denominator from being zero. This embodiment constructs a dimensionless data pollution index through the above steps. This index enables real-time self-awareness of input data quality through weighted analysis. This reflects the heterogeneous importance of different data, through the smoothing coefficient. This ensures the robustness of the algorithm; it solves the problem of blindly trusting dirty data under high noise interference, paving the way for subsequent risk quantification and resilience entropy. It provides key, quantified inputs on uncertainty; In addition, to ensure the long-term validity of the baseline forecast, the baseline forecast model needs to be periodically retrained using verified clean historical data to adapt to the dynamic changes in mine geological conditions and mining activities, and to prevent systematic bias in the data contamination index caused by model drift.

[0021] Example 3: The resilience entropy assessment unit calculates the basic safety entropy through the following steps: Obtain factor observations of key physical risk factors; Obtain the preset factor safety baseline value corresponding to the factor observation value; Obtain the preset critical hazard value of the factor corresponding to the factor observation value; By combining the observed values ​​of the factors with the safety baseline values ​​of the factors, and by combining the critical hazard values ​​of the factors, normalization is performed to calculate the risk level of each factor. The risk levels of each factor are weighted and summed according to the preset factor weights to obtain the basic safety entropy; This embodiment, based on Embodiment 1, calculates the basic safety entropy for the resilience entropy assessment unit. The specific steps have been refined; the purpose of this calculation is to integrate different physical monitoring data, such as geostress. Micro-vibration energy gas concentration The units of measurement should be unified to construct a standardized basic risk indicator; The specific steps include: Obtain factor observations of key physical risk factors : Factor observation, referring to the first factor observation. Specific observed values ​​of key physical risk factors, such as The units are not uniform, and data is collected in real time by sensors; Obtain factor observations Corresponding preset factor safety benchmark value : Factor safety benchmark value, dimensions and Same, such as corresponding ; Obtain factor observations Corresponding preset factor critical risk value : : Critical hazard value of factor, dimensions and Same, such as corresponding ; and These are all preset thresholds, determined based on geological conditions and safety regulations, and must meet certain conditions. ; Combining factor observations With factor safety benchmark value And combined with the critical risk value of the factor After normalization, the risk level of each factor is calculated. ; The calculation model for this normalization process is as follows: The parameters in the formula are explained as follows: : No. Individual factor risk levels, dimensionless; through... Cut off to ensure it is non-negative; when When ; when , the critical danger value is exceeded; ; when , the critical danger value is exceeded; , indicating that the critical danger value has been exceeded; In this embodiment, a linear normalization model is used for simplifying calculation; in other embodiments, if there is an explicit nonlinear relationship, such as an exponential relationship, between a specific physical risk factor and disaster risk, a nonlinear model such as may be used for normalization processing, where is a nonlinear coefficient, to accurately depict the accelerated cumulative effect of risk; According to the preset factor weight , the risk degree of each factor is weighted and summed to obtain the basic safety entropy ; The calculation model of this weighted sum processing is as follows: The parameters in the formula are explained as follows: : basic safety entropy, dimensionless; the value is used as a basis for calculating the mine system resilience entropy ; : total number of key physical risk factors; : the weight of the i th factor, dimensionless, and ; the determination method of this parameter is to determine it through historical disaster data regression analysis or expert calibration method, reflecting the contribution of different factors to the total risk; : risk degree, dimensionless, calculated in the previous step; Through the above steps, the resilience entropy evaluation unit of this embodiment converts all physical factors of different dimensions into dimensionless risk metrics through the calculation of the risk degree , solves the problem of heterogeneous data being difficult to compare, and constructs the basic safety entropy through the weighted sum of the factor weight , realizes the comprehensive evaluation of the system physical risk, and provides a standardized and comparable benchmark for the subsequent introduction of the data pollution index for risk correction. Embodiment 4: The resilience entropy evaluation unit corrects the basic safety entropy, including the following steps:

[0022] ​​​Based on the preset pollution sensitivity coefficient and preset pollution impact index, and combined with the data pollution index, a nonlinear correction factor is constructed. Multiplying the basic safety entropy by the nonlinear correction factor yields the mine system resilience entropy; This embodiment, based on embodiment 3, adds a resilience entropy assessment unit to address the challenges posed by high-noise data environments, i.e., data contamination index. The problem of accurately estimating system risk at higher levels requires further analysis of the basic security entropy. Make corrections; The motivation for this correction is: basic security entropy. It is based on dirty data, that is The calculation shows that when data pollution is severe, i.e. High time, It may appear normal, but the system is actually facing significant risks due to data distortion; therefore, it is necessary to introduce... right Make corrections to achieve penetrating perception of the system's true safe distance; The specific steps include: Based on the preset pollution sensitivity coefficient Compared with the preset pollution impact index And combined with the data pollution index Construct a nonlinear correction factor; In this embodiment, the mathematical form of the nonlinear correction factor is: ; Basic security entropy Multiplying by the nonlinear correction factor yields the toughness entropy of the mine system. ; The revised calculation model is as follows: in, : Resilience entropy of a mining system, dimensionless and They have the same dimensions; their value is the final measure of the system's true risk and is used for subsequent risk level determination. Basic security entropy, dimensionless; Data pollution index, dimensionless; calculated and input in real time by the data pollution quantification unit; Pollution sensitivity coefficient, dimensionless. This parameter is used to adjust the system's sensitivity to data corruption. Pollution impact index, dimensionless, usually set as This parameter is used to reflect the amplifying effect of data contamination on risk; for example, it can be set to... , to reflect exponential amplification; With The determination method is: based on a calibration dataset containing historical disaster events Optimization fitting is performed; the dataset Each record in the dataset contains a set of historical monitoring data calculated from the historical basic safety entropy , historical data pollution index , and a true label representing whether a disaster has occurred , for example, 1 for occurrence and 0 for non-occurrence; calibration With The process is to find a set of values, so that the calculated based on these historical data achieves optimal early warning performance when predicting , such as the area under the receiver operating characteristic curve (AUC); When the data quality is good , the nonlinear correction factor , at this time , the system risk is determined by the physical index; when the data pollution is serious , the nonlinear correction factor , at this time will be nonlinearly amplified, that is, when does not change, the distorted by dirty data will increase rapidly; The embodiment introduces to nonlinearly correct , and obtains ; the physical risk and the uncertainty risk caused by data pollution are innovatively coupled in the same evaluation model; this solves the major defect of traditional safety entropy models that fail in high-noise environments due to reliance on dirty data, realizes conservative estimation of the system's true risk , and greatly improves the penetration and reliability of risk early warning.

[0023] Embodiment 5: The resilience entropy evaluation unit discriminates and processes the mine system resilience entropy, including the following steps: The mine system resilience entropy is compared with a preset first-level early warning threshold; The mine system resilience entropy is compared with a preset second-level danger threshold; When the mine system resilience entropy is less than the first-level early warning threshold, the risk level is determined as a normal state; When the mine system resilience entropy is greater than or equal to the first early warning threshold and less than the second danger threshold, the risk level is determined as the first early warning; When the mine system resilience entropy is greater than or equal to the second danger threshold, the risk level is determined as the second early warning; Based on the embodiment 4, the resilience entropy evaluation unit obtains the mine system resilience entropy After that, in order to convert it into a discrete signal available for the decision unit, further discrimination processing is performed on to determine the risk level; The system needs to preset two danger thresholds, which have the same dimension as and are dimensionless: : the first early warning threshold; when , it indicates that the system resilience decreases and the data uncertainty increases; : the second danger threshold; when , it indicates that the system has approached the danger boundary; and must satisfy ; The determination method of these two thresholds is: through analyzing the historical instability data, a value is determined, when exceeds the value, the historical decision accuracy decreases or the disaster risk significantly increases, which is taken as the calibration basis of and ; The specific discrimination steps include: The mine system resilience entropy is compared with the preset first early warning threshold ; The mine system resilience entropy is compared with the preset second danger threshold ; When , the risk level is determined as the normal state; When , the risk level is determined as the first early warning; When , the risk level is determined as the second early warning; This embodiment maps the continuous value into three discrete risk levels of the normal state, the first early warning, and the second early warning by setting two thresholds and , and provides a clear and explicit trigger signal for the subsequent anti-fragile decision unit, so that the decision logic, i.e. the switching of the risk aversion coefficient can be performed according to these determined levels, which greatly simplifies the complexity of the decision model and enhances the operability of the system.

[0024] Example 6: The utility function calculation for an antifragile decision-making unit includes the following steps: To obtain the expected economic benefits of the pre-set alternative decisions; Calculate the system collapse probability based on the resilience entropy of the mine system; The decision utility of the candidate decision is obtained by subtracting the product of the risk aversion coefficient and the system collapse probability from the expected economic benefits. Identify the candidate decision that maximizes the decision utility as the optimal decision; Based on Example 1, this embodiment, in order to resolve the conflict between mine safety and profitability, constructs and calculates a utility function based on risk decision theory to determine the optimal decision. ; The utility function The motivation is to provide a mathematical framework for systems to choose between short-term benefits and long-term survival; traditional decision-making pursues... This system pursues Its core idea is that when the probability of collapse... When increasing, through the risk aversion coefficient Utility Impose severe punishment; The specific solution steps include: Obtain preset candidate decisions Expected economic benefits : : Decision-making options, such as producing the optimal solution Or secure solution ; :decision making The expected economic benefits, in units of yuan; through a certain decision ,like The expected output is obtained by subtracting production costs from the estimated output; specifically, the decision... The expected economic benefits can be expressed as ,in The market price per unit of product. In decision-making The expected output is below This corresponds to the total production cost; for example, the optimal production solution. Corresponding to higher expected output And a safe solution For example, a slowdown mode corresponds to a lower expected output. ; Solving for the resilience entropy of the mine system System crash probability : System crash probability, dimensionless; a concept... an increasing function, Calculated by the resilience entropy assessment unit; for example, the function can be set as follows: The fitting coefficients Also based on the calibration dataset in Embodiment 4. The determination method is as follows: using statistical methods such as maximum likelihood estimation to find... Values ​​that make the function Best fit and The probability relationship between them; Expected economic benefits Subtract risk aversion coefficient With system crash probability The product of these factors yields the alternative decision. Decision utility ; The calculation model for this decision utility function is as follows: in, :decision making The utility, dimensions and Consistent, such as yuan; Risk aversion coefficient, dimensions and Consistent, such as element; this parameter is The function, the value of which is dynamically adjusted by this unit according to the risk level; and As defined in steps 1 and 2; Determine decision utility The corresponding decision when the maximum value is reached As the optimal decision ; Right now In other words, the decision engine will consider all potential decisions, such as... and Calculate their decision utility separately and And select the decision with the highest decision utility value as Output; This embodiment constructs a utility function. Economic benefits and risk losses, by Quantification unifies the comparison under the same mathematical framework; it provides a computable basis for the system to make a rational choice between profit maximization and collapse probability minimization, and makes the anti-fragile decision from a concept to an executable algorithm.

[0025] Embodiment 7: The anti-fragile decision unit dynamically adjusts the risk aversion coefficient, including the following steps: When the risk level is in the normal state, the risk aversion coefficient is set to a preset low coefficient value; When the risk level is in the first warning state, the risk aversion coefficient is switched to a preset high coefficient value; When the risk level is in the second warning state, the risk aversion coefficient is switched to a preset critical coefficient value; wherein the critical coefficient value is greater than the high coefficient value; Wherein, the high coefficient value ensures that the product of the risk aversion coefficient and the system collapse probability is greater than the expected economic benefit; This embodiment is based on embodiment 6, and the core step of the anti-fragile decision unit to implement the anti-fragile decision is to dynamically adjust the risk aversion coefficient according to the risk level ; It is not a fixed value, but or its corresponding risk level segmented function; the system presets three coefficient values: : low coefficient value, dimension of yuan; for example, yuan; this is a small value, allowing the system to make decisions in the normal state mainly dominated by ; : high coefficient value, dimension of yuan; for example, yuan; : critical coefficient value, dimension of yuan; for example, yuan; Wherein, ; And The determination method must meet a key logic: its value must ensure that the expected loss of the risk penalty term is much greater than any possible short-term gains, such as about yuan; The specific adjustment steps are as follows: When the risk level is in the normal state, that is, the resilience entropy evaluation unit determines , the risk aversion coefficient is set to a preset low coefficient value ; at this time ; since is small, Item is also small, , the decision will be biased to select The largest The production of optimal solution; When the risk level is a first warning, that is, , the risk aversion coefficient Switch to the preset high coefficient value ; At this time, the utility function is ; Since The maximum risk penalty term Item occupies an absolute dominant position in , ensuring that the optimal decision switches to the safe solution; When the risk level is a second warning, that is, , the risk aversion coefficient Switch to the preset critical coefficient value ; In this warning state, the utility function is ; Since The maximum The term occupies an absolute dominant position in , making Become a huge negative value; At this time, any , which Lightly increases , its Will be much smaller than The , the safe solution Aims to reduce , thereby reducing ; Wherein, the high coefficient value And the critical coefficient value The value should meet a key constraint: to ensure that in the warning state, for any decision to be selected to produce benefits, its risk penalty term is greater than its expected economic benefit, so that the final utility of the decision is always negative or lower than the utility of the safe decision, so as to force the system to select the safe solution as the optimal decision; This embodiment realizes the mutation of decision logic by setting the risk aversion coefficient As a nonlinear segmented function of the risk level; When the system detects that the risk ( ) Approaches the threshold , the system can automatically and instantaneously switch from the pursuit of benefit mode ( ) To the collapse avoidance mode ( ), and when the risk further deteriorates to the second warning threshold , it switches to the extreme avoidance mode with stronger punishment ( ); this stepwise non-linear switching forces the utility function to automatically switch from high-risk to conservative , implementing an antifragile decision.

[0026] Embodiment 8: The closed-loop scheduling control unit matches the preset scheduling instruction set, including the following steps: The preset scheduling instruction set includes production optimal mode instructions, first-level safety mode instructions, and second-level safety mode instructions; When the optimal decision is a production optimal solution and the risk level is a normal state, the hierarchical control instruction is determined as the production optimal mode instruction; When the optimal decision is a safety solution and the risk level is a first-level early warning, the hierarchical control instruction is switched to the first-level safety mode instruction; When the optimal decision is a safety solution and the risk level is a second-level early warning, the hierarchical control instruction is switched to the second-level safety mode instruction; This embodiment is based on Embodiment 7, and the closed-loop scheduling control unit is responsible for converting the optimal decision output by the antifragile decision unit , which has already implied risk level information, because is and is into specific operation instructions for the mine site; The specific steps of matching the preset scheduling instruction set by this unit are as follows: The preset scheduling instruction set includes: Production optimal mode instructions: in this embodiment, corresponding to the production optimal solution ; this instruction corresponds to a complex, intelligent tunneling mode that relies on accurate data, and pursues yield-cost optimization; First-level safety mode instructions: in this embodiment, corresponding to one of the safety solutions , denoted as ; this instruction corresponds to a traditional and conservative operation mode, such as reducing the tunneling speed by 30% and increasing the support density by 20%; Second-level safety mode instructions: in this embodiment, corresponding to a more conservative form of the safety solution , denoted as ; this instruction corresponds to suspending tunneling operations in high-risk areas and switching to a mode that relies less on data, such as manual observation; When the optimal decision is a production optimal solution and the risk level is a normal state, the hierarchical control instruction is determined as the production optimal mode instruction; in , , ​, the system executes ; When the optimal decision is the safe solution and the risk level is a first-level warning, the hierarchical control instruction is switched to a first-level safety mode instruction ; in , , , the system matches the instruction corresponding to the first-level warning in and executes; When the optimal decision is the safe solution and the risk level is a second-level warning, the hierarchical control instruction is switched to a second-level safety mode instruction ; in , , , the system matches the instruction corresponding to the second-level warning in and executes; The embodiment associates the abstract optimal decision output by the upstream decision unit with the actual operation mode of the mine, such as speed reduction, accurate shutdown, and automatic association, by predefining a hierarchical instruction set corresponding to the risk level and the decision target; the last mile from risk perception to anti-fragile decision and then to on-site physical execution is connected, and a complete closed loop from algorithm to control is realized.

[0027] Embodiment 9: The closed-loop scheduling control unit is further used for: After executing the first-level safety mode instruction or the second-level safety mode instruction, the mine system resilience entropy is continuously monitored; When the mine system resilience entropy falls below the first-level warning threshold, the risk aversion coefficient of the anti-fragile decision unit is restored to a low coefficient value; The hierarchical control instruction is switched back to the production optimal mode instruction; The embodiment is based on embodiment 8, to form a complete resilience management closed loop, that is, not only can it avoid danger, but also can restore safety, the closed-loop scheduling control unit also has a state recovery function and is linked with the anti-fragile decision unit; Specifically, After executing the first-level safety mode instruction or the second-level safety mode instruction , the system continuously monitors the mine system resilience entropy through the resilience entropy evaluation unit; the execution or , such as reducing the disturbance of tunneling, usually makes the physical risk factor decrease, thereby causing to decrease, and further decline; When the toughness entropy of the mining system Falling back to the Level 1 warning threshold After the following time and a period of stability, Antifragile decision-making units will use risk aversion coefficients Automatically restore to low-order coefficient value The system detected that the risk level had returned to normal, and the decision-making logic automatically switched back. ; Hierarchical control command to switch back to optimal production mode command This switch happens automatically because... Restore to Subsequently, in the utility function solution of Example 6, the antifragile decision-making unit... , making Become again The largest optimal decision The closed-loop scheduling control unit uses this information. Match and execute instruction; This embodiment ensures that the system can not only detect... When the risk increases, the system automatically switches to dimensionality reduction. It can still be executed. lead to After returning, automatically switch back to the previous dimension. This forms a complete dynamic resilience closed loop of perception-decision-execution-recovery, avoiding permanent waste of efficiency caused by excessive conservatism, and truly achieving a dynamic balance between safety and efficiency on the time scale.

[0028] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A big data analysis and prevention decision-making system for disasters in deep mines, characterized in that: include: The data pollution quantification unit is used to collect multi-source heterogeneous data streams and baseline prediction data, perform noise pollution quantification analysis, and obtain the data pollution index. The resilience entropy assessment unit is used to collect physical risk factors, calculate the basic safety entropy, and correct the basic safety entropy according to the data pollution index to obtain the mine system resilience entropy. The mine system resilience entropy is then processed to obtain the risk level. The antifragile decision-making unit is used to dynamically adjust the risk aversion coefficient based on the risk level and to perform utility function calculation on the preset alternative decisions to obtain the optimal decision; The closed-loop scheduling control unit is used to generate and execute hierarchical control instructions by matching a preset set of scheduling instructions with the optimal decision and risk level.

2. The deep mine disaster big data analysis and prevention decision-making system as described in claim 1, characterized in that, The data pollution quantification unit performs noise pollution quantification analysis, which includes the following steps: Acquire real-time observations from each sensor; Obtain the theoretical readings output by the benchmark prediction model trained on historical data; The relative deviation between real-time observations and theoretical readings is calculated to obtain the relative deviation of various types of data. The relative deviations of various types of data are weighted and summed according to the preset normalization weights to obtain the data pollution index.

3. The deep mine disaster big data analysis and prevention decision-making system as described in claim 1, characterized in that, The resilience entropy assessment unit calculates the basic safety entropy through the following steps: Obtain factor observations of key physical risk factors; Obtain the preset factor safety baseline value corresponding to the factor observation value; Obtain the preset critical hazard value of the factor corresponding to the factor observation value; By combining the observed values ​​of the factors with the safety baseline values ​​of the factors, and by combining the critical hazard values ​​of the factors, normalization is performed to calculate the risk level of each factor. The risk levels of each factor are weighted and summed according to the preset factor weights to obtain the basic safety entropy.

4. The deep mine disaster big data analysis and prevention decision-making system as described in claim 1, characterized in that, The resilience entropy assessment unit corrects the basic safety entropy through the following steps: Based on the preset pollution sensitivity coefficient and preset pollution impact index, and combined with the data pollution index, a nonlinear correction factor is constructed. Multiplying the basic safety entropy by the nonlinear correction factor yields the mine system resilience entropy.

5. The deep mine disaster big data analysis and prevention decision-making system as described in claim 1, characterized in that, The resilience entropy assessment unit performs the following steps to determine the resilience entropy of the mine system: The mine system resilience entropy is compared with the preset first-level early warning threshold; The mine system resilience entropy is compared with the preset secondary hazard threshold. When the mine system resilience entropy is less than the first-level early warning threshold, the risk level is determined to be the normal state; When the mine system's resilience entropy is greater than or equal to the Level 1 warning threshold and less than the Level 2 danger threshold, the risk level is determined to be Level 1 warning. When the resilience entropy of the mine system is greater than or equal to the level 2 hazard threshold, the risk level is determined to be level 2 warning.

6. The deep mine disaster big data analysis and prevention decision-making system as described in claim 1, characterized in that, The utility function calculation for an antifragile decision-making unit includes the following steps: To obtain the expected economic benefits of the pre-set alternative decisions; Calculate the system collapse probability based on the resilience entropy of the mine system; The decision utility of the candidate decision is obtained by subtracting the product of the risk aversion coefficient and the system collapse probability from the expected economic benefits. The candidate decision that maximizes the decision utility is identified as the optimal decision.

7. The deep mine disaster big data analysis and prevention decision-making system as described in claim 1, characterized in that, The dynamic adjustment of the risk aversion coefficient by the antifragile decision-making unit includes the following steps: When the risk level is normal, the risk aversion coefficient is set to a preset low coefficient value; When the risk level is at Level 1 warning, the risk aversion coefficient will be switched to a preset high coefficient value; When the risk level is at Level 2 warning, the risk aversion coefficient will be switched to the preset critical coefficient value; Among them, the high coefficient value ensures that the product of the risk aversion coefficient and the system collapse probability is greater than the expected economic benefits.

8. The deep mine disaster big data analysis and prevention decision-making system as described in claim 1, characterized in that, The closed-loop scheduling control unit matches the preset scheduling instruction set, including the following steps: The preset scheduling instruction set includes production optimal mode instructions, first-level safety mode instructions, and second-level safety mode instructions; When the optimal decision is the production optimal solution and the risk level is normal, the hierarchical control instruction will be determined as the production optimal mode instruction. When the optimal decision is a safe solution and the risk level is a Level 1 warning, the hierarchical control command will be switched to the Level 1 safety mode command. When the optimal decision is a safe solution and the risk level is a level 2 warning, the hierarchical control command will be switched to the level 2 safety mode command.

9. The deep mine disaster big data analysis and prevention decision-making system as described in claim 8, characterized in that, The closed-loop scheduling control unit is also used for: After executing Level 1 or Level 2 safety mode commands, continuously monitor the mine system resilience entropy; When the mine system resilience entropy falls below the first-level early warning threshold, the antifragile decision-making unit will restore the risk aversion coefficient to a low value. The hierarchical control command switches back to the optimal production mode.

Images