Non-coal mine safety state assessment method based on multi-dimensional data fusion

Abstract

The invention relates to the technical field of mine safety monitoring, and discloses a non-coal mine safety state assessment method based on multi-dimensional data fusion, which comprises the following steps: S1, constructing a multi-dimensional scoring model, constructing a multi-dimensional scoring model covering people, machines, rings and management, fusing multi-dimensional data and preprocessing and normalizing, and S2, dynamically adding and subtracting for updating. Based on a security event library rule, adding and subtracting the positive and negative security events, S3, performing multi-factor formula quantification, S4, performing graded and classified early warning, and S5, performing trend analysis and output. According to the invention, quantifiable and comparable evaluation of the safety state is realized. Through a unified score index, not only can the safety level of a single mine at a certain moment be objectively measured, but also the safety conditions of different mines in different time periods can be conveniently compared and analyzed, and weak links and advanced typical types can be found out. The multi-dimensional data fusion enables the scoring to be more comprehensive and accurate, reduces the false alarm and missing alarm caused by single monitoring, and improves the reliability of evaluation.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of mine safety monitoring, in particular to a non-coal mine safety state evaluation method based on multi-dimensional data fusion. BACKGROUND

[0002] The non-coal mine safety state refers to the safety degree and risk level jointly presented by personnel operation, equipment operation, working environment, management measures and other factors in the production and operation process of non-coal mines (such as metal mines, non-metal mines, quarries, etc.). It reflects whether the mine is in a state that can effectively avoid accidents, protect personnel safety and ensure normal production, covering different levels of no danger, low risk, medium risk, high risk, etc., and is a comprehensive representation of the overall safety production conditions and potential hazards of the mine.

[0003] Traditional non-coal mine safety monitoring mainly relies on single sensor early warning and manual inspection. For example, by installing independent gas, carbon monoxide, etc. sensors, each threshold value is reached to issue an alarm; at the same time, safety personnel are arranged to regularly inspect the mine for hidden dangers.

[0004] Current non-coal mine safety supervision has many shortcomings, mainly in the following aspects: high false alarm rate, incomplete safety behavior evaluation, and isolated data. The independent operation of each monitoring system makes the early warning information scattered, and the single threshold alarm method is prone to false alarms and omissions, which weakens the trust of management personnel in the early warning system; at the same time, there is a lack of effective evaluation means for personnel unsafe behavior, and it is impossible to comprehensively quantify human unsafe factors through post-inspection; in addition, data from different sources (personnel positioning, environmental monitoring, equipment status, etc.) are separated from each other, forming an information island, making it difficult to comprehensively analyze the safety situation, and therefore, we propose a non-coal mine safety state evaluation method based on multi-dimensional data fusion. SUMMARY

[0005] In view of the shortcomings of the prior art, the present application provides a non-coal mine safety state evaluation method based on multi-dimensional data fusion, which solves the problem of multiple shortcomings in the prior art of non-coal mine safety supervision.

[0006] To achieve the above purpose, the present application is realized by the following technical scheme: a non-coal mine safety state evaluation method based on multi-dimensional data fusion, comprising the following steps:

[0007] S1: Construct a multi-dimensional scoring model

[0008] A multi-dimensional scoring model covering "people, machines, environment, management" is constructed, multi-dimensional data is fused and preprocessed and normalized, and after calculating the sub-scores, a comprehensive safety score P is generated according to the weights;

[0009] S2: Dynamic plus-minus score update

[0010] Based on the multi-dimensional scoring model established in S1, the initial score of the multi-dimensional scoring model is set, and the positive and negative safety events are scored based on the safety event library rules, and the current score P(t) is updated in real time;

[0011] S3: Multi-factor formula quantification

[0012] A multi-factor scoring formula P = a·A + β·B + γ·C + δ·D is used to fuse the sub-scores of each dimension in S1 to obtain a comprehensive score, where a + β + γ + δ = 1;

[0013] S4: Hierarchical classification warning

[0014] Based on the comprehensive score in S3, the single score abnormality in S2, and the overall score trend change, a hierarchical classification warning is triggered and the corresponding notification strategy is executed;

[0015] S5: Trend analysis and output

[0016] The overall score history data in S1-S3 is stored, and trend charts, prediction reports, and comparative analysis results are generated according to the dynamic score trend in S2, and the S4 warning mechanism is linked.

[0017] Preferably, in the S1 of constructing a multi-dimensional scoring model, the multi-dimensional data includes personnel positioning data, safety monitoring data, equipment operation data, and implicit data, and the implicit data includes historical hidden danger records and operation procedure execution situation.

[0018] Preferably, the preprocessing normalization algorithm in S1 is as follows:

[0019] Different dimensions and different value ranges of data are converted into comparable indicators in the 0-1 or 0-100 interval, and the method includes:

[0020] Min-Max Normalization:

[0021] Formula:

[0022] Z-Score standardization:

[0023] Formula: (Translation mapping to 0-100)

[0024] Where μ is the mean and σ is the standard deviation, suitable for data following normal distribution, including equipment operation efficiency indicators.

[0025] Preferably, in the S1 construction of the multi-dimensional score model, the sub-scores include a personnel safety sub-score A, an environment safety sub-score B, a device operation safety sub-score C, and an implicit management data sub-score D, and the weight coefficients are set according to expert experience or obtained through machine learning algorithm based on historical data training.

[0026] Preferably, in the S2 dynamic score adding and subtracting updating, the initial score of the multi-dimensional score model is set as 100 points, the positive behaviors in the safety event library rules include standard wearing of protective equipment and qualified equipment maintenance, and the negative events include illegal entry into a forbidden area and over-limit of a dangerous gas.

[0027] Preferably, in the S2, the safety event library rules are divided into positive safety events and negative safety events according to event properties, the positive safety events correspond to adding items, and the negative safety events correspond to subtracting items, the positive safety events include personnel safety behavior, device management, and management measures, and the negative safety events include personnel illegal behavior, environment and device abnormality, and management omission.

[0028] Preferably, in the S3, A, B, C, and D respectively represent sub-scores of personnel, environment, device, and implicit data dimensions, and alpha, beta, gamma, and delta are weight coefficients of the corresponding dimensions, which are set according to the importance of each dimension to the overall safety or determined through data analysis.

[0029] Preferably, in the S4 graded classification early warning, when the comprehensive safety score P is lower than a preset threshold, the system immediately triggers a safety alarm to prompt that there is a major safety risk in the mine, and if any key monitoring parameter is over-limit or a major illegal behavior of personnel occurs, even if the comprehensive score has not yet fallen below the threshold, the early warning is also triggered.

[0030] Preferably, in the S4 graded classification early warning, the safety situation is judged in combination with the change trend of the score, and when the score continuously decreases for multiple periods or the decrease amplitude exceeds a certain proportion in a short time, even if the current score is still above the threshold, the system also issues a trend early warning in advance.

[0031] Preferably, in the S5 trend analysis and output, a trend analysis algorithm is used to process the score data, including calculating a moving average and a growth rate index, to predict the safety situation in the future short term, when the prediction result shows that the safety score will continue to decrease and approach an alarm line, the system feeds back this information to the S4 graded classification early warning, in addition, the automatic trend output also supports generating a comparative analysis report, including horizontally comparing the safety scores of different mines or different teams, and outputting a ranking and a differential index.

[0032] The application provides a non-coal mine safety state evaluation method based on multi-dimensional data fusion.

[0033] 1、The present application realizes the safety state quantifiable, the evaluation of contrast. Through the unified score index, not only can objective measure single mine in a moment of safety level, also facilitate the safety condition comparison and analysis between different mines, different time periods, find out the weak link and advanced typical. Multidimensional data fusion makes the score more comprehensive and accurate, reduces the false alarm caused by single monitoring, improves the reliability of evaluation.

[0034] 2、The present application has the initiative early warning function, enhances the foresight of accident prevention. The system not only alarms in time to the overrun condition, but also can perceive the risk sign in advance through trend prediction, issue early warning, and prevent the trouble from the bud. This change from passive response to active prevention helps to take measures to curb the evolution of hidden danger in time, and reduce accidents.

[0035] 3、The present application improves the supervision efficiency, with the help of the present application, the supervision personnel can quickly identify high risk points through the score and early warning information, without spending a lot of time for on-site inspection and data summary. Automatic data collection and analysis reduces the workload of manual sorting and judgment, and makes the supervision force change from "extensive net" to "precise fishing", and puts more effort into key supervision and hidden danger rectification. Overall, the method improves the efficiency of safety supervision and enterprise self-management.

[0036] 4、The present application supports differentiated management, according to the safety score and risk level, it is possible to implement hierarchical and classified supervision for mine enterprises. The supervision department can determine the inspection frequency and intensity according to the score: the mine with weak safety management ability (low score) is listed as the key supervision object, and the inspection frequency is strengthened; the unit with good safety condition can appropriately reduce the inspection frequency, so as to realize the reasonable allocation of supervision resources. At the same time, the enterprise can also evaluate the safety performance of each department according to the score, and promote the implementation of safety responsibility system. This differentiated and refined management method helps to improve the overall safety level of the industry BRIEF DESCRIPTION OF DRAWINGS

[0037] Figure 1 The present application is a non-coal mine safety state evaluation method based on multidimensional data fusion flow chart;

[0038] Figure 2 The present application is a safety event library scoring process schematic diagram. DETAILED DESCRIPTION

[0039] The technical solutions in the embodiments of the present application will be described in detail below with reference to the drawings in the present application specification. Obviously, the described embodiments are only part of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor belong to the scope of protection of the present application.

[0040] Embodiment:

[0041] Please refer to the attached Figure 1 - attached Figure 2 , the embodiment of the present application provides a kind of based on multi-dimensional data fusion non-coal mine safety state evaluation method, comprising the following steps:

[0042] S1: construct multi-dimensional score model

[0043] Multi-dimensional score model covering "people, machine, environment, management" is constructed, multi-dimensional data is fused and preprocessed normalization, and after calculating sub-score, comprehensive safety score P is generated according to weight;

[0044] S2: dynamic score updating

[0045] Based on the multi-dimensional score model established in S1, set the initial score of multi-dimensional score model and based on safety event library rule, positive and negative safety events are scored, and current score P (t) is updated in real time;

[0046] S3: multi-factor formula quantification

[0047] Multi-factor score formula P=α·A+β·B+γ·C+δ·D is used to fuse the sub-score of each dimension in S1 to obtain comprehensive score, wherein α+β+γ+δ=1;

[0048] S4: grading classification early warning

[0049] Based on the comprehensive score in S3, single score index exception in S2 and overall score trend change, trigger grading classification early warning and execute corresponding notification strategy;

[0050] S5: trend analysis and output

[0051] Overall score history data in S1-S3 is stored, and trend chart, prediction report and comparative analysis result are generated according to the dynamic score trend in S2, and S4 early warning mechanism is linked.

[0052] In the S1 of constructing multi-dimensional score model, multi-dimensional data includes personnel positioning data, safety monitoring data, equipment operation data and implicit data, and the implicit data includes historical hidden danger record, operation procedure execution situation.

[0053] The pre-processing normalization algorithm in S1 is as follows:

[0054] Different dimensions, different value range data are converted into 0-1 or 0-100 interval comparable indicators, and the method comprises:

[0055] Min-Max Normalization:

[0056] Formula:

[0057] Z-Score Standardization:

[0058] Formula: (Translated to 0-100)

[0059] Where μ is the mean and σ is the standard deviation, applicable to data following a normal distribution, including device operation efficiency indicators.

[0060] In the S1 construction of the multi-dimensional score model, the sub-scores include personnel safety sub-score A, environmental safety sub-score B, device operation safety sub-score C, and implicit management data sub-score D. The weight coefficients are set according to expert experience or obtained through machine learning algorithms based on historical data training, including the following algorithms:

[0061] I. Personnel Safety Sub-score (A)

[0062] Algorithm logic: deduct the total score of violations from the full score to reflect the impact of unsafe behavior.

[0063] Formula:

[0064]

[0065] Where:

[0066] A max = 100 (Personnel Dimension Full Score);

[0067] V i is the number of occurrences of the i-th type of violation (e.g., "not wearing a safety helmet" occurs 2 times, V i = 2);

[0068] k i is the deduction coefficient of the violation (e.g., "violation of restricted area" k i = 20).

[0069] II. Environmental Safety Sub-score (B)

[0070] Algorithm logic: based on the degree of deviation of environmental parameters from the safety threshold, weighted calculation of the score.

[0071] Formula:

[0072]

[0073] Where:

[0074] n is the number of environmental parameters (such as gas, carbon monoxide, etc.);

[0075] x jCurrent value of the jth parameter (e.g. gas concentration 0.6%);

[0076] T j Safety threshold of the parameter (e.g. gas threshold 1.0%);

[0077] w j Parameter weight (e.g. gas w j = 0.4, carbon monoxide w j = 0.3, satisfying ∑w j = 1)

[0078] Three, equipment operation safety sub-score (C)

[0079] Algorithm logic: combine with equipment reliability index (failure rate, maintenance on time rate) calculation, positive indicators add points, negative indicators deduct points.

[0080] Formula:

[0081] C = 100 - (k f × f + k d × (1 - d))

[0082] Wherein:

[0083] f is the equipment failure rate (e.g. monthly failure times / total running times);

[0084] d is the maintenance on time rate (e.g. on time completion of maintenance times / total maintenance times);

[0085] k f , k d is the weight coefficient (e.g. k f = 50, k d = 30, reflecting the greater impact of failure rate).

[0086] Four, implicit management data sub-score (D)

[0087] Algorithm logic: based on the implementation of management measures (hidden danger investigation, training, etc.), quantify the safety level of management dimension.

[0088] Formula:

[0089] D = k h × h + k t × t + k p × p wherein:

[0090] h is the hidden danger rectification rate (hidden danger rectification number / total hidden danger number);

[0091] t is the safety training qualified rate (qualified number / total number);

[0092] p is the operation procedure execution rate (compliant operation times / total operation times).

[0093] k h 、k t 、k p are weight coefficients (such as k h = 0.5, k t = 0.3, k p = 0.2, satisfying ∑k = 1), and the final result is multiplied by 100 to convert to percentage.

[0094] In the S2 dynamic plus-minus score updating, the initial score of the multi-dimensional scoring model is set to 100 points, the positive behaviors in the safety event library rules include standard wearing of protective equipment and qualified equipment maintenance, and the negative events include illegal entry into restricted areas and dangerous gas over-limit. Through the plus-minus score mechanism, the system can finely depict the change of safety state over time, realize real-time quantitative evaluation of safety status, and this mechanism ensures that the scoring model is sensitive to emergencies and timely reflects the field dynamics, and at the same time, the score can be restored after short-term fluctuations, avoiding long-term lag.

[0095] The safety event library rules in S2 are divided into positive safety events and negative safety events according to the nature of the events, wherein the positive safety events correspond to the plus items, and the negative safety events correspond to the minus items, wherein the positive safety events include personnel safety behavior class, equipment management class and management measure class, and the negative safety events include personnel violation behavior class, environment and equipment anomaly class and management oversight class, including the following steps:

[0096] Step 1: initialization of comprehensive score and safety event library rule loading

[0097] Firstly, the initial comprehensive safety score of the multi-dimensional scoring model is set to 100 points (i.e. P base = 100 points) as the benchmark value for scoring calculation; at the same time, the pre-set safety event library rules are loaded, which are divided into positive safety events (plus items) and negative safety events (minus items) according to the nature of the events, wherein the positive safety events cover personnel safety behavior class, equipment management class and management measure class, and the negative safety events cover personnel violation behavior class, environment and equipment anomaly class and management oversight class, and each class of events corresponds to clear score influence parameters (such as basic score, weight coefficient, etc.), for example, “standard wearing of protective equipment” corresponds to a basic plus score k add1 = 5 points, and “illegal entry into restricted areas” corresponds to a basic minus score k subl = 20 points.

[0098] Step 2: real-time monitoring and event triggering determination

[0099] The system collects multi-dimensional data of “person, machine, environment, management” in real time, and identifies whether a safety event occurs and the event level through an event triggering determination algorithm: for environmental parameters (such as gas concentration xj ), device status, etc. can be monitored in real time, and threshold comparison algorithm is used for determination. If a certain environmental parameter x j ≥ T j × k alert (where T j is the safety threshold of the parameter, and k alert = 0.75 is the early warning coefficient), it is determined that a negative event is triggered, for example, when the gas concentration x j = 0.8% and the threshold T j = 1.0%, because 0.8 ≥ 1.0 × 0.75, the negative event of “slight overrun of dangerous gas” is triggered; for events that need to be confirmed by human, such as hidden danger investigation and rectification, safety training, etc., a compliance determination algorithm is used. If the event completion rate (such as training participation rate r = actual number of participants / number of participants) satisfies r ≥ 0.9 (qualified coefficient k pass = 0.9), it is determined that a positive event is triggered, for example, when the training participation rate r = 0.95, the positive event of “safety training meets the standard” is triggered.

[0100] Step 3: Calculate the plus-minus score of a single event

[0101] For the triggered safety events, the plus-minus score of a single event is calculated according to the event type and level: for positive events, the plus score formula ΔP add = w i × k i × q i is used, where w i is the event type weight (such as personnel safety behavior class w 人 = 0.4), k i is the event basic score (such as “device maintenance qualified” k 设 = 15 points), and q i is the event execution quality (such as maintenance acceptance qualified rate q = 0.9). For example, the single plus score of “device maintenance qualified” is ΔP add = 0.3 × 15 × 0.9 = 4.05 points; for negative events, the deduction formula ΔP sub = w j × k j × s j is used, where w j is the event type weight (such as environment and device anomaly class w 环 = 0.5), k j is the event basic score (such as “illegal entry into restricted area” k 违 = 20 points), and s j is the event severity (such as the number coefficient of illegal entry into restricted area s = 1). For example, the single deduction score of “illegal entry into restricted area” is ΔP sub = 0.3 × 20 × 1 = 6 points.

[0102] Step 4: Processing the cumulative effect of events and updating the real-time score

[0103] If the same event occurs repeatedly within a preset period (such as 24 hours), the cumulative effect needs to be calculated: for positive events, use the continuous scoring formula ΔP add_cum = ΔP add ×(1+c / k cap ), where c is the number of consecutive compliance times (such as 3 consecutive days of standard protective equipment c = 3), k cap = 10 is the upper limit coefficient of the score, for example, the third "standard protective equipment" score is ΔP add_cum = 5 × (1 + 3 / 10) = 6.5 points; for negative events, use the progressive deduction formula ΔP sub_cum = ΔP sub ×(1+r / k scale ), where r is the number of repetitions (such as the second violation of the restricted area within 1 month r = 2), k scale = 2 is the progressive coefficient, for example, the second "violation of the restricted area" deduction is ΔP sub_cum = 6 × (1 + 2 / 2) = 12 points; then combine the cumulative score, and calculate the current score by the real-time score update formula P(t) = P(t-1) + ΔP add_cum - ΔP sub_cum , where P(t-1) is the score at the last time, for example, the last time score P(t-1) = 95 points, the current cumulative score 6.5 points, and the cumulative deduction 12 points, then the current score P(t) = 95 + 6.5 - 12 = 89.5 points.

[0104] Step 5: Introducing time decay mechanism to dynamically adjust the score

[0105] To avoid the long-term over-influence of historical events on the current score, a time decay mechanism is introduced to update the score by the formula , where λ = 0.99 is the decay coefficient (representing a 1% decay per day), t is the current time, t i , t j are the times of positive and negative events, respectively, for example, the "equipment maintenance qualified" score (ΔP add_i = 4.05 points) occurred 3 days ago, and the decayed score at the current time is 4.05 × 0.99 3 ≈ 3.93 points, while the "violation of the restricted area" deduction (ΔP sub_j = 6 points) occurred 1 day ago, and the decayed score is 6 × 0.99 1 ≈ 5.94 points, combined with the initial score and the decayed value of other events, the final dynamically adjusted comprehensive safety score is obtained.

[0106] A, B, C, D in S3 respectively represent sub-scores of personnel, environment, equipment and implicit data dimensions, and a, b, g and d are weight coefficients of the corresponding dimensions, which are set according to the importance of each dimension to the overall safety or determined through data analysis.

[0107] When the comprehensive safety score P is lower than the preset threshold, the system immediately triggers a safety alarm to prompt that there is a major safety risk in the mine. For example, the safety score of 70 points can be set as the early warning threshold, and a red alarm is issued when the score falls below 70. In addition, the present application considers the case that the dramatic abnormality of some single indicators can be offset by the comprehensive score, and therefore sets a multi-condition early warning rule: on the one hand, if any key monitoring parameter is out of limit (such as the gas concentration exceeds the critical value) or a major violation of personnel behavior occurs, even if the comprehensive score has not yet fallen below the threshold, the early warning should also be triggered; on the other hand, the safety situation is judged in combination with the change trend of the score, and when the score continuously decreases for multiple periods or the decrease amplitude exceeds a certain proportion in a short time, the system can issue a trend early warning in advance. For example, let AP = P(t)-P(t-1) be the single-period change amount of the score, if there are several large negative amplitudes in succession, it is determined that the safety state is deteriorating rapidly, even if the current score is still above the threshold, an orange early warning prompt is issued in time. The early warning logic also includes grading and classification of early warning information: according to the severity of the risk, it is divided into general early warning and major early warning, different levels correspond to different notification strategies (such as general early warning reminding the on-site team to rectify, major early warning reporting to the supervision center), through the above early warning logic, the present application can realize the transformation from passive alarm to active early warning, prompting intervention measures before the danger develops into an accident, and improving the forward-looking of safety management.

[0108] In the S5 trend analysis and output, the score data is processed by using a trend analysis algorithm, including calculating a moving average, a growth rate index, predicting a safety situation in a short term in the future, and feeding back information to the S4 grading classification warning when the prediction result shows that the safety score will continue to decline and approach the warning line. A manager can intuitively view the safety score daily, weekly, and monthly change trajectory of a mine or work area, and identify the improvement or deterioration trend of the safety situation. On the other hand, the system processes the score data by using a trend analysis algorithm, for example, calculates a moving average, a growth rate, and the like, predicts a safety situation in a short term in the future, and feeds back information to the warning mechanism when the prediction result shows that the safety score may continue to decline and approach the warning line. In addition, the automatic trend output also supports generating a comparative analysis report, for example, comparing the safety scores of different mines or different teams horizontally, outputting a ranking and a differentiated index. This visual and automatic trend output function facilitates the supervision department and the enterprise manager to master the macro safety dynamics, provides data support for differentiated supervision, and at the same time, all the score and warning data can be archived for reference, used for accident investigation and analysis, and the basis for safety management decision, including the following algorithms:

[0109] I. Moving average algorithm (smooth historical score data) To eliminate the interference of short-term fluctuations on trend analysis, a moving average algorithm is used to smooth the historical comprehensive safety score, and the calculation formula is as follows:

[0110]

[0111] Wherein: MA(t, n) represents the n-period moving average value at the t time (for example, n = 7 for daily moving average, and n = 4 for weekly moving average); P(t-i) represents the original comprehensive safety score at the t-i time; and n is the moving window size (set according to the data granularity, for example, n = 24 for hourly monitoring, and n = 7 for daily monitoring). Text explanation: By calculating the average value of scores in consecutive n periods, the score fluctuations of a single day or hour are smoothed, and the medium and long-term trend of the safety state is more clearly reflected. For example, if the scores of a mine in the last 7 days are [90, 88, 92, 85, 87, 83, and 80], the 7-day moving average value is 87.14, which represents the overall safety level of this week.

[0112]

[0113]

[0114] II. Growth rate calculation algorithm (quantify the trend change amplitude)

[0115] By calculating the growth rate index of the score, the improvement or deterioration speed of the safety state is judged, and the core formula includes:

[0116] Single-period growth rate:​

[0117] Multi-period average growth rate:

[0118] Where:

[0119] G(t) represents the growth rate of the tth period relative to the (t-1)th period (positive value for improvement, negative value for deterioration);

[0120] AG(t, n) represents the average growth rate of the last n periods (e.g., n=3 represents the average change amplitude of the last 3 days).

[0121] Text interpretation: Determine the short-term fluctuation direction by single-period growth rate, and identify the trend stability by multi-period average growth rate. For example, if the score growth rate for the last 3 days is [-2%, -3%, -5%], the 3-day average growth rate AG(t, 3) = (-2% -3% -5%) / 3 = -3.3%, indicating that the safety state continues to deteriorate and the amplitude expands.

[0122] Three, short-term prediction algorithm (predict future safety score)

[0123] Use linear regression or exponential smoothing algorithm to predict the safety score in the future short term, the core formula is as follows:

[0124] Linear regression prediction (suitable for trend stable scenarios):

[0125] Where, regression coefficients a (intercept) and b (slope) are obtained by least squares fitting of historical data:

[0126] ( is the mean of historical scores, is the mean of time variable, k is the prediction period, e.g., k=1 represents prediction of the next 1 day)

[0127] Exponential smoothing prediction (suitable for scenarios with large trend fluctuations):

[0128] Where: -α is the smoothing coefficient (0<α<1, e.g., α=0.7 means more emphasis on recent data); is the predicted value of the tth period, P(t) is the actual value of the tth period.

[0129] Text interpretation: By fitting the time trend of historical scores, the score of the next k periods is predicted. For example, if linear regression gives a=105, b=-2, then the score after 3 days is If the current t = 10, the predicted value is 105 - 2 x 13 = 79 points. When the predicted value is equal to or less than the early warning threshold (EWT), the result is fed back to the S4 early warning module to trigger early warning.

[0130] While the embodiments of the application have been illustrated and described, it will be understood by those skilled in the art that various changes, modifications, substitutions, and alterations can be made therein without departing from the spirit and scope of the application, which is defined by the appended claims and their equivalents.

Claims

1. A method for assessing the safety status of non-coal mines based on multi-dimensional data fusion, characterized in that, Includes the following steps: S1: Constructing a multidimensional scoring model A multi-dimensional scoring model covering "people, machines, environment, and management" is constructed, multi-dimensional data is integrated and preprocessed and normalized, and after calculating sub-scores, a comprehensive safety score P is generated according to the weights. S2: Dynamic score update Based on the multidimensional scoring model established in S1, the initial score of the multidimensional scoring model is set and the positive and negative security events are added or subtracted based on the rules of the security event database, and the current score P(t) is updated in real time. S3: Multi-factor formula quantification The comprehensive score is obtained by fusing the sub-scores of each dimension in S1 using the multi-factor scoring formula P=α·A+β·B+γ·C+δ·D, where α+β+γ+δ=1; S4: Tiered and Categorized Early Warning Based on the comprehensive score in S3, the anomalies of single addition and subtraction indicators in S2, and the overall score trend changes, a hierarchical and classified early warning is triggered and the corresponding notification strategy is executed. S5: Trend Analysis and Output Store historical overall scores from S1 to S3, and generate trend charts, forecast reports, and comparative analysis results based on the dynamic scoring trends in S2, linking with the S4 early warning mechanism.

2. The method for assessing the safety status of non-coal mines based on multi-dimensional data fusion according to claim 1, characterized in that, In the S1 multidimensional scoring model, the multidimensional data includes personnel location data, safety monitoring data, equipment operation data, and implicit data. The implicit data includes historical hazard records and the execution status of operating procedures.

3. The method for assessing the safety status of non-coal mines based on multi-dimensional data fusion according to claim 1, characterized in that, The preprocessing normalization algorithm in S1 is as follows: Methods for converting data of different units and value ranges into comparable indicators in the range of 0-1 or 0-100 include: Min-Max Normalization: official: Z-Score standardization: formula: (Mapped to 0-100 after translation) Where μ is the mean and σ is the standard deviation, it is applicable to data that follows a normal distribution, including equipment operating efficiency indicators.

4. The method for assessing the safety status of non-coal mines based on multi-dimensional data fusion according to claim 1, characterized in that, In the multidimensional scoring model constructed by S1, the sub-scorings include personnel safety sub-scoring A, environmental safety sub-scoring B, equipment operation safety sub-scoring C, and implicit management data sub-scoring D. The weight coefficients are set according to expert experience or obtained by training based on historical data through machine learning algorithms.

5. The method for assessing the safety status of non-coal mines based on multi-dimensional data fusion according to claim 1, characterized in that, In the S2 dynamic scoring update, the initial score of the multidimensional scoring model is set to 100 points. Positive behaviors in the safety event database rules include proper wearing of protective equipment and qualified equipment maintenance, while negative events include unauthorized entry into restricted areas and exceeding limits for hazardous gases.

6. The method for assessing the safety status of non-coal mines based on multi-dimensional data fusion according to claim 1, characterized in that, The rules in the S2 safety event database are divided into positive safety events and negative safety events according to the nature of the events. Positive safety events correspond to bonus items, and negative safety events correspond to deduction items. Positive safety events include: personnel safety behavior, equipment management, and management measures. Negative safety events include: personnel violations, environmental and equipment anomalies, and management oversights.

7. The method for assessing the safety status of non-coal mines based on multi-dimensional data fusion according to claim 1, characterized in that, In S3, A, B, C, and D represent sub-scores for personnel, environment, equipment, and implicit data dimensions, respectively, while α, β, γ, and δ are the weight coefficients for the corresponding dimensions. These are set based on the importance of each dimension to overall safety or determined through data analysis.

8. The method for assessing the safety status of non-coal mines based on multi-dimensional data fusion according to claim 1, characterized in that, In the S4 graded classification early warning system, when the comprehensive safety score P is lower than the preset threshold, the system immediately triggers a safety alarm, indicating that there is a major safety risk in the mine. If any key monitoring parameter exceeds the limit or a major violation by personnel occurs, an early warning will be triggered even if the comprehensive score has not yet fallen below the threshold.

9. The method for assessing the safety status of non-coal mines based on multi-dimensional data fusion according to claim 1, characterized in that, In the S4 graded classification early warning system, the security situation is judged by the changing trend of the score. When the score declines continuously for multiple periods or the drop exceeds a certain percentage in a short period of time, the system will issue a trend warning in advance, even if the current score is still above the threshold.

10. The method for assessing the safety status of non-coal mines based on multi-dimensional data fusion according to claim 1, characterized in that, In the S5 trend analysis and output, trend analysis algorithms are used to process the scoring data, including calculating moving averages and growth rate indicators, to predict the safety situation in the short term. When the prediction results show that the safety score will continue to decline and approach the warning line, the system feeds this information back to the S4 graded classification early warning. In addition, the automatic trend output also supports the generation of comparative analysis reports, including horizontal comparison of the safety scores of different mines or different work groups, and output rankings and differentiated indicators.

Images