Mine safety risk early warning method and system

Abstract

The invention relates to the technical field of risk early warning, in particular to a mine safety risk early warning method and system, and the method comprises the following steps: screening abnormal behavior data related to tunnel stability monitoring based on real-time collected monitoring data, updating a behavior pattern library through employing a PrefixSpan algorithm, recognizing sequence features matched with historical accident data, and carrying out the early warning of the safety risk of a mine. And generating a sequence behavior early warning index, and performing time sequence analysis on the historical harmful gas detection data based on the sequence behavior early warning index. According to the method, a risk mode is accurately identified through real-time data flow analysis, faster early warning of potential dangers is realized, risk factors can be dynamically evaluated through combination of sequence feature identification and environmental change analysis, the real-time performance and correlation of early warning information are guaranteed, and the early warning efficiency is improved by refining a risk area and optimizing a data acquisition strategy. The operation efficiency and the response speed of the whole safety monitoring system are improved, and it is ensured that quick and accurate response can be made at the critical moment.

Description

Technical Field

[0001] The present invention relates to the technical field of risk warning, and in particular to a mine safety risk warning method and system. Background Art

[0002] The field of risk warning technology is a field that focuses on identifying, analyzing and warning of potential dangers in advance to improve safety and reduce accidents. In mine safety management, because mining operations face many potential risks, including collapse, explosion, water damage and harmful gas leakage, by integrating sensor technology, data analysis and machine learning algorithms, it is possible to monitor the environment and operating conditions of the mine in real time, and conduct real-time analysis and prediction of abnormal situations, so as to send early warning signals to staff in time, reduce accidents and ensure the life safety and health of miners.

[0003] Among them, the mine safety risk warning method aims to prevent and reduce safety accidents in mining operations through technical means. It usually includes the deployment of multiple sensors, such as gas detectors, pressure sensors and geological radars, to monitor various potential dangerous factors inside the mine. The data collected by the sensors, combined with data analysis technology and algorithms, can evaluate the risk factors that lead to disasters, thereby providing early warning information in real time to help personnel take risk avoidance measures in time, thereby protecting personnel's lives and minimizing economic losses.

[0004] The ability of existing technologies to identify and respond to environmental changes is often limited by processing speed, resulting in the inability to timely update risk assessments or adjust early warning strategies in emergency situations. Existing technologies are cumbersome when integrating new data sources or adapting to new monitoring needs, and lack sufficient flexibility to adapt to different operating environments or security issues, causing the early warning system to fail at critical moments and be unable to provide the necessary safety guarantees, increasing the risk of casualties and potential economic losses. Summary of the invention

[0005] The purpose of the present invention is to solve the shortcomings of the prior art and to propose a mine safety risk early warning method and system.

[0006] In order to achieve the above object, the present invention adopts the following technical solution: a mine safety risk early warning method, comprising the following steps:

[0007] S1: Based on the real-time collected monitoring data, screen the abnormal behavior data associated with tunnel stability monitoring, use the PrefixSpan algorithm to update the behavior pattern library, identify the sequence features that match the historical accident data, and generate sequence behavior warning indicators;

[0008] S2: Based on the sequence behavior warning indicators, perform time series analysis on historical harmful gas detection data, associate the update frequency with environmental change data, evaluate the risk level of environmental changes, and generate an environmental risk level;

[0009] S3: Based on the environmental risk level, extract the abnormal pattern in the miner positioning data stream, use the Apriori algorithm to perform real-time analysis on the abnormal pattern, determine the risk area, adjust the monitoring parameters and optimize the data collection according to the analysis results, and generate the security monitoring optimization result;

[0010] S4: Based on the security monitoring optimization results, threshold analysis is performed on real-time data to determine the risk level, and corresponding warning signals are issued using CEP technology to generate real-time warning response signals.

[0011] The present invention is improved in that the step of screening the abnormal behavior data is specifically as follows:

[0012] S111: Based on the real-time collected monitoring data, the original monitoring data associated with the tunnel stability monitoring is collected, and for each data point, the deviation from the mean of the historical data is calculated using the formula:

[0013] Δx i =x i ―μ

[0014] Get a list of deviation values, where x i represents a single monitoring data point, μ represents the mean of historical data, Δx i Represents data point x i Deviation value of

[0015] S112: Based on the deviation value list, threshold screening is performed to identify abnormal data points using the formula:

[0016] SD = {Δx i ||Δx i |>k·σ}

[0017] Get the abnormal data point set SD, where Δx i For data point x i The deviation value, k is the adjustment coefficient, and σ is the standard deviation of the deviation value;

[0018] S113: Analyze the abnormal data point set using the formula:

[0019]

[0020] Get the abnormal behavior data SE, where d represents the data point in the abnormal data point set, and w d is the weight coefficient, and SD is the set of abnormal data points.

[0021] The present invention is improved in that the step of obtaining the sequence behavior warning indicator is specifically as follows:

[0022] S121: Use the PrefixSpan algorithm to process the behavior pattern library and identify frequent sequence patterns, using the formula:

[0023] QS2=PrefixSpan(M Q ,Qβ min )

[0024] Get the frequent sequence pattern set QS2, where M Q is the behavior pattern library, Qβ min is the minimum support threshold;

[0025] S122: Based on the frequent sequence pattern set, matching is performed with historical accident data using the formula:

[0026] QF={qs∈QS2|match(qs,H Q )×p Q >α Q}

[0027] Get the matching sequence feature set QF, where qs is the sequence pattern, H Q is the historical accident dataset, p Q is the probability of occurrence, α Q is the matching threshold, QS2 is the set of frequent sequence patterns, match(qs,H Q ) represents the sequence pattern qs and the historical accident data set H Q The degree of match between

[0028] S123: Perform risk scoring on the matched sequence feature set using the formula:

[0029]

[0030] Get the sequence behavior warning indicator QR, where f Q is the behavior pattern in QF, w Qf is the weight, risk(f Q ) is the potential risk score of the behavior pattern, c Q is the constant risk adjustment term.

[0031] The present invention is improved in that the steps of performing time series analysis are specifically as follows:

[0032] S211: Based on the sequence behavior warning indicator, collect and organize historical harmful gas detection data, filter out noise and remove abnormal values ​​from the data, and obtain cleaned data;

[0033] S212: Based on the cleaned data, perform time series analysis to detect trends and periodic changes in the data, apply the ARMA model, and use the formula:

[0034] X trend =ARMA(Z clean ,p r ,q r )

[0035] Get time series trend data X trend , where p r is the autoregressive order, q r is the moving average order, Z clean It is the cleaned data.

[0036] The present invention is improved in that the steps of obtaining the environmental risk level are specifically as follows:

[0037] S221: Based on the time series analysis results, combined with the real-time monitored environmental change data, use the formula:

[0038] I env =ω1X trend +ω2EQ t

[0039] Get the environmental index I env , where X trend is time series trend data, EQ t is the real-time environmental change data, ω1 and ω2 are weight factors;

[0040] S222: Based on the environmental index, the risk level of environmental change is evaluated using the judgment formula:

[0041]

[0042] Get the environmental risk level RQ, where I env is the environmental index, θ high and θ low are high and low risk thresholds.

[0043] The present invention is improved in that the step of performing real-time analysis on the abnormal pattern is specifically as follows:

[0044] S311: Based on the environmental risk level, collect the original location data of the miners, process the data, and use the formula:

[0045]

[0046] Get standardized data XD clean (t), where XD(t) is the original position data at time point t, is the mean of the data set, s XD is the standard deviation of the data set;

[0047] S312: Based on the standardized data, abnormal pattern recognition is performed using the LOF algorithm and the formula:

[0048] M abnormal (t) = LOF(XD clean (t),k v )>1.5,

[0049] Get the abnormal data point mark M abnormal (t), where XD clean (t) is the standardized data, k v is the number of neighbors used in the LOF algorithm;

[0050] S313: Based on the abnormal data point mark, the relationship between abnormal patterns is identified by association rule mining, using the formula:

[0051]

[0052] Get the support degree of the association rule Support(X A ), where X A is the item set of abnormal patterns, N is the total number of transactions, M abnormal (t) is the abnormal data point mark.

[0053] The present invention is improved in that the steps of obtaining the security monitoring optimization result are specifically as follows:

[0054] S321: According to the analysis results, adjust the sensitivity of the monitoring process to optimize the response capability to potential risks, using the formula:

[0055] KP new =KP old +α K ×(KP target ―KP old )

[0056] Get the new monitoring parameter value KP new , where KP target is the target parameter value, KP old is the currently monitored parameter value, α K is the smoothing factor;

[0057] S322: Based on the new monitoring parameter value, reconfigure the data collection frequency, and optimize resource allocation and data quality, using the formula:

[0058] KC new =KC old×(1+γ K ×IFactor)

[0059] Get the new data collection frequency KC new , among which, KC old is the original data collection frequency, IFactor is the impact factor, γ K It is the adjustment coefficient, which is used to control the variation of the acquisition frequency.

[0060] The present invention is improved in that the step of acquiring the real-time early warning response signal is specifically as follows:

[0061] S411: Based on the security monitoring optimization result, compare with the real-time data, calculate the deviation between the real-time data and the threshold, and use the formula:

[0062]

[0063] Get the initial warning sensitivity Gs, where G d Indicates the current real-time monitoring data, G u is the threshold value;

[0064] S412: Based on the preliminary warning sensitivity and combined with the abnormal pattern recognition result, the formula is adopted:

[0065]

[0066] Get the weighted value G of the warning signal w , where Gs is the initial warning sensitivity, p C is the probability of occurrence of abnormal patterns, α C is the adjustment coefficient;

[0067] S413: Based on the weighted value of the warning signal and the current risk level, the formula is adopted:

[0068] GR=min(G w +r Z ,1)

[0069] Get the real-time warning response signal GR, where G w is the weighted value of the early warning signal, r Z is the current risk level.

[0070] A mine safety risk early warning system, the system comprising:

[0071] The behavior pattern analysis module screens data anomalies associated with tunnel stability based on real-time collected monitoring data, updates the behavior pattern library using the PrefixSpan algorithm, identifies sequence features that match historical accident data, and generates sequence behavior warning indicators;

[0072] The environmental risk analysis module analyzes historical harmful gas detection data based on the sequence behavior warning indicators, associates and analyzes the update frequency with environmental change data, evaluates the risk level of environmental changes, and generates an environmental risk level;

[0073] The abnormal pattern recognition module extracts abnormal patterns from the miner positioning data stream based on the environmental risk level, uses the Apriori algorithm for real-time analysis, and generates preliminary risk area recognition results;

[0074] The monitoring parameter optimization module determines the risk area based on the preliminary risk area identification result, adjusts the monitoring parameters and optimizes the data collection, and generates the security monitoring optimization result;

[0075] The real-time warning transmission module performs threshold analysis on the real-time data based on the security monitoring optimization result, makes a risk level judgment, and uses CEP technology to send out a warning signal to generate a real-time warning response signal.

[0076] Compared with the prior art, the advantages and positive effects of the present invention are:

[0077] In the present invention, risk patterns are accurately identified through real-time data flow analysis, achieving faster early warning of potential dangers. By combining sequence feature recognition with environmental change analysis, risk factors can be dynamically evaluated to ensure the real-time and relevance of early warning information. By refining risk areas and optimizing data collection strategies, the operating efficiency and response speed of the entire safety monitoring system are improved, ensuring that a quick and accurate response can be made at critical moments, significantly reducing the possibility of accidents. BRIEF DESCRIPTION OF THE DRAWINGS

[0078] Figure 1 A flow chart of a mine safety risk early warning method is proposed for the present invention;

[0079] Figure 2 This is a flowchart for screening abnormal behavior data in the present invention;

[0080] Figure 3 This is a flow chart for obtaining the sequence behavior warning indicator in the present invention;

[0081] Figure 4 A flowchart of time series analysis in the present invention;

[0082] Figure 5 This is a flow chart for obtaining the environmental risk level in the present invention;

[0083] Figure 6 It is a flow chart of real-time analysis of abnormal patterns in the present invention;

[0084] Figure 7A flowchart for obtaining the security monitoring optimization results in the present invention;

[0085] Figure 8 This is a flow chart for obtaining the real-time early warning response signal in the present invention. DETAILED DESCRIPTION

[0086] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0087] In the description of the present invention, it should be understood that the terms "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside" and the like indicate positions or positional relationships based on the positions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, in the description of the present invention, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0088] Example

[0089] See also Figure 1 The present invention provides a technical solution: a mine safety risk early warning method, comprising the following steps:

[0090] S1: Based on the real-time collected monitoring data, screen the abnormal behavior data associated with tunnel stability monitoring, use the PrefixSpan algorithm to update the behavior pattern library, identify the sequence features that match the historical accident data, and generate sequence behavior warning indicators;

[0091] S2: Based on the sequence behavior warning indicators, the historical harmful gas detection data is analyzed in time series, the update frequency and environmental change data are correlated, and the risk level of environmental changes is evaluated to generate the environmental risk level;

[0092] S3: Based on the environmental risk level, extract the abnormal patterns in the miner positioning data stream, use the Apriori algorithm to analyze the abnormal patterns in real time, determine the risk area, adjust the monitoring parameters and optimize data collection according to the analysis results, and generate security monitoring optimization results;

[0093] S4: Based on the security monitoring optimization results, threshold analysis is performed on real-time data to determine the risk level, and corresponding warning signals are issued using CEP technology to generate real-time warning response signals.

[0094] Sequence behavior warning indicators include behavior anomaly index, sequence consistency score, and warning sensitivity. Environmental risk levels include risk alert level, environmental change index, and risk threshold. Security monitoring optimization results include monitoring stability, data response speed, and monitoring accuracy. Real-time warning response signals include warning start time, response type, and signal effective distance.

[0095] See also Figure 2 ,The specific steps for screening abnormal behavior data are:

[0096] S111: Based on the real-time collected monitoring data, the original monitoring data associated with the tunnel stability monitoring is collected, and for each data point, the deviation from the mean of the historical data is calculated using the formula:

[0097]

[0098] Get a list of deviation values, where x i represents a single monitoring data point, μ represents the mean of historical data, Represents data point x i Deviation value of

[0099] S112: Based on the deviation value list, threshold screening is performed to identify abnormal data points using the formula:

[0100]

[0101] Get the abnormal data point set SD, where For data point x i The deviation value, k is the adjustment coefficient, and σ is the standard deviation of the deviation value;

[0102] S113: Analyze the abnormal data point set using the formula:

[0103]

[0104] Get the abnormal behavior data SE, where d represents the data point in the abnormal data point set, and w d is the weight coefficient, and SD is the set of abnormal data points.

[0105] Monitoring data point x collected at the target time point i is 100, and the mean μ of historical data is calculated to be 95;

[0106] The calculation process is:

[0107]

[0108] result Indicates the deviation of the current data point from the historical average.

[0109] Using the above deviation

[0110] The standard deviation of historical deviations, σ, is 3, and k = 2.0;

[0111] Calculated as:

[0112] k·σ=2.0×3=6

[0113] Since |Δx i |=5 is less than 6, and the data point will not be identified as an anomaly.

[0114] The identified abnormal data point set SD includes two points, whose deviation values ​​are d1=7 and d2=-8 respectively; the corresponding weights are and

[0115] SE=1.5×|7|+2.0×|―8|=1.5×7+2.0×8=10.5+16=26.5

[0116] The result SE = 26.5 represents the weighted sum of the outlier data.

[0117] See also Figure 3 , the specific steps for obtaining the sequence behavior warning indicator are:

[0118] S121: Use the PrefixSpan algorithm to process the behavior pattern library and identify frequent sequence patterns using the formula:

[0119] QS2=PrefixSpan(M Q ,Qβ min )

[0120] Get the frequent sequence pattern set QS2, where M Q is the behavior pattern library, Qβ min is the minimum support threshold;

[0121] S122: Based on the frequent sequence pattern set, match with the historical accident data, using the formula:

[0122] QF={qs∈QS2|match(qs,H Q )×p Q >α Q}

[0123] Get the matching sequence feature set QF, where qs is the sequence pattern, H Q is the historical accident dataset, p Q is the probability of occurrence, α Q is the matching threshold, QS2 is the set of frequent sequence patterns, match(qs,HQ ) represents the sequence pattern qs and the historical accident data set H Q The degree of match between

[0124] S123: Risk score the matched sequence feature set using the formula:

[0125]

[0126] Get the sequence behavior warning indicator QR, where f Q is the behavior pattern in QF, w Qf is the weight, risk(f Q ) is the potential risk score of the behavior pattern, c Q is the constant risk adjustment term.

[0127] Detected behavior pattern library M Q contains 100 behavior patterns, and sets the minimum support threshold Qβ min =0.05;

[0128] Applying PrefixSpan algorithm to analyze M Q , identify behavioral patterns that are repeated at least 5 times;

[0129] 15 patterns that met the conditions were captured;

[0130] QS2 includes these 15 frequent behavior patterns, indicating that the pattern appears in at least 5% of the data.

[0131] Assume that QS2 contains a behavior pattern qs in the historical accident data set H Q The medium match is 0.12, p Q =0.08;

[0132] α Q Set to 0.1 or 10%;

[0133] match(qs,H Q )×p Q =0.12×0.08=0.0096

[0134] Since 0.0096 is less than 0.1, qs is not considered to be relevant to the historical accident data and is not included in QF.

[0135] Assume there are two behavior modes f in QF Q1 and f Q2 , the weights are w Qf1 =1.2, w Qf2 =0.8, the potential risk scores are:

[0136] risk(f Q1)=0.3

[0137] risk(f Q2 )=0.5

[0138] QR=(1.2×0.3+0.01)+(0.8×0.5+0.01)=0.36+0.01+0.4+0.01=0.78

[0139] QR = 0.78 represents the comprehensive risk score of this set of behavior patterns, which is used to decide whether preventive measures need to be taken.

[0140] See also Figure 4 , the specific steps for time series analysis are:

[0141] S211: Based on the sequence behavior warning indicators, collect and organize the historical harmful gas detection data, filter out the noise and remove the abnormal values ​​of the data, and obtain the cleaned data;

[0142] S212: Based on the cleaned data, perform time series analysis to detect trends and cyclical changes in the data, apply the ARMA model, and use the formula:

[0143] X trend =ARMA(Z clean ,p r ,q r )

[0144] Get time series trend data X trend , where p r is the autoregressive order, q r is the moving average order, Z clean It is the cleaned data.

[0145] The collected gas concentration values ​​are: [20ppm, 30ppm, 200ppm, 25ppm, 22ppm], with 200ppm being an abnormally high value;

[0146] Apply median filtering and calculate the median of each window:

[0147] For the first window [20, 30, 200], the median is 30ppm;

[0148] For the second window [30, 200, 25], the median is 30ppm;

[0149] For the third window [200, 25, 22], the median is 25ppm.

[0150] The output cleaned data sequence is [30ppm, 30ppm, 25ppm], which excludes outliers and retains more stable collection results.

[0151] Using the above [30ppm, 30ppm, 25ppm], let p r =0.5 and q r =0.5;

[0152] Calculation process:

[0153] X trend,1 =0.5×30+0.5×error1

[0154] X trend,2 =0.5×30+0.5×error2

[0155] X trend,3 =0.5×25+0.5×error3

[0156] Assuming the error term is small, the trend series is [30, 30, 25].

[0157] See also Figure 5 , the specific steps for obtaining the environmental risk level are:

[0158] S221: Based on the results of time series analysis, combined with real-time monitoring of environmental change data, use the formula:

[0159] I env =ω1X trend +ω2EQ t

[0160] Get the environmental index I env , where X trend is time series trend data, EQ t is the real-time environmental change data, ω1 and ω2 are weight factors;

[0161] S222: Based on the environmental index, assess the risk level of environmental changes using the following judgment formula:

[0162]

[0163] Get the environmental risk level RQ, where I env is the environmental index, θ high and θ low are high and low risk thresholds.

[0164] X trend The average value in the most recent week of observations was 50 ppm;

[0165] EQ t Monitoring in place 25℃;

[0166] Assume ω1=0.6 and ω2=0.4.

[0167] Calculation process:

[0168] Calculates a composite environmental index based on the provided weighting factors and data points:

[0169] I env =0.6×50+0.4×25=30+10=40

[0170] Results I env =40 indicates a comprehensive environmental quality assessment that takes into account gas trends and temperature factors.

[0171] Above I env The value of is 40, let θ high is 45, θ low is 30, then 40 is between 45 and 30, so the RQ is medium.

[0172] See also Figure 6 ,The steps for real-time analysis of abnormal patterns are as follows:

[0173] S311: Based on the environmental risk level, collect the miners’ original location data and process the data using the formula:

[0174]

[0175] Get standardized data XD clean (t), where XD(t) is the original position data at time point t, is the mean of the data set, s XD is the standard deviation of the data set;

[0176] S312: Based on the standardized data, abnormal pattern recognition is performed using the LOF algorithm using the formula:

[0177] M abnormal (t) = LOF(XD clean (t),k v )>1.5

[0178] Get the abnormal data point mark M abnormal (t), where XD clean (t) is the standardized data, k v is the number of neighbors used in the LOF algorithm;

[0179] S313: Based on the abnormal data point markers, association rule mining is used to identify the relationship between abnormal patterns, using the formula:

[0180]

[0181] Get the support degree of the association rule Support(X A ), where X A is the item set of abnormal patterns, N is the total number of transactions, M abnormal (t) is the abnormal data point mark, Indicates that the calculated item set X appears as a subset in M abnormal The number of times in (t).

[0182] Data is continuously collected through sensors in a mine. For example, at time points t=1 to t=10, the collected data is [10, 12, 15, 9, 14, 12, 13, 11, 16, 9];

[0183] Calculate the average

[0184]

[0185] Calculate the standard deviation s XD :

[0186]

[0187] use and XD Calculate XD clean (t):

[0188]

[0189] XD clean The value represents the number of standard deviations from the mean position and is used to identify data points that deviate from the norm.

[0190] Let k v =2, calculate the LOF value of t=1 considering its two nearest neighbor points t=2 and t=3;

[0191] The LOF algorithm is used for calculation. If the LOF value is greater than 1.5, t is considered to be abnormal, indicating that the density of t is significantly lower than that of its neighbors.

[0192] Assume M abnormal The time points [1, 5, 9] are abnormal;

[0193] If X A is {1,9}, then:

[0194]

[0195] A support of 0.2 indicates that among all transactions, 20% contain the abnormal pattern X A .

[0196] See also Figure 7,The specific steps for obtaining security monitoring optimization results are:

[0197] S321: Based on the analysis results, adjust the sensitivity of the monitoring process and optimize the response to potential risks using the formula:

[0198] KP new =KP old +α K ×(KP target ―KP old )

[0199] Get the new monitoring parameter value KP new , where KP target is the target parameter value, KP old is the currently monitored parameter value, α K is the smoothing factor;

[0200] S322: Based on the new monitoring parameter values, reconfigure the data collection frequency and optimize resource allocation and data quality, using the formula:

[0201] KC new =KC old ×(1+γ K ×IFactor)

[0202] Get the new data collection frequency KC new , among which, KC old is the original data collection frequency, IFactor is the impact factor, γ K It is the adjustment coefficient, which is used to control the variation of the acquisition frequency.

[0203] The following data is collected:

[0204] KP old is 50;

[0205] KP target is 60;

[0206] α K Set to 0.5.

[0207] Calculation process:

[0208] KP new =50+0.5×(60-50)=55

[0209] The new monitoring parameter value is 55, which helps to monitor environmental changes more sensitively.

[0210] The following data was collected:

[0211] KC old is 10Hz;

[0212] IFactor is set to 0.1;

[0213] γ K Set to 1.

[0214] Calculation process:

[0215] KC new =10×(1+1×0.1)=11

[0216] The new data collection frequency of 11Hz will increase the frequency of data collection to more effectively monitor potential high-risk areas.

[0217] See also Figure 8 ,The specific steps for obtaining the real-time warning response signal are:

[0218] S411: Based on the security monitoring optimization results, compare with the real-time data, calculate the deviation between the real-time data and the threshold, and use the formula:

[0219]

[0220] Get the initial warning sensitivity Gs, where G d Indicates the current real-time monitoring data, G u is the threshold value;

[0221] S412: Based on the initial warning sensitivity and combined with the abnormal pattern recognition results, the formula is used:

[0222]

[0223] Get the weighted value G of the warning signal w , where Gs is the initial warning sensitivity, p C is the probability of occurrence of abnormal patterns, α C is the adjustment coefficient;

[0224] S413: Based on the weighted value of the warning signal and the current risk level, the formula is:

[0225] GR=min(G w +r Z ,1)

[0226] Get the real-time warning response signal GR, where G w is the weighted value of the early warning signal, r Z is the current risk level.

[0227] Set the threshold G u is 100, real-time monitoring data G d is 90.

[0228] Calculation process:

[0229] First calculate G d and G u The absolute value of the difference:

[0230] |G d ―G u |=|90―100|=10

[0231] Calculate the ratio:

[0232]

[0233] Calculate Gs:

[0234] Gs=max(0,1―0.1)=0.9

[0235] Gs=0.9 means that the sensitivity between the monitoring data and the threshold is 0.9, which is a relatively high sensitivity value. This means that the current real-time data is close to the threshold and needs attention.

[0236] α C is 0.5;

[0237] Using the above Gs = 0.9, the probability of occurrence p is detected C =0.2.

[0238] Calculation process:

[0239] calculate Values:

[0240] e ―0.5·0.2 =e ―0.1 ≈0.905

[0241] Calculate G w :

[0242] G w =0.9·(1-0.905)=0.9·0.095=0.0855

[0243] G w =0.0855 represents the weighted warning signal value, the lower G w A value of 0 indicates that the current risk level is low.

[0244] The following data is collected:

[0245] r Z is 0.4;

[0246] Combined with G w =0.0855.

[0247] Calculation process:

[0248] Calculate Gw +r Z :

[0249] 0.0855+0.4=0.4855

[0250] Since 0.4855<1, GR=0.4855;

[0251] The early warning response signal GR=0.4855 indicates that the current risk level is low and the system status is stable.

[0252] A mine safety risk early warning system, the system comprising:

[0253] The behavior pattern analysis module screens data anomalies associated with tunnel stability based on real-time collected monitoring data, updates the behavior pattern library using the PrefixSpan algorithm, identifies sequence features that match historical accident data, and generates sequence behavior warning indicators;

[0254] The environmental risk analysis module analyzes historical harmful gas detection data based on sequence behavior warning indicators, correlates and analyzes update frequency and environmental change data, evaluates the risk level of environmental changes, and generates environmental risk levels;

[0255] The abnormal pattern recognition module extracts abnormal patterns from the miner positioning data stream based on the environmental risk level, uses the Apriori algorithm for real-time analysis, and generates preliminary risk area identification results;

[0256] The monitoring parameter optimization module determines the risk area based on the preliminary risk area identification results, adjusts the monitoring parameters and optimizes data collection to generate security monitoring optimization results;

[0257] The real-time warning transmission module performs threshold analysis on real-time data and makes risk level judgments based on the security monitoring optimization results, and uses CEP technology to send out warning signals and generate real-time warning response signals.

[0258] The above are only preferred embodiments of the present invention and are not intended to limit the present invention in other forms. Any technician familiar with the profession may use the technical contents disclosed above to change or modify them into equivalent embodiments with equivalent changes and apply them to other fields. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. A mine safety risk early warning method, characterized in that: The mine safety risk early warning method specifically comprises the following steps: Based on the real-time collected monitoring data, the abnormal behavior data associated with tunnel stability monitoring is screened, the PrefixSpan algorithm is used to update the behavior pattern library, the sequence features that match the historical accident data are identified, and the sequence behavior warning indicators are generated; Based on the sequence behavior warning indicators, a time series analysis is performed on the historical harmful gas detection data, the update frequency and environmental change data are correlated and analyzed, and the risk level of environmental changes is evaluated to generate an environmental risk level; Based on the environmental risk level, the abnormal patterns in the miner positioning data stream are extracted, and the Apriori algorithm is used to analyze the abnormal patterns in real time to determine the risk area. The monitoring parameters are adjusted and the data collection is optimized according to the analysis results to generate the security monitoring optimization results; Based on the security monitoring optimization results, threshold analysis is performed on real-time data to determine the risk level, and corresponding warning signals are issued using CEP technology to generate real-time warning response signals.

2. The mine safety risk early warning method according to claim 1 is characterized in that: The steps of screening abnormal behavior data are specifically as follows: Based on the real-time collected monitoring data, the original monitoring data associated with tunnel stability monitoring is collected, and for each data point, the deviation from the mean of historical data is calculated using the formula: In the formula, x i is a single monitoring data point, μ represents the mean of historical data, For data point x i Deviation value of Based on the deviation value list, threshold screening is performed to identify abnormal data points using the formula: Get the abnormal data point set SD, where For data point x i The deviation value, k is the adjustment coefficient, and σ is the standard deviation of the deviation value; The abnormal data point set is analyzed using the formula: Get the abnormal behavior data SE, where d is the data point in the abnormal data point set, w d is the weight coefficient, and SD is the set of abnormal data points.

3. The mine safety risk early warning method according to claim 1, characterized in that: The steps for obtaining the sequence behavior warning indicator are specifically as follows: The PrefixSpan algorithm is used to process the behavior pattern library and identify frequent sequence patterns. The formula QS2=PrefixSpan(M Q ,Qβ min ) Get the frequent sequence pattern set QS2, where M Q is the behavior pattern library, Qβ min is the minimum support threshold; Based on the frequent sequence pattern set, matching is performed with historical accident data using the formula: QF={qs∈QS2|match(qs,H Q )×p Q >α Q } Get the matching sequence feature set QF, where qs is the sequence pattern, H Q is the historical accident dataset, p Q is the probability of occurrence, α Q is the matching threshold, QS2 is the set of frequent sequence patterns, match(qs,H Q ) represents the sequence pattern qs and the historical accident data set H Q The degree of match between The risk score of the matched sequence feature set is calculated using the formula: Get the sequence behavior warning indicator QR, where T Q is the behavior pattern in QF, w Qf is the weight, risk(f Q ) is the potential risk score of the behavior pattern, c Q is the constant risk adjustment term.

4. The mine safety risk early warning method according to claim 1, characterized in that: The steps of performing time series analysis are specifically as follows: Based on the sequence behavior warning indicators, historical harmful gas detection data is collected and sorted, and noise and outliers are filtered out of the data to obtain cleaned data; Based on the cleaned data, time series analysis is performed to detect trends and cyclical changes in the data, and the ARMA model is applied using the formula: X trend =ARMA(Z clean ,p r ,q r ) Get time series trend data X trend , where p r is the autoregressive order, q r is the moving average order, Z clean It is the cleaned data.

5. The mine safety risk early warning method according to claim 1, characterized in that: The steps for obtaining the environmental risk level are specifically as follows: Based on the time series analysis results, combined with the real-time monitoring of environmental change data, the formula is used: I env =ω1X trend +ω2EQ t Get the environmental index I env , where X trend is time series trend data, EQ t is the real-time environmental change data, ω1 and ω2 are weight factors; Based on the environmental index, the risk level of environmental changes is assessed using the following judgment formula: Get the environmental risk level RQ, where I env is the environmental index, θ high and θ low are high and low risk thresholds.

6. The mine safety risk early warning method according to claim 1, characterized in that: The steps of performing real-time analysis on abnormal patterns are specifically as follows: Based on the environmental risk level, the original location data of the miners is collected and processed using the formula: Get standardized data XD clean (t), where XD(t) is the original position data at time point t, is the mean of the data set, s XD is the standard deviation of the data set; Based on the standardized data, abnormal pattern recognition is performed using the LOF algorithm using the formula: M abnormal (t)=LOF(XD clean (t),k v )>1.5, Get the abnormal data point mark M abnormal (t), where XD clean (t) is the standardized data, k v is the number of neighbors used in the LOF algorithm; Based on the abnormal data point markers, association rule mining is used to identify the relationship between abnormal patterns, using the formula: Get the support degree of the association rule Support(X A ), where X A is the item set of abnormal patterns, N is the total number of transactions, M abnormal (t) is the abnormal data point mark.

7. The mine safety risk early warning method according to claim 1, characterized in that: The steps for obtaining the security monitoring optimization result are specifically as follows: Based on the analysis results, the sensitivity of the monitoring process is adjusted to optimize the response to potential risks using the formula: KP new =KP old +α K ×(KP target -KP old ), Get the new monitoring parameter value KP new , where KP tarhet is the target parameter value, KP old is the currently monitored parameter value, α K is the smoothing factor; Based on the new monitoring parameter values, the data collection frequency is reconfigured, and resource allocation and data quality are optimized, using the formula: KC new =KC old ×(1+γ K ×IFactor) Get the new data collection frequency KC new , among which, KC old is the original data collection frequency, IFactor is the impact factor, γ K It is the adjustment coefficient, which is used to control the variation of the acquisition frequency.

8. The mine safety risk early warning method according to claim 1, characterized in that: The steps of obtaining the real-time early warning response signal are specifically as follows: Based on the security monitoring optimization results, the real-time data is compared with the real-time data, and the deviation between the real-time data and the threshold is calculated using the formula: Get the initial warning sensitivity Gs, where G d Indicates the current real-time monitoring data, G u is the threshold value; Based on the preliminary warning sensitivity and combined with the abnormal pattern recognition results, the formula is adopted: Get the weighted value G of the warning signal w , where Gs is the initial warning sensitivity, p C is the probability of occurrence of abnormal patterns, α C is the adjustment coefficient; Based on the weighted value of the warning signal and the current risk level, the formula is adopted: GR=min(G w +r Z ,1), Get the real-time warning response signal GR, where G w is the weighted value of the early warning signal, r Z is the current risk level.

9. A mine safety risk early warning system, characterized in that: According to the mine safety risk early warning method according to any one of claims 1 to 8, the system comprises: The behavior pattern analysis module screens data anomalies associated with tunnel stability based on real-time collected monitoring data, updates the behavior pattern library using the PrefixSpan algorithm, identifies sequence features that match historical accident data, and generates sequence behavior warning indicators; The environmental risk analysis module analyzes historical harmful gas detection data based on the sequence behavior warning indicators, associates and analyzes the update frequency with environmental change data, evaluates the risk level of environmental changes, and generates an environmental risk level; The abnormal pattern recognition module extracts abnormal patterns from the miner positioning data stream based on the environmental risk level, uses the Apriori algorithm for real-time analysis, and generates preliminary risk area recognition results; The monitoring parameter optimization module determines the risk area based on the preliminary risk area identification result, adjusts the monitoring parameters and optimizes the data collection, and generates the security monitoring optimization result; The real-time warning transmission module performs threshold analysis on the real-time data based on the security monitoring optimization result, makes a risk level judgment, and uses CEP technology to send out a warning signal to generate a real-time warning response signal.

Images