CN114087021B - Rock burst multi-parameter dynamic trend early warning method - Google Patents

Abstract

The invention provides a multi-parameter dynamic trend early warning method of rock burst, belonging to the technical field of underground excavation engineering and coal rock dynamic disaster early warning. The method includes: according to the real-time monitoring data of the on-site microseismic monitoring system, using the Mann-Kendall trend test method to describe the change trend of the "time-space-strong" multivariate precursory early warning indicators in the process of rockburst gestation and evolution. According to the pre-warning rule of the rockburst precursor, the early warning is carried out; the early warning efficiency of each index is evaluated; the index is optimized based on the principle of maximizing early warning efficiency; the early warning efficiency corresponding to the optimized index is used as the weight to carry out multi-index fusion to obtain the comprehensive shock risk comprehensive abnormal index; The comprehensive anomaly index of impact risk is compared with the corresponding quantitative grading standard to determine the impact risk level; index selection and weight update are carried out regularly driven by the actual monitoring data on site. The invention can provide efficient and accurate decision-making basis for prevention and control of underground rock burst.

Description

A multi-parameter dynamic trend early warning method for rock burst

technical field

The invention relates to the technical field of underground excavation engineering and coal and rock dynamic disaster early warning, in particular to a multi-parameter dynamic trend early warning method of rock burst.

Background technique

Rockburst is one of the main dynamic disasters affecting underground production in coal mines, often causing serious consequences such as casualties and major property losses. In recent years, with the depletion of shallow mineral resources, coal mining continues to expand deep into the earth, the coal seam structure of the stope and the occurrence conditions of surrounding rocks around the roadway have become increasingly complex, and the internal dynamic response characteristics of the coal and rock mass have become more intense, resulting in shock The frequency of ground pressure accidents is rising rapidly, and accurate and efficient monitoring and early warning methods for this disaster are the key to ensuring safe underground production.

At present, the commonly used online monitoring methods in underground mines include microseismic, geosound, electromagnetic radiation, ground stress, etc. Among them, the microseismic monitoring system can monitor the fracture of coal and rock mass in a large area of the mine, and is widely used in the monitoring and early warning of rock burst. , the relevant personnel can achieve the purpose of early warning by analyzing the change trend of the energy, frequency and other indicators of microseismic events. However, the judgment of the change trend of the early warning indicators of shock hazard precursors depends on artificial experience, which is inefficient and has a low degree of reliability. At the same time, due to the different dimensions reflected by each early warning indicator on the gestation and evolution process of shock hazard, conflicting results may occur. Cause the relevant personnel to misjudge the actual dangerous state. Therefore, it is necessary to propose an early-warning method that can automatically and efficiently determine the time-series change trend of the early warning indicators of shock hazard precursors and can provide a comprehensive, single, and quantitative shock hazard level.

SUMMARY OF THE INVENTION

The embodiment of the present invention provides a multi-parameter dynamic trend early warning method for rock burst, which can perform efficient and accurate monitoring and early warning on rock burst. The technical solution is as follows:

The embodiment of the present invention provides a multi-parameter dynamic trend early warning method for rock burst, including:

S101, according to the real-time monitoring data of the on-site microseismic monitoring system, the Mann-Kendall trend test method is used to describe the change trend of the "time-space-strong" multivariate precursory early warning indicators in the process of rockburst gestation and evolution, and determine whether the change trend of each index conforms to The characterization law of rockburst precursors serves as the early warning criterion for early warning;

S102, using a confusion matrix to evaluate the early warning effectiveness of each indicator;

S103, perform index selection based on the principle of maximizing early warning efficiency;

S104, using the early warning efficiency corresponding to the preferred index as a weight to perform multi-index fusion to obtain a comprehensive anomaly index of impact risk;

S105, compare the comprehensive anomaly index of impact risk with the corresponding quantitative grading standard to determine the impact risk level;

S106, regularly perform index selection and weight update driven by the actual monitoring data on site.

Further, according to the real-time monitoring data of the on-site microseismic monitoring system, the Mann-Kendall trend test method is used to describe the change trend of the "time-space-strong" multivariate precursory early warning indicators in the process of rockburst gestation and evolution. Whether the trend conforms to the rockburst precursory characterization law is used as an early warning criterion for early warning, including:

Collect the real-time monitoring data of the on-site microseismic monitoring system, and preprocess the original monitoring data with a specific time window and slip step length to obtain the time series of the multivariate precursory early warning indicators of rockburst hazard "time-space-strong"; The "time-space-strong" multivariate precursory early warning indicators of rock burst danger include:

Daily total frequency, frequency deviation value, average total frequency, earthquake absence, A(b) value, microseismic activity scale, algorithm complexity, P(b) value and time information entropy reflecting the time dimension;

Microseismic activity and source concentration that reflect spatial dimensions;

The daily maximum energy, daily total energy, daily average energy, energy deviation value, average total energy, total fault area and b value reflecting the intensity dimension;

Among them, daily maximum energy, daily total energy, daily total frequency, daily average energy, energy deviation value, frequency deviation value, average total energy, microseismic activity, seismic absence, A(b) value, total fault area, average total frequency , microseismic activity scale, and algorithm complexity belong to positive early warning indicators, that is, the higher the value, the greater the shock risk; the degree of focus concentration, b value, P(b) value, and time information entropy belong to negative early warning indicators, that is, The lower the value, the greater the impact risk;

The Mann-Kendall trend test method is used to determine the real-time change trend of each early-warning index, and the early-warning criterion is based on whether the change trend of each index conforms to the law of rockburst precursor characterization.

Further, the preprocessing of the original monitoring data with a specific time window and slip step size refers to converting the original monitoring data from an irregular time series to a regular time series, including:

Define a sliding time window with a length of Δt, and divide the collected monitoring data time series into n data sets with a length of Δt, which correspond to the time at the end of the time window, that is, the data set at time T _i is denoted as X _i [x ₁ ,x ₂ ,x ₃ ,...,x _k ], k≤t, 0<i≤n, where Δt is the length of the sliding time window, and T _i is the corresponding end of the i-th time window time;

Calculate the early warning index values _yi corresponding to all samples in X _i , and arrange the early warning index values _yi in order to obtain the converted regular time series, denoted as Y[y ₁ , y ₂ , y ₃ ,..., y _n ].

Further, the described utilization of Mann-Kendall trend test method to determine the real-time variation trend of each early warning index includes:

Divide the early warning indicator rule time series Y[y ₁ , y ₂ , y ₃ ,..., y _n ] into m data sets with a time window length of △a corresponding to the time A _i at the end of the time window, namely A The data set at time _i is _Yi [y ₁ , y ₂ , y ₃ ,..., y _q ], 10≤q≤a, 0< _i≤m , calculate the test statistic S of Yi:

q表示Y_i[y₁,y₂,y₃,...,y_q]数据集的长度，y_p表示第p个数据，p＝1,2,3...q-1，y_j表示第j个数据，j＝p+1,2,3...q；where sgn( ) represents the symbolic function,

q represents the length of the data set Yi [y ₁ , y ₂ , y ₃ ,..., y _q ], y _p represents the p- _th data, p=1, 2, 3...q-1, y _j Represents the jth data, j=p+1,2,3...q;

According to the _obtained test statistic S of Yi, determine the test standard _Z of Yi:

When Z>0, the early warning index at time T _i has an increasing trend; when Z < 0, the early warning index at time T _i has a decreasing trend; when Z=0, the early warning index at time T _i does not have an obvious trend of change.

Further, in this embodiment, the pre-warning based on whether the change trend of each index conforms to the pre-warning rule of rockburst precursors as the pre-warning criterion includes:

The real-time change trend of each index is compared with the characterization law of rockburst precursors. If the positive early warning index has an increasing trend, an early warning will be issued, if the negative early warning index has a decreasing trend, an early warning will be issued, and other changing trends will not be issued an early warning.

Further, the early warning efficiency is expressed as:

表示预警为有冲击危险并且实际发生了冲击地压事件，

表示预警为无冲击危险但实际发生了冲击地压事件；Precision表示精确率，

其中

表示预警为有冲击危险并且实际发生了冲击地压事件，

表示预警为有冲击危险但实际未发生冲击地压事件。Among them, EFF means early warning efficiency; Recall means recall rate,

Indicates that the warning is in danger of shock and a shock event has actually occurred,

Indicates that the early warning is no risk of shock but a shock event has actually occurred; Precision indicates the accuracy rate,

in

Indicates that the warning is in danger of shock and a shock event has actually occurred,

Indicates that the warning is a risk of shock but no shock event has actually occurred.

Further, the index optimization based on the principle of maximizing early warning efficiency includes:

The early warning efficacy values of all early warning indicators are arranged in descending order, and the top n% of the indicators are screened for the next step of data fusion, where n is a positive integer and the number of indicators is rounded up.

Further, the multi-index fusion obtained by using the early warning efficiency corresponding to the preferred index as a weight to obtain a comprehensive anomaly index of impact risk includes:

The early warning efficiency corresponding to the preferred index is input into the comprehensive rockburst early warning model as a weight, and the rockburst comprehensive early warning model uses the comprehensive abnormal index method to perform multi-index fusion to obtain the comprehensive shock risk comprehensive abnormal index:

Among them, Q is the comprehensive anomaly index of shock risk; e is the natural base; n is the total number of preferred indicators; EFF _k is the early warning efficiency corresponding to the k-th index; W _k(+/-) means the k-th positive/negative direction The abnormal membership of early warning indicators,

For positive early warning indicators, the value of abnormal membership is:

For negative early warning indicators, the value of abnormal membership is:

Further, the shock hazard level includes: no shock hazard state, weak shock hazard state, medium shock hazard state and strong shock hazard state.

Further, the periodic optimization of indicators and updating of weights driven by on-site actual monitoring data includes:

Periodically select indicators driven by the actual monitoring data on site, and input the early warning performance corresponding to the optimal indicators as weights into the comprehensive rockburst early warning model, so as to realize self-feedback update of the comprehensive rockburst early warning model.

The beneficial effects brought by the technical solutions provided by the embodiments of the present invention include at least:

1) It can automatically determine the real-time change trend of early warning indicators, and use the change trend of early warning indicators to carry out rockburst early warning;

2) Using the comprehensive anomaly index of shock hazard to quantitatively describe the hazard of rock burst, avoiding the phenomenon of high false alarm/missing rate caused by a single early warning indicator and conflict of early warning levels caused by different early warning indicators;

3) Regular self-feedback update driven by on-site actual monitoring data, has strong scalability and adaptability, adapts to complex and changeable working conditions downhole, and is helpful for efficient and accurate monitoring and early warning of rock burst , and provide efficient and accurate decision-making basis for the prevention and control of downhole rock burst.

Description of drawings

In order to illustrate the technical solutions in the embodiments of the present invention more clearly, the following briefly introduces the accompanying drawings used in the description of the embodiments. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can also be obtained from these drawings without creative effort.

1 is a schematic flowchart of a multi-parameter dynamic trend early warning method for rock burst provided by an embodiment of the present invention;

2 is a detailed schematic flowchart of a multi-parameter dynamic trend early warning method for rock burst provided by an embodiment of the present invention;

Fig. 3 is the schematic diagram of the process of preprocessing the original monitoring data in the present embodiment;

Fig. 4 is the time series change curve of the daily total frequency F _sum early warning index in the present embodiment;

Fig. 5 is the time series change curve of frequency deviation value _DF early warning index in the present embodiment;

预警指标的时序变化曲线；Fig. 6 is the average total frequency in this embodiment

Time series change curve of early warning indicators;

Fig. 7 is the time series change curve of the early-warning index of lack of earthquake M _m in the present embodiment;

Fig. 8 is the time series change curve of A (b) value early warning index in the present embodiment;

Fig. 9 is the time series variation curve of the microseismic activity scale F early warning index in the present embodiment;

Fig. 10 is the time series change curve of the algorithm complexity AC early warning index in the present embodiment;

Fig. 11 is the time series change curve of the early warning index of P(b) value in the present embodiment;

Fig. 12 is the time series change curve of the time information entropy Q _t early warning index in the present embodiment;

Fig. 13 is the time series variation curve of the microseismic activity _SD early warning index in the present embodiment;

Fig. 14 is the time series variation curve of the epicenter concentration degree λ early warning index in the present embodiment;

Fig. 15 is the time series change curve of the daily maximum energy E _max early warning index in the present embodiment;

Fig. 16 is the time series change curve of the daily total energy E _sum early warning index in the present embodiment;

Fig. 17 is the time series change curve of the daily average energy E _avg early warning index in the present embodiment;

Fig. 18 is the time series change curve of the energy deviation value _DE early warning index in the present embodiment;

预警指标的时序变化曲线；Figure 19 shows the average total energy in this example

Time series change curve of early warning indicators;

Fig. 20 is the time series change curve of the early warning index of the total area A(t) of the fault layer in the present embodiment;

Fig. 21 is the time series change curve of the b value early warning index in the present embodiment;

FIG. 22 is a time-series change curve of the comprehensive anomaly index Q of impact risk in this embodiment.

Detailed ways

In order to make the objectives, technical solutions and advantages of the present invention clearer, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings.

As shown in FIG. 1 , an embodiment of the present invention provides a multi-parameter dynamic trend early warning method for rock burst, including:

S101, according to the real-time monitoring data of the on-site microseismic monitoring system, use the Mann-Kendall trend test method to describe the variation trend of the "time-space-strong" multivariate precursory early warning indicators during the gestation and evolution of rockburst, with Whether the change trend of each index conforms to the characteristic law of rockburst precursors is used as an early warning criterion for early warning; specifically, the following steps may be included:

A1. Collect the real-time monitoring data of the on-site microseismic monitoring system, and preprocess the original monitoring data with a specific time window and slip step to obtain the time series of multiple precursor early warning indicators of rockburst hazard “time-space-strong”; , the "time-space-strong" multivariate precursory early warning indicators of rock burst danger include but are not limited to:

缺震M_m、A(b)值、微震活动标度F、算法复杂度AC、P(b)值和时间信息熵Q_t；其中，A(b)值和P(b)值为地震学基本定律之一的“G-R”关系式(古登堡-里克特方程)中经验常数b的衍生指标；The daily total frequency F _sum reflecting the time dimension, the frequency deviation value D _F , and the average total frequency

Seismic absence M _m , A(b) value, microseismic activity scale F, algorithm complexity AC, P(b) value and time information entropy Q _t ; where A(b) value and P(b) value for seismology A derivative index of the empirical constant b in the "GR" relation (the Gutenberg-Richter equation), one of the fundamental laws;

Among them, the value of the total daily frequency F _sum is calculated by the following formula:

Among them, n represents the total frequency of microseismic events in the previous 24 hours (hours);

The value of the frequency deviation value _DF is calculated by the following formula:

表示前30d(天)内微震事件平均日频次；Among them, F _t represents the frequency of microseismic events in the first 24 hours,

Indicates the average daily frequency of microseismic events within the previous 30 days (days);

的取值由下式计算：Average total frequency

The value of is calculated by the following formula:

Among them, T represents the length of the time window, and F _(t) represents the frequency of microseisms at time t;

The value of the lack of earthquake M _m is calculated by the following formula:

Among them, m is the total number of energy level bins; lgE _i is the ith energy level; _Ni is the actual microseismic number of the ith energy level; the value of A(b) is calculated by the following formula:

Among them, N is the total number of microseisms, and _Mi is the energy level of microseismic events;

The value of the microseismic activity scale F is calculated by the following formula:

F ₀ =10 ^6.11+1.09M

Among them, T represents the length of the time window, and _Mi is the energy level of the microseismic event;

The value of the algorithmic complexity AC is calculated by the following formula:

Among them, n represents the number of changes in the microseismic energy level within the time window, _Mmax represents the maximum microseismic energy level, and M _min represents the minimum microseismic energy level;

The value of P(b) is calculated by the following formula:

Among them, N is the total number of microseisms, and _Mi is the energy level of microseismic events;

The value of time information entropy Q _t is calculated by the following formula:

Among them, n is the frequency of microseismic events in the time window, and t _i is the occurrence time of the ith mine shock;

Microseismic activity S _D and source concentration λ reflecting the spatial dimension; where,

The value of the microseismic activity S _D is calculated by the following formula:

Among them, N is the total number of microseisms, E _i is the energy of the microseismic event, and M is the maximum energy level of the microseismic;

The value of the source concentration λ is calculated by the following formula:

Among them, λ ₁ , λ ₂ , λ ₃ are the characteristic roots of the covariance matrix composed of the microseismic event coordinates (x, y, z) in the time window;

断层总面积A(t)和b值；其中，The daily maximum energy E _max , the daily total energy E _sum , the daily average energy _{E avg} , the energy deviation value DE , and the average total energy reflecting the intensity dimension

The total fault area A(t) and b values; where,

The value of the daily maximum energy E _max is calculated by the following formula:

E _max =max(E ₁ ,E ₂ ,...,E _n )

Among them, E _i is the ith microseismic event in the first 24 hours;

The value of the total daily energy E _sum is calculated by the following formula:

Among them, n represents the total frequency of microseismic events in the previous 24 hours, and E _i represents the energy of each microseismic event;

The value of the daily average energy E _avg is calculated by the following formula:

Among them, n represents the total frequency of microseismic events in the previous 24 hours, and E _i represents the energy of each microseismic event;

The value of the energy deviation value D _E is calculated by the following formula:

表示前30d内微震事件平均日能量；Among them, E _t represents the total energy of microseismic events in the first 24 hours,

Represents the average daily energy of microseismic events within the first 30 days;

的取值由下式计算：average total energy

The value of is calculated by the following formula:

Among them, T represents the length of the time window, and E _(t) represents the microseismic energy at time t;

The value of the total fault area A(t) is calculated by the following formula:

Among them, _k0 is the lower limit of the energy level of the microseismic in the time window, k is the energy level of each microseismic, and N(k) is the number of microseismic events with energy level k in the time window;

The value of b is calculated by the following formula:

Among them, m is the total number of energy level bins; lgE _i is the ith energy level; _Ni is the actual microseismic number of the ith energy level.

The above early warning indicators can build a precursor early warning indicator library. Among the above early warning indicators, daily maximum energy, daily total energy, daily total frequency, daily average energy, energy deviation value, frequency deviation value, average total energy, microseismic activity, lack of earthquake, A (b) value, total fault area, average total frequency, microseismic activity scale, and algorithm complexity are positive early warning indicators, that is, the higher the value, the greater the shock risk; the degree of focus concentration, b value, P(b) Value and time information entropy are negative early warning indicators, that is, the lower the value, the greater the impact risk.

In this embodiment, the preprocessing of the original monitoring data with a specific time window and slip step size refers to converting the original monitoring data from an irregular time series to a regular time series, including:

Define a sliding time window with a length of Δt, and divide the collected monitoring data time series into n data sets with a length of Δt, which correspond to the time at the end of the time window, that is, the data set at time T _i is denoted as X _i [x ₁ ,x ₂ ,x ₃ ,...,x _k ], k≤t, 0<i≤n, where Δt is the length of the sliding time window, and T _i is the corresponding end of the i-th time window time;

Calculate the early warning index values _yi corresponding to all samples in X _i , and arrange the early warning index values _yi in order to obtain the converted regular time series, denoted as Y[y ₁ , y ₂ , y ₃ ,..., y _n ], as shown in Figure 3.

A2, use the Mann-Kendall trend test method to determine the real-time change trend of each early warning index, and use whether the change trend of each index conforms to the law of rockburst precursor characterization as the early warning criterion for early warning, which may include the following steps:

Divide the early warning indicator rule time series Y[y ₁ , y ₂ , y ₃ ,..., y _n ] into m data sets with a time window length of △a corresponding to the time A _i at the end of the time window, namely A The data set at time _i is _Yi [y ₁ , y ₂ , y ₃ ,..., y _q ], 10≤q≤a, 0< _i≤m , calculate the test statistic S of Yi:

q表示Y_i[y₁,y₂,y₃,...,y_q]数据集的长度，y_p表示第p个数据，p＝1,2,3...q-1，y_j表示第j个数据，j＝p+1,2,3...q；where sgn( ) represents the symbolic function,

q represents the length of the data set Yi [y ₁ , y ₂ , y ₃ ,..., y _q ], y _p represents the p- _th data, p=1, 2, 3...q-1, y _j Represents the jth data, j=p+1,2,3...q;

According to the _obtained test statistic S of Yi, determine the test standard _Z of Yi:

When Z>0, the early warning index at time T _i has an increasing trend; when Z < 0, the early warning index at time T _i has a decreasing trend; when Z=0, the early warning index at time T _i does not have an obvious trend of change. Repeat the above calculation until the change trend of the early warning indicators corresponding to all times is obtained;

The real-time change trend of each index is compared with the characterization law of rockburst precursors. If the positive early warning index has an increasing trend, an early warning will be issued, if the negative early warning index has a decreasing trend, an early warning will be issued, and other changing trends will not be issued an early warning.

S102, using a confusion matrix to evaluate the early warning effectiveness of each indicator;

In this embodiment, the confusion matrix is specifically the following 2×2 matrix:

In this embodiment, the early-warning efficiency of the early-warning index obtained according to the confusion matrix is:

其中

表示预警为有冲击危险并且实际发生了冲击地压事件，

表示预警为无冲击危险但实际发生了冲击地压事件；Precision表示精确率，

其中

表示预警为有冲击危险并且实际发生了冲击地压事件，

表示预警为有冲击危险但实际未发生冲击地压事件。。Among them, EFF means early warning efficiency; Recall means recall rate,

in

Indicates that the warning is in danger of shock and a shock event has actually occurred,

Indicates that the early warning is no risk of shock but a shock event has actually occurred; Precision indicates the accuracy rate,

in

Indicates that the warning is in danger of shock and a shock event has actually occurred,

Indicates that the warning is a risk of shock but no shock event has actually occurred. .

In this embodiment, the trend range of EFF is 0 to 1, and the closer the value is to 1, the better the warning effect is.

S103, perform index selection based on the principle of maximizing early warning efficiency;

In this embodiment, the early warning efficacy values of all early warning indicators are arranged in descending order, and the top n% of the indicators are screened for the next step of data fusion, where n is a positive integer and the number of indicators is rounded up.

S104, using the early warning efficiency corresponding to the preferred index as a weight to perform multi-index fusion to obtain a comprehensive anomaly index of impact risk;

In this embodiment, the early warning efficiency corresponding to the preferred index is input into the comprehensive rockburst early warning model as the weight, and the comprehensive rockburst early warning model uses the comprehensive abnormal index method to perform multi-index fusion to obtain the comprehensive shock risk comprehensive abnormal index:

Among them, Q is the comprehensive anomaly index of shock risk; e is the natural base, which is approximately equal to 2.718; n is the total number of preferred indicators; _{EFF k} is the early warning efficiency corresponding to the k-th index; The abnormal membership degree of the positive/negative early warning index, which ranges from 0 to 1. The specific calculation method is as follows:

For positive early warning indicators, the value of abnormal membership is:

For negative early warning indicators, the value of abnormal membership is:

S105, compare the impact risk comprehensive abnormality index with the corresponding quantitative classification standard to determine the impact risk level.

In this embodiment, the impact hazard quantitative classification standard is specifically:

When 0≤Q≤0.25, there is no impact danger state;

When 0.25<Q≤0.5, it is a weak shock dangerous state;

When 0.5<Q≤0.75, it is in the dangerous state of medium impact;

When 0.75<Q≤1, it is a dangerous state of strong shock.

S106, regularly perform index selection and weight update driven by the actual monitoring data on site.

In this embodiment, index optimization is carried out regularly driven by the actual monitoring data on site, and the early warning performance corresponding to the preferred index is input into the rockburst comprehensive early warning model as a weight, so as to realize self-feedback update of the rockburst comprehensive early warning model.

In this embodiment, for example, at the site every one month or more (or when the working face is replaced, the geological conditions are greatly changed, a large-energy mine earthquake or a ground pressure event occurs, etc.), all existing Steps S101-S104 are repeated for the historical monitoring data, and the adaptive indicators are re-selected from the precursory early warning indicator library, and the corresponding early warning performance is input into the rockburst comprehensive early warning model as a weight, so as to achieve the final self-feedback update of the rockburst comprehensive early warning model. The purpose of improving adaptability.

To sum up, the multi-parameter dynamic trend early warning method for rock burst described in the embodiment of the present invention has at least the following beneficial effects:

1) It can automatically determine the real-time change trend of early warning indicators, and use the change trend of early warning indicators to carry out rockburst early warning;

2) Using the comprehensive anomaly index of shock hazard to quantitatively describe the hazard of rock burst, avoiding the phenomenon of high false alarm/missing rate caused by a single early warning indicator and conflict of early warning levels caused by different early warning indicators;

3) Regular self-feedback update driven by on-site actual monitoring data, has strong scalability and adaptability, adapts to complex and changeable working conditions downhole, and is helpful for efficient and accurate monitoring and early warning of rock burst , and provide efficient and accurate decision-making basis for the prevention and control of downhole rock burst.

In order to better understand the present invention, combined with specific application scenarios, the multi-parameter dynamic trend early warning method for rock burst provided by this embodiment is further described:

In this embodiment, taking a certain working face of a coal mine as an example, first collect the mining period of the working face (February 1, 2018 to February 1, 2019, including September 24, 2018 to October 2018) The original microseismic monitoring data and related information on the 19th due to the mining stop due to the inspection of the working face and no monitoring data): During the mining process of the working face, a total of 15 rock burst events occurred, and early warning for these 15 rock burst events was given. Proceed as follows:

Collect the real-time monitoring data of the on-site microseismic monitoring system, preprocess the original monitoring data with 15 days as the time window and 1 day as the slip step, and calculate the time to obtain the “time-space-strong” multivariate precursory early warning indicators of rock burst danger. Sequence, each early warning index includes: daily total frequency, frequency deviation value, average total frequency, earthquake lack, A(b) value, microseismic activity scale, algorithm complexity, P(b) value, time information entropy, microseismic activity degree , focal concentration degree, daily maximum energy, daily total energy, daily average energy, energy deviation value, average total energy, total fault area, b value, a total of 18 indicators. Show. The positive early warning indicators include: daily maximum energy, daily total energy, daily total frequency, daily average energy, energy deviation value, frequency deviation value, average total energy, microseismic activity, lack of earthquake, A(b) value, fault total Area, average total frequency, microseismic activity scale, algorithm complexity; negative early warning indicators include: focus degree, b value, P(b) value, time information entropy.

The Mann-Kendall trend test method is used to determine the real-time change trend of each early warning index. The trend judgment results of each early warning index at the time of the occurrence of 15 rock burst events are shown in Table 1:

Table 1 The results of trend determination of early warning indicators before rock burst

Note: "/" indicates that the indicator has no obvious trend of change.

Combining the change trend of each index with the rockburst precursor characterization law for early warning, if the indicator occurs within 5 days before the rockburst event, it means that the early warning of the rockburst danger is correct.

The confusion matrix is used to evaluate the early warning efficiency of each indicator, and the calculation results are shown in Table 2:

Table 2 Early warning efficiency evaluation of rockburst early warning indicators

依照预警效能最大化原则，优选前兆预警指标库中预警效能排名前40％内的指标对应的预警效能输入冲击地压综合预警模型中，按照表2中计算结果选取了E_max,D_F,E_avg,E_sum,D_E,F_sum,AC,P(b)等8个指标。Based on the principle of maximizing early warning efficiency, the indicators are selected. According to Table 2, it can be seen that the early warning efficiency of each early warning indicator is arranged in the following order:

According to the principle of maximizing the early warning efficiency, the early warning efficiency corresponding to the indicators within the top 40% of the early warning efficiency in the precursor early warning index library is selected and input into the rockburst comprehensive early warning model, and E _max , D _F , E are selected according to the calculation results in Table 2. _avg , E _sum , D _E , F _sum , AC, P(b) and other 8 indicators.

Using the comprehensive abnormal index method to fuse multiple indicators, the shock risk comprehensive abnormal index Q is obtained, and the time series change curve of Q is shown in Figure 22. Different values of Q correspond to different shock hazard levels, as shown in Table 3.

Table 3 Rockburst hazard classification

Coal rock dynamic disaster risk comprehensive abnormal index Q Levels of danger shock danger 0≤Q＜0.25 Class I no shock hazard 0.25≤Q＜0.5 Class II Weak shock hazard 0.5≤Q＜0.75 Class III medium shock hazard 0.75≤Q≤1 Level IV Strong shock hazard

The early warning efficiency of the comprehensive anomaly index Q of impact risk reaches 0.563 when 8 indicators including E _max , D _F , E _avg , E _sum , D _E , F _sum , AC, P(b) are selected, which is higher than any single early warning Early warning performance of indicators. After that, when the site is replaced every one month or more (or when the working face is replaced, the geological conditions have changed greatly, and a large-energy mine earthquake or rock burst event occurs, etc., the selection of indicators and weight determination shall be performed again to suit the site. Complex working conditions change.

The above are only preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present invention shall be included in the protection of the present invention. within the range.

Claims (7)

1. a multi-parameter dynamic trend early warning method for rock burst, is characterized in that, comprises: S101, according to the real-time monitoring data of the on-site microseismic monitoring system, the Mann-Kendall trend test method is used to describe the change trend of the "time-space-strong" multivariate precursory early warning indicators during the gestation and evolution of rockburst, and determine whether the change trend of each index conforms to The characterization law of rockburst precursors serves as the early warning criterion for early warning; S102, using a confusion matrix to evaluate the early warning effectiveness of each indicator; S103, perform index selection based on the principle of maximizing early warning efficiency; S104, using the early warning efficiency corresponding to the preferred index as a weight to perform multi-index fusion to obtain a comprehensive anomaly index of impact risk; S105, compare the comprehensive anomaly index of impact risk with the corresponding quantitative grading standard to determine the impact risk level; S106, regularly perform index selection and weight update driven by the actual monitoring data on site; Among them, according to the real-time monitoring data of the on-site microseismic monitoring system, the Mann-Kendall trend test method is used to describe the change trend of the "time-space-strong" multivariate precursory early warning indicators in the process of rockburst gestation and evolution. Whether it complies with the rockburst precursor characterization rule is used as an early warning criterion, including: Collect the real-time monitoring data of the on-site microseismic monitoring system, and preprocess the original monitoring data with a specific time window and slip step length to obtain the time series of the multivariate precursory early warning indicators of rockburst hazard "time-space-strong"; The "time-space-strong" multivariate precursory early warning indicators of rock burst danger include: The daily total frequency, frequency deviation value, average total frequency, earthquake absence, A(b) value, microseismic activity scale, algorithm complexity, P(b) value and time information entropy reflecting the time dimension; where, A(b) value and P(b) value are derived indices of the empirical constant b in the Gutenberg-Richter equation, one of the fundamental laws of seismology, and the A(b) value is expressed as:

Among them, N is the total number of microseisms, and _Mi is the energy level of microseismic events; The P(b) value is expressed as:

Microseismic activity and source concentration that reflect spatial dimensions; Daily maximum energy, daily total energy, daily average energy, energy deviation value, average total energy, total fault area and b value reflecting the intensity dimension; Among them, daily maximum energy, daily total energy, daily total frequency, daily average energy, energy deviation value, frequency deviation value, average total energy, microseismic activity, seismic absence, A(b) value, total fault area, average total frequency , microseismic activity scale and algorithm complexity belong to positive early warning indicators, that is, the higher the value, the greater the impact risk; the degree of focus concentration, b value, P(b) value, and time information entropy belong to negative early warning indicators, that is, The lower the value, the greater the impact risk; The Mann-Kendall trend test method is used to determine the real-time change trend of each early warning index, and the early warning criterion is based on whether the change trend of each index conforms to the prewarning rule of rockburst precursory; Wherein, the preprocessing of the original monitoring data with a specific time window and slip step refers to converting the original monitoring data from an irregular time series to a regular time series, including: Define a sliding time window with a length of Δt, and divide the collected monitoring data time series into n data sets with a length of Δt corresponding to the time at the end of the time window, that is, the data set at time T _i is recorded as X _i [x ₁ ,x ₂ ,x ₃ ,...,x _k ], k≤t, 0<i≤n, where Δt is the length of the sliding time window, and T _i is the time corresponding to the end of the i-th time window; Calculate the early warning index values _yi corresponding to all samples in X _i , and arrange the early warning index values _yi in order to obtain the converted regular time series, denoted as Y[y ₁ , y ₂ , y ₃ ,..., y _n ]; Wherein, the use of Mann-Kendall trend test method to determine the real-time change trend of each early warning index includes: Divide the early warning indicator rule time series Y[y ₁ , y ₂ , y ₃ ,...,y _n ] into m data sets with a time window length of Δa corresponding to the time A _i at the end of the time window, namely A _i The data set at the moment is Yi [y ₁ , y ₂ , y ₃ ,...,y _q ], _10≤q≤a , 0< _i≤m , calculate the test statistic S of Yi:

where sgn( ) represents the symbolic function,

q represents the length of the data set Yi [y ₁ , y ₂ , y ₃ ,..., y _q ], y _p represents the p- _th data, p=1, 2, 3...q-1, y _j Represents the jth data, j=p+1,2,3...q; According to the _obtained test statistic S of Yi, determine the test standard _Z of Yi:

When Z>0, the early warning index at time T _i has an increasing trend; when Z < 0, the early warning index at time T _i has a decreasing trend; when Z=0, the early warning index at time T _i has no obvious trend of change. 2. The multi-parameter dynamic trend early-warning method of rock burst according to claim 1, wherein the pre-warning based on whether the variation trend of each index conforms to the pre-warning characterization rule of rock burst as a pre-warning criterion comprises: The real-time change trend of each index is compared with the characterization law of rockburst precursors. If the positive early warning index has an increasing trend, an early warning will be issued, if the negative early warning index has a decreasing trend, an early warning will be issued, and other changing trends will not be issued an early warning. 3. The rockburst multi-parameter dynamic trend early warning method according to claim 1, wherein the early warning efficiency is expressed as:

Among them, EFF means early warning efficiency; Recall means recall rate,

Indicates that the warning is in danger of shock and a shock event has actually occurred,

Indicates that the early warning is no risk of shock but a shock event has actually occurred; Precision indicates the accuracy rate,

in

Indicates that the warning is in danger of shock and a shock event has actually occurred,

Indicates that the warning is a risk of shock but no shock event has actually occurred. 4. The multi-parameter dynamic trend early-warning method for rock burst according to claim 1, characterized in that, the index optimization based on the principle of maximizing early-warning effectiveness comprises: The early warning efficacy values of all early warning indicators are arranged in descending order, and the top n% of the indicators are screened for the next step of data fusion, where n is a positive integer and the number of indicators is rounded up. 5. The multi-parameter dynamic trend early-warning method for rock burst according to claim 1, characterized in that, using the early-warning efficiency corresponding to the preferred index as a weight to perform multi-index fusion to obtain a comprehensive abnormality index of impact risk comprising: The early-warning efficiency corresponding to the preferred index is input into the comprehensive early-warning model of rock burst as the weight, and the comprehensive early-warning model of rock burst uses the comprehensive abnormal index method to perform multi-index fusion to obtain the comprehensive abnormal index of shock risk:

Among them, Q is the comprehensive anomaly index of shock risk; e is the natural base; n is the total number of preferred indicators; EFF _k is the early warning efficiency corresponding to the k-th index; W _k(+/-) means the k-th positive/negative direction The abnormal membership of early warning indicators, For positive early warning indicators, the value of abnormal membership is:

For negative early warning indicators, the value of abnormal membership is:

6 . The multi-parameter dynamic trend pre-warning method of rockburst according to claim 5 , wherein the impact danger level includes: no impact danger state, weak impact danger state, medium impact danger state and strong impact danger state. 7 . 7. The multi-parameter dynamic trend early-warning method of rock burst according to claim 1, characterized in that, said regular index selection and weight update under the driving of actual on-site monitoring data include: Periodically select indicators driven by the actual monitoring data on site, and input the early warning performance corresponding to the optimal indicators as weights into the comprehensive rockburst early warning model, so as to realize the self-feedback update of the comprehensive rockburst early warning model.

Images