CN111980736A - A method of risk prediction and roof fall warning of mine bolt-supported roadway - Google Patents

Abstract

The invention discloses a method for predicting the risk of a mine bolt supporting roadway and early warning of roof fall. The method includes: S1, processing the pressure and displacement data of the corresponding bolt in a cloud server, and calculating the pressure ratio of the corresponding bolt and Displacement ratio, the bolts with the same support type are grouped into the same group, and the bolts in the support structure are sorted according to their critical degree; S2, the optimal coefficients of the pressure ratio and displacement ratio are obtained respectively, and the pressure ratio coefficient is obtained. and the optimization objective function of the displacement ratio coefficient; S3, quantify the pressure ratio and displacement ratio of the corresponding bolt; S4, use the soft-spacing SVM to find the classification hyperplane and its normal vector. The invention provides a method for predicting the risk of a mine bolt-supported roadway and early warning of roof fall. According to the historical data of pressure and displacement of a mine pressure monitoring system, the invention trains a risk-degree model of the bolt-support roadway through machine learning, and provides the current The risk degree of the roadway supported by bolts and the early warning strategy of roof fall, and the occurrence time of roof fall is predicted.

Description

A method of risk prediction and roof fall warning of mine bolt-supported roadway

technical field

The invention relates to a method for risk prediction and roof fall warning of a mine bolt-supported roadway.

Background technique

At present, coal is the most important one-time energy in my country, and the coal industry is the main basic industry of the national economy and plays an important role in the development of the national economy. However, in the process of coal mining, due to the harsh production environment, the complex production process, and the influence of complex ground stress, mining dynamic pressure and other factors, the working face, roof fall and frame, as well as roadway bottom drum, two-gang deformation and coal rock Prominent and other disasters have seriously affected the safe and efficient mining of coal and the safety of personnel and equipment. The bolt support project improves the stress state of the surrounding rock of the roadway through the restraint effect of the bolt on the separation and expansion of the surrounding rock. The support body and the surrounding rock work together to form a complete and stable bearing circle along the roadway, give full play to the role of bolt support, and play the role of actively strengthening the surrounding rock and maintaining the roadway. Coal mine safety production requires safe and reliable roadway support, ensuring the stability of the roadway surrounding rock and avoiding the occurrence of roof collapse accidents, which are the top priorities for avoiding casualties in coal mines.

However, the current mine pressure monitoring has problems such as data recording lag, low measurement accuracy, and measurement is easily affected by the environment, resulting in large errors. Most of them are still manually observed, and the amount of data collection is small and discontinuous, which cannot realize real-time transmission and analysis of monitoring data, and cannot provide accurate and timely early warning.

Some mines have adopted the mine pressure monitoring system based on the Internet of Things, but at present it is mainly in the stage of data collection, storage and preliminary analysis, and has not reached the level of in-depth data analysis and mining. Once a roof fall accident occurs, it is often characterized by strong suddenness and high speed, and the above-mentioned systems often fail to achieve early protection against accidents due to insufficient data utilization.

SUMMARY OF THE INVENTION

The technical problem to be solved by the present invention is to overcome the deficiencies of the prior art and provide a method for predicting the risk of a mine bolt support roadway and for early warning of roof fall. The risk degree model of the bolted roadway is trained. According to the risk degree model of the bolted roadway and combined with the current pressure and displacement parameters, the risk of the current bolted roadway and the early warning strategy for roof fall are given, and the occurrence time of roof fall is predicted. .

In order to solve the above-mentioned technical problems, the technical scheme of the present invention is:

A method for risk prediction and roof fall warning of mine bolt-supported roadway, comprising:

S1: Process the pressure and displacement data of the corresponding bolts in the cloud server, calculate the pressure ratio and displacement ratio of the corresponding bolts, and group the bolts with the same support type into the same group. The rods are sorted by criticality;

S2, obtain the optimal coefficients of the pressure ratio and displacement ratio respectively, and obtain the optimal objective function of the pressure ratio coefficient and the displacement ratio coefficient;

S3, quantify the pressure ratio and displacement ratio of the corresponding bolt;

S4, use soft margin SVM to find the classification hyperplane and its normal vector;

S5, obtain the mean and standard deviation of the data obtained after the accident data is projected on the normal vector axis, obtain the mean and standard deviation of the data obtained after the normal working data is projected on the normal vector axis, and then obtain through the conditional entropy The risk degree of the bolt-supported roadway, and the risk degree of the bolt-supported roadway is predicted;

S6, generating a roof fall warning strategy, when the risk of the roadway supported by the bolt is greater than the warning threshold, the warning is started.

Further, the step S1 includes:

Let the pressure ratio of the i-th recorded data of the j-th anchor rod be:

Let the displacement ratio of the i-th recorded data of the j-th anchor rod be:

Among them, i is the ith sample data, i=1,2,...,N;

j is the j-th anchor, j=1,2,...,M, the first anchor is the most critical anchor, and the critical degree gradually decreases in the following order;

P _ji is the pressure value of the i-th recorded data of the j-th anchor;

P _j0 is the pressure rating of the jth anchor;

L _ji is the displacement value of the i-th record data of the j-th anchor;

L _j0 is the maximum allowable displacement value of the j-th anchor.

Further, the step S2 includes:

The optimized objective function of the pressure ratio coefficient and displacement ratio coefficient is:

in,

C ₁ is the pressure ratio coefficient;

C ₂ is the displacement ratio coefficient;

y _i is the class label of the i-th data, and when y _i is -1, it means an accident occurs:

Use gradient descent to find C, ie (C1, C2):

趋近0时，迭代所得的C^*为最优解；can be obtained, when L(C) is the smallest, that is

When approaching 0, the C ^* obtained by iteration is the optimal solution;

where α represents the step size.

Further, the step S3 includes:

The pressure ratio and displacement ratio of the corresponding bolt are quantified, and the quantization interval is 0.1, and the

Further, the step S4 includes:

S41, establish a data vector x=(x ⁽¹⁾ ,x ⁽²⁾ ,...,x ^(j) ,...,x ^(M) );

S42, establish a coefficient vector w=(w ⁽¹⁾ ,w ⁽²⁾ ,...,w ^(j) ,...,w ^(M) ), where w ^(j) is the coefficient corresponding to the corresponding feature x ^(j) ;

S43, x _i is the ith training data vector, y _i is the class label of x _i ; when y _i is -, it means an accident occurs, when y _i is +1, it means that the work is normal, and N is the number of training data;

S44, use soft margin SVM to find the classification hyperplane with the largest geometric margin, and the problem can be expressed as a constrained optimization problem:

S.ty _i (wx _i +b)≥1-ξ _i

ξ _i ≥ 0i=1, 2,...N;

Among them, F is the penalty coefficient; ξ is the slack variable; ξ _i is the slack variable of the i-th training data; b is the bias;

Obtain the optimal solution w ^* of the optimal classification hyperplane and the coefficient vector, where w ^* is the normal vector of the optimal classification hyperplane:

为拉格朗日乘子向量中对偶问题的解的第i个元素。in,

is the ith element of the solution to the dual problem in the vector of Lagrangian multipliers.

Further, the step S5 includes:

S51, obtain the data obtained by projecting the data on the normal vector, and obtain the mean value and label difference of the two data categories of accidents and normal work:

Among them, NA is the number of samples of category y ₌ 1; _{NB is the number of samples of category y=-1; μ B is} the mean value of the data obtained after the accident data is projected on the w ^* vector axis; μ _A is the normal operation The mean value of the data obtained after the data is projected on the w ^* vector axis; δ _A is the standard deviation of the data obtained after the normal work data is projected on the w ^* vector axis; δ _B is the accident data after the w ^* vector axis projection. the standard deviation of the obtained data;

S52, predict the risk degree of the roadway supported by bolts:

Let x _A be the sum of all sample data with y=1, and Z _A be the projection of x _A on the w ^* vector axis;

Let the current data be x _c , Zc=w ^* x _c ;

Then the current risk degree V of bolt support roadway is:

When wx _c ≤ μ _A +δ _A , the risk V is 0;

When wx _c ≥ μ _B -δ _B , the risk V is 1;

When μ _A +δ _A ≤wx _c ≤μ _B -δ _B , the risk V is:

in,

H(Z _C |Z _A ) is the conditional entropy of Z _C under a given z _A condition;

H(μ _B |Z _A ) is the conditional entropy of μ _B at a given z _A condition.

Further, the step S6 includes:

S61, when the risk degree V>γ, start an early warning;

Wherein, γ is an early warning threshold, and the range of the early warning threshold γ is 0.3-0.4;

S62, predict the occurrence time of roof fall:

T=(1-V)T ₀ ;

in,

T is the time from the current data recording time, and it is predicted that the roof will fall;

T ₀ is the reference estimated working time when the current bolt support structure works normally.

Having adopted the above-mentioned technical scheme, the present invention has the following beneficial effects:

1. The present invention avoids the disadvantages of strong subjectivity and randomness of manual risk judgment and early warning based on monitoring data, and reduces the labor cost of the mine.

2. The present invention can predict the risk degree of the current bolt support system and the occurrence time of roof fall, which is beneficial to prevent accidents in advance.

Description of drawings

Fig. 1 is the principle block diagram of the structure of the mine pressure monitoring system of the present invention;

Fig. 2 is a flow chart of a method for risk prediction and roof fall warning of a mine bolting roadway according to the present invention.

Detailed ways

In order to make the content of the present invention easier to understand clearly, the present invention will be described in further detail below according to specific embodiments and in conjunction with the accompanying drawings.

Figure 1 shows the structure of the mine pressure monitoring system. The acquisition nodes include pressure nodes and displacement nodes, which are the basic units constituting the monitoring system. They are located at the bottom of the entire system and are installed on the underground anchoring equipment. They are responsible for collecting pressure and displacement data, and storing the data. It is transmitted to the cloud server through the ZigBee coordinator, and the pressure and displacement data in the cloud server are used to train the risk model of the roadway for bolt support.

As shown in Figure 2, a method of risk prediction and roof fall warning of mine bolt support roadway, which includes:

S1: Process the pressure and displacement data of the corresponding bolts in the cloud server, calculate the pressure ratio and displacement ratio of the corresponding bolts, and group the bolts with the same support type into the same group. The rods are sorted by criticality;

S2, obtain the optimal coefficients of the pressure ratio and displacement ratio respectively, and obtain the optimal objective function of the pressure ratio coefficient and the displacement ratio coefficient;

S3, quantify the pressure ratio and displacement ratio of the corresponding bolt;

S4, use soft margin SVM to find the classification hyperplane and its normal vector;

S5, obtain the mean and standard deviation of the data obtained after the accident data is projected on the normal vector axis, obtain the mean and standard deviation of the data obtained after the normal working data is projected on the normal vector axis, and then obtain through the conditional entropy The risk degree of the bolt-supported roadway, and the risk degree of the bolt-supported roadway is predicted;

S6, generating a roof fall warning strategy, when the risk of the roadway supported by the bolt is greater than the warning threshold, the warning is started.

Further, the step S1 includes:

Let the pressure ratio of the i-th recorded data of the j-th anchor rod be:

Let the displacement ratio of the i-th recorded data of the j-th anchor rod be:

Among them, i is the ith sample data, i=1,2,...,N;

j is the j-th anchor, j=1,2,...,M, the first anchor is the most critical anchor, and the critical degree gradually decreases in the following order;

P _ji is the pressure value of the i-th recorded data of the j-th anchor;

P _j0 is the pressure rating of the jth anchor;

L _ji is the displacement value of the i-th record data of the j-th anchor;

L _j0 is the maximum allowable displacement value of the j-th anchor.

Further, the step S2 includes:

The optimized objective function of the pressure ratio coefficient and displacement ratio coefficient is:

in,

C ₁ is the pressure ratio coefficient;

C ₂ is the displacement ratio coefficient;

y _i is the class label of the i-th data, and when y _i is -1, it means an accident occurs:

Use gradient descent to find C, ie (C1, C2):

趋近0时，迭代所得的C^*为最优解；can be obtained, when L(C) is the smallest, that is

When approaching 0, the C ^* obtained by iteration is the optimal solution;

where α represents the step size.

Further, the step S3 includes:

The pressure ratio and displacement ratio of the corresponding bolt are quantified, and the quantization interval is 0.1, and the

Further, the step S4 includes:

S41, establish a data vector x=(x ⁽¹⁾ ,x ⁽²⁾ ,...,x ^(j) ,...,x ^(M) );

S42, establish a coefficient vector w=(w ⁽¹⁾ ,w ⁽²⁾ ,...,w ^(j) ,...,w ^(M) ), where w ^(j) is the coefficient corresponding to the corresponding feature x ^(j) ;

S43, x _i is the ith training data vector, y _i is the class label of x _i ; when y _i is -, it means an accident occurs, when y _i is +1, it means that the work is normal, and N is the number of training data;

S44, use soft margin SVM to find the classification hyperplane with the largest geometric margin, and the problem can be expressed as a constrained optimization problem:

S.ty _i (wx _i +b)≥1-ξ _i

ξ _i ≥ 0i=1, 2,...N;

Among them, F is the penalty coefficient; ξ is the slack variable; ξ _i is the slack variable of the i-th training data; b is the bias; by transforming the original problem into a dual problem, using the KKT condition, find the optimal solution of the dual problem, The optimal solution w ^* of the optimal classification hyperplane and the coefficient vector can be obtained, and w ^* is the normal vector of the optimal classification hyperplane:

为拉格朗日乘子向量中对偶问题的解的第i个元素。in,

is the ith element of the solution to the dual problem in the vector of Lagrangian multipliers.

Further, the step S5 includes:

S51, obtain the data obtained by projecting the data on the normal vector, and obtain the mean value and label difference of the two data categories of accidents and normal work:

Among them, NA is the number of samples of category y ₌ 1; _{NB is the number of samples of category y=-1; μ B is} the mean value of the data obtained after the accident data is projected on the w ^* vector axis; μ _A is the normal operation The mean value of the data obtained after the data is projected on the w ^* vector axis; δ _A is the standard deviation of the data obtained after the normal work data is projected on the w ^* vector axis; δ _B is the accident data after the w ^* vector axis projection. the standard deviation of the obtained data;

S52, predict the risk degree of the roadway supported by bolts:

Let x _A be the sum of all sample data with y=1, and Z _A be the projection of x _A on the w ^* vector axis;

Let the current data be x _c , Zc=w ^* x _c ;

Then the current risk degree V of bolt support roadway is:

When wx _c ≤ μ _A +δ _A , the risk V is 0;

When wx _c ≥ μ _B -δ _B , the risk V is 1;

When μ _A +δ _A ≤wx _c ≤μ _B -δ _B , the risk V is:

in,

H(Z _C |Z _A ) is the conditional entropy of Z _C under a given z _A condition;

H(μ _B |Z _A ) is the conditional entropy of μ _B under a given z _A condition;

The final risk degree V is a number ranging from 0 to 1. The closer the risk degree V is to 0, the lower the risk degree is, and the closer to 1 the risk degree is, the greater the risk degree is. The user can decide preventive maintenance and advance protection according to the size of the risk degree V .

Further, the step S6 includes:

S61, when the risk degree V>γ, start an early warning;

Among them, γ is an early warning threshold, and the range of the early warning threshold γ is 0.3-0.4, and the user can independently select the threshold setting within this range;

The range of the above warning threshold γ is reasonable, which is a good compromise between avoiding false alarms and avoiding false alarms.

S62, predict the occurrence time of roof fall:

T=(1-V)T ₀ ;

in,

T is the time from the current data recording time, and it is predicted that the roof will fall;

T ₀ is the benchmark estimated working time when the current bolt support structure works normally, which can be estimated and set by experts or experienced personnel according to the actual situation.

The specific embodiments described above further describe in detail the technical problems, technical solutions and beneficial effects solved by the present invention. It should be understood that the above are only specific embodiments of the present invention, and are not intended to limit the present invention. invention, any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included within the protection scope of the present invention.

Claims (7)

1. The method for predicting the risk degree of a mine anchor bolt supporting roadway and warning roof collapse is characterized by comprising the following steps of:

s1, processing the pressure and displacement data of the corresponding anchor rods in the cloud server, calculating the pressure ratio and the displacement ratio of the corresponding anchor rods, dividing the anchor rods with the same support type into the same group, and sequencing the anchor rods in the support structure according to the key degree;

s2, respectively calculating the optimal coefficients of the pressure ratio and the displacement ratio to obtain the optimal objective functions of the pressure ratio coefficient and the displacement ratio coefficient;

s3, quantifying the pressure ratio and the displacement ratio of the corresponding anchor rods;

s4, solving a classification hyperplane and a normal vector thereof by using a soft space SVM;

s5, solving the mean value and the standard deviation of the data obtained after the accident data are projected on the normal vector axis, solving the mean value and the standard deviation of the data obtained after the normal working data are projected on the normal vector axis, then solving the danger degree of the anchor bolt supporting roadway through the conditional entropy, and predicting the danger degree of the anchor bolt supporting roadway;

and S6, generating a roof collapse early warning strategy, and starting early warning when the danger degree of the anchor bolt supporting roadway is greater than an early warning threshold value.

2. The method for predicting the risk degree of the mine bolting roadway and warning the roof fall according to claim 1, wherein the step S1 includes:

let the pressure ratio of the ith recording data of the jth anchor rod be:

let the displacement ratio of the ith recording data of the jth anchor rod be:

wherein, i is the ith sample data, i is 1, 2.

j is the jth anchor rod, j is 1,2, a.

P_jiRecording the pressure value of data for the ith anchor rod;

P_j0the pressure rating of the jth anchor rod;

L_jirecording the displacement value of the data for the ith anchor rod;

L_j0the allowable maximum displacement value of the jth anchor rod.

3. The method for predicting the risk degree of the mine bolting roadway and warning the roof fall according to claim 2, wherein the step S2 includes:

the optimized objective function of the pressure ratio coefficient and the displacement ratio coefficient is as follows:

wherein,

C₁is the pressure ratio coefficient;

C₂is a displacement ratio coefficient;

y_ifor class marking of ith data, y_iA value of-1 indicates the occurrence of an accident:

c, i.e. (C1, C2) using a gradient descent method:

can be obtained when L (C) is minimum, i.e.

When approaching 0, iterating the obtained C^*Is the optimal solution;

where α represents the step size.

4. The method for predicting the risk degree of the mine bolting roadway and warning the roof fall according to claim 3, wherein the step S3 includes:

quantizing the pressure ratio and the displacement ratio of the corresponding anchor rods, wherein the quantization interval is 0.1 to obtain

5. The method for predicting the risk degree of the mine bolting roadway and warning the roof fall according to claim 4, wherein the step S4 includes:

s41, establishing a data vector x ═ x⁽¹⁾,x⁽²⁾,…,x^(j),…,x^(M))；

S42, establishing coefficient vector w ═ w (w)⁽¹⁾,w⁽²⁾,…,w^(j),…,w^(M)) Wherein w is^(j)For the corresponding feature x^(j)The corresponding coefficient;

S43，x_ifor the ith training data vector, y_iIs x_iClass label of (1); y is_iFor-time to indicate an accident, y_iWhen the value is +1, the work is normal, and N is the number of training data;

s44, using a soft space SVM to solve the classification hyperplane with the maximum geometric space:

S.t y_i(w.x_i+b)≥1-ξ_i

ξ_i≥0 i＝1，2，...N；

wherein F is a penalty coefficient; xi is a relaxation variable; xi_iRelaxation variables for the ith training data; b is an offset;

obtaining the optimal solution of the optimal classification hyperplane and coefficient vectorw^*，w^*Namely, the normal vector of the optimal classification hyperplane:

wherein,

is the ith element of the solution to the dual problem in the lagrange multiplier vector.

6. The method for predicting the risk degree of the mine bolting roadway and warning the roof fall according to claim 5, wherein the step S5 includes:

s51, obtaining data obtained by projecting the data on a normal vector, and obtaining the mean value and the labeling difference of two data types of accidents and normal work:

wherein N is_AIs as in class y-1The number of the books; n is a radical of_BThe number of samples in the category y-1; mu.s_BFor data of accidents at w^*Mean value of the data obtained after vector axis projection; mu.s_ANormal data for operation at w^*Mean value of the data obtained after vector axis projection;_Anormal data for operation at w^*Standard deviation of data obtained after vector axis projection;_Bfor data of accidents at w^*Standard deviation of data obtained after vector axis projection;

s52, predicting the anchor bolt support roadway risk:

let x_AFor the sum of sample data for all y ═ 1, Z_AIs x_AAt w^*Projection of the vector axis;

let current data be x_c，Zc＝w^*x_c；

Then the current bolting roadway risk degree V is:

when wx_c≤μ_A+_AWhen the risk degree V is 0;

when wx_c≥μ_B-_BThe risk V is 1;

when mu is_A+_A≤wx_c≤μ_B-_BThe risk V is:

wherein,

H(Z_C|Z_A) To give at a given z_AUnder the condition of Z_CThe conditional entropy of (a);

H(μ_B|Z_A) To give at a given z_AUnder the condition of_BThe conditional entropy of (1).

7. The method for predicting the risk degree of the mine bolting roadway and warning the roof fall according to claim 6, wherein the step S6 includes:

s61, when the risk degree V is larger than gamma, starting early warning;

wherein gamma is an early warning threshold value, and the range of the early warning threshold value gamma is 0.3-0.4;

s62, predicting the roof fall occurrence time:

T＝(1-V)T₀；

wherein,

t is the time which starts to time from the current data recording time and predicts the time of roof fall;

T₀and estimating the working time for the reference when the current anchor rod supporting structure works normally.

Images