CN116882548B - Tunneling roadway coal and gas outburst prediction method based on dynamic probability reasoning - Google Patents
Tunneling roadway coal and gas outburst prediction method based on dynamic probability reasoning Download PDFInfo
- Publication number
- CN116882548B CN116882548B CN202310710975.2A CN202310710975A CN116882548B CN 116882548 B CN116882548 B CN 116882548B CN 202310710975 A CN202310710975 A CN 202310710975A CN 116882548 B CN116882548 B CN 116882548B
- Authority
- CN
- China
- Prior art keywords
- risk
- variables
- probability
- variable
- reasoning
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 54
- 230000005641 tunneling Effects 0.000 title claims abstract description 47
- 239000003245 coal Substances 0.000 title claims abstract description 42
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 6
- 238000005553 drilling Methods 0.000 claims description 35
- 238000012360 testing method Methods 0.000 claims description 29
- 238000004458 analytical method Methods 0.000 claims description 27
- 238000005520 cutting process Methods 0.000 claims description 27
- 238000004364 calculation method Methods 0.000 claims description 17
- 238000005065 mining Methods 0.000 claims description 16
- 238000010206 sensitivity analysis Methods 0.000 claims description 16
- 238000003795 desorption Methods 0.000 claims description 14
- 230000008569 process Effects 0.000 claims description 14
- 230000002441 reversible effect Effects 0.000 claims description 13
- 230000008859 change Effects 0.000 claims description 12
- 238000003745 diagnosis Methods 0.000 claims description 9
- 238000011835 investigation Methods 0.000 claims description 9
- 230000007704 transition Effects 0.000 claims description 9
- 230000001965 increasing effect Effects 0.000 claims description 8
- 230000002265 prevention Effects 0.000 claims description 8
- 230000006378 damage Effects 0.000 claims description 5
- 230000009471 action Effects 0.000 claims description 4
- 238000010276 construction Methods 0.000 claims description 4
- 230000007246 mechanism Effects 0.000 claims description 4
- 238000007476 Maximum Likelihood Methods 0.000 claims description 3
- 230000033228 biological regulation Effects 0.000 claims description 3
- 230000003247 decreasing effect Effects 0.000 claims description 3
- 230000002542 deteriorative effect Effects 0.000 claims description 3
- 230000035945 sensitivity Effects 0.000 claims description 3
- 238000013480 data collection Methods 0.000 claims description 2
- 238000012163 sequencing technique Methods 0.000 claims 1
- 238000011161 development Methods 0.000 abstract description 4
- 230000001427 coherent effect Effects 0.000 abstract description 3
- 230000003068 static effect Effects 0.000 description 9
- 230000008901 benefit Effects 0.000 description 6
- 230000000694 effects Effects 0.000 description 5
- 230000002829 reductive effect Effects 0.000 description 4
- 230000002159 abnormal effect Effects 0.000 description 3
- 238000013517 stratification Methods 0.000 description 3
- 125000002015 acyclic group Chemical group 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000036961 partial effect Effects 0.000 description 2
- 238000000638 solvent extraction Methods 0.000 description 2
- 238000009825 accumulation Methods 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000032798 delamination Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 230000006866 deterioration Effects 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000002349 favourable effect Effects 0.000 description 1
- 238000013467 fragmentation Methods 0.000 description 1
- 238000006062 fragmentation reaction Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000001939 inductive effect Effects 0.000 description 1
- 230000004941 influx Effects 0.000 description 1
- 230000002045 lasting effect Effects 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 230000035699 permeability Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000007619 statistical method Methods 0.000 description 1
- 239000000758 substrate Substances 0.000 description 1
- 230000008719 thickening Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N5/00—Computing arrangements using knowledge-based models
- G06N5/04—Inference or reasoning models
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N7/00—Computing arrangements based on specific mathematical models
- G06N7/01—Probabilistic graphical models, e.g. probabilistic networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0635—Risk analysis of enterprise or organisation activities
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/10—Services
- G06Q50/26—Government or public services
- G06Q50/265—Personal security, identity or safety
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Human Resources & Organizations (AREA)
- General Physics & Mathematics (AREA)
- Economics (AREA)
- Strategic Management (AREA)
- Tourism & Hospitality (AREA)
- Entrepreneurship & Innovation (AREA)
- Marketing (AREA)
- General Business, Economics & Management (AREA)
- Development Economics (AREA)
- Mathematical Physics (AREA)
- Evolutionary Computation (AREA)
- Operations Research (AREA)
- Educational Administration (AREA)
- Software Systems (AREA)
- Game Theory and Decision Science (AREA)
- Quality & Reliability (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Computing Systems (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Primary Health Care (AREA)
- Health & Medical Sciences (AREA)
- Computer Security & Cryptography (AREA)
- Probability & Statistics with Applications (AREA)
- Algebra (AREA)
- Computational Mathematics (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
- Pure & Applied Mathematics (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention relates to a tunneling roadway coal and gas outburst prediction method based on dynamic probability reasoning, and belongs to the field of tunneling roadway coal and gas outburst disaster prediction methods. In order to make risk variable reasoning more comprehensive and accurate and to construct 13 variables and introduce expansion (dynamic Bayesian network) in Bayesian network reasoning, the development of underground tunneling working faces with risk connection is provided as a time slice, a risk prediction scheme of continuous probability reasoning is realized, and two methods of parameter learning under a expectation maximization algorithm and expert reasoning decision under a fuzzy set theory are combined to determine the probability distribution of initial Bayesian network nodes without influence. The structure is compact and coherent, the prediction is flexible, and the accuracy is high.
Description
Technical Field
The invention relates to a tunneling roadway coal and gas outburst prediction method based on dynamic probability reasoning, and belongs to the field of tunneling roadway coal and gas outburst disaster prediction methods.
Background
The existing gas outburst prediction method is difficult to comprehensively reflect the outburst risk due to the fact that the critical value prediction of a single index, and the multi-index mathematical modeling and prediction method is widely focused. The related research is based on ideal solution similarity sorting, a development theory, a fuzzy set analysis, a D-S evidence fusion theory, a limited interval cloud model and the like. Although having certain application effects, certain limitations exist. For example: weight calculation is generally introduced, and the prediction effect is not stable enough due to the diversity of weight determination methods; the prominent risk is a complex process which continuously changes along with the mining progress, the current static working face risk prediction based on drilling and geophysical prospecting generally lacks consideration of time factors, analysis is performed only on the basis of the cycle prediction data, and the reference of the risk of the previous cycle to the current cycle risk prediction is ignored; the prediction method focuses on result prediction, i.e. unidirectional prediction. At present, aiming at different actually existing risk states, key factors and anti-outburst control decision bases are lacked.
Disclosure of Invention
Aiming at the defects of the prior art, the tunneling roadway coal and gas outburst prediction method based on dynamic probability reasoning is simple in steps, convenient to use, free of weight calculation, capable of avoiding the instability influence of weights on the traditional outburst prediction method, capable of rapidly identifying key variables causing risk corresponding states under any cycle by combining reverse diagnosis reasoning and sensitivity analysis.
In order to achieve the technical purpose, the invention discloses a tunneling roadway coal and gas outburst prediction flow based on dynamic probability reasoning, which comprises the following steps:
Step 1, analyzing outstanding risk variables of a tunneling roadway based on geological conditions of a mine where a working face is located, wherein the geological conditions comprise: gas occurrence state, conventional drilling test parameters, coal seam conditions and geological structures; the gas outburst risk variables of the tunneling roadway comprise: mining depth X1, gas pressure X2, gas content X3, damage type X4, geological structure complexity degree X5, firmness coefficient X6, drilling cuttings quantity X7, drilling gas emission initial speed X8, drilling cuttings desorption index X9, soft layering thickness X10, local coal thickness change X11, fault structure X12 and fold structure X13; wherein variables relating to the area prediction process and the work surface prediction process include X1, X2, X3, X4, X6, and X10; the working face prediction process comprises X5, X7, X8 and X9; defining the degree of the prominent risk as a target variable Y, determining an inherent logic relationship between the prominent risk variables X1-X13 and the target variable Y, and establishing a directed acyclic graph in the Bayesian network, wherein the directed acyclic graph comprises 14 corresponding nodes;
step 2, dividing the state category of each node in the directed acyclic graph; dividing a tunneling working face circulation time slice, taking the prominent risk variables X1-X13 as root nodes and intermediate nodes, and taking the target variable Y as leaf nodes; establishing the connection of adjacent time slices, constructing time state transition probability, determining initial prior probability and conditional probability distribution parameters of each node, establishing a dynamic Bayesian network salient prediction model, and calculating by adopting a mode combining parameter learning and expert decision analysis;
Step 3, based on a Bayesian network calculation principle, when the working face prediction drilling test is not started, utilizing the initial probability distribution comprehensive reasoning of each root node of the Bayesian network to obtain the probability result of the leaf node, namely the probability distribution of the prominent risk degree of the initial working face; along with continuous test of drilling test and geological parameters of continuous tunneling circulation, continuously updating variable information of geological structure complexity degree X5, firmness coefficient X6, drilling cuttings quantity X7, drilling gas emission initial speed X8 and drilling cuttings desorption index X9, and combining probability distribution of the salient risk Y t-1 of the previous circulation and time state transition probability of construction, so that probability distribution of the salient risk Y t of the current circulation can be obtained; in the process of predicting any cycle, carrying out reverse diagnosis analysis and sensitivity analysis at any time;
And 4, determining key variables influencing the current risk state in any working face cycle based on reverse diagnosis analysis, and then pertinently adopting control, judging variables with larger influence on a prediction result based on sensitivity analysis, wherein the higher the sensitivity of the variables is, the more the representative variables need to continuously pay attention to the changes, so that the testing precision is improved, and the salient risk is prevented from deteriorating.
Further, based on a protruding mechanism, the soft layering thickness, the local coal thickness change, the fault structure and the buckling structure of the mine where the working face is located directly influence the complexity of the geological structure; the mining depth is related to the quantity of drill cuttings based on stress; the gas pressure X2, the gas content, the damage type and the firmness coefficient are connected with the initial gas emission speed of the drilling hole, the drill cuttings desorption index and the soft layering thickness to different degrees; and constructing a directed acyclic graph by combining the logical relations of the variables, dividing root nodes, intermediate nodes and leaf nodes of the directed acyclic graph, and properly increasing or decreasing and adjusting the variables by combining the variable test conditions of actual mines.
Further, the state categories of the nodes in the directed acyclic graph respectively comprise 2-4 state classifications, and the prominent risks are divided into 3 types: safety L1, threat L2 and outburst risk L3 are used as reference bases for further pre-acquisition outburst prevention decision of each cycle;
Defining the prominent risk variables X1, X2, X3, X4, X6, X10, X11, X12 and X13 as root nodes, and preferentially calculating initial prior probability according to the statistical frequency data of the corresponding variables in the historical accidents;
Defining the prominent risk variables X5, X7, X8 and X9 as intermediate nodes, adopting a expectation maximization algorithm of maximum likelihood estimation, and carrying out parameter learning through the historical mining cycle sample data of the mine where the prominent risk variables X5, X7, X8 and X9 are positioned to determine the conditional probability distribution P (Y|X5, X7, X8 and X9) of the target variable Y under the action of the intermediate nodes X5, X7, X8 and X9; the rest variables use expert decision analysis method;
the prior probability and the conditional probability distribution of the residual target nodes Xi (i=1, 2, …, 13) are determined through probability interval division, expert survey data collection and defuzzification in sequence; the probability interval division is combined with the triangle fuzzy number construction, so that the problem of ambiguity that the prominent risk variable is difficult to correspond to a single state is solved;
Leading in an expert confidence index theta to ensure that expert investigation results are more reliable, adopting an alpha weighted estimation method to deblur the expert investigation results, and maximally reserving the information integrity after deblurring;
And obtaining an initial prominent risk dynamic Bayesian network reasoning prediction model based on the established directed acyclic graph and the variable probability distribution parameters.
Further, the risk variables X1 and X6 affect the risk variable X7 together, the risk variables X2, X3 and X6 affect the risk variable X8 together, the risk variables X2, X3 and X4 affect the risk variable X9 together, the risk variables X10, X11, X12 and X13 affect the risk variable X5 together, and the risk variables X5, X7, X8 and X9 affect the target variable Y together.
Further, the specific steps of the reverse diagnostic analysis are:
setting the corresponding risk state probability of the target node Y to 1 in combination with the current actual prominent risk state, i.e., P (y=l1) =1, P (y=l2) =1 or P (y=l3) =1;
Based on the Bayesian network calculation principle, the probability distribution of each node Xi (i=1, 2, …, 13) can be obtained by reverse reasoning, and compared with the probability distribution before setting, each variable is orderly sequenced according to the probability change degree, so that the key variable of the current risk state is determined.
Further, the sensitivity analysis specifically comprises the following steps: the sensitivity analysis method can identify the variables with larger variation amplitude and larger contribution to the fluctuation regulation and control of the risk results in the current tunneling cycle.
Compared with the prior art, the invention has the beneficial effects that:
The influence of the prominent risk of the previous cycle on the current cycle prominent risk cycle is not considered in the traditional working face risk static drilling prediction, and various methods are used for predicting the risk of an independent single cycle, so that the risk prediction between the cycles has great fluctuation and low accuracy. The method introduces expansion (dynamic Bayesian network) in Bayesian network inference, and proposes a risk prediction scheme of taking a tunneling working face with risk connection in the pit as a time slice and realizing continuous probabilistic reasoning. The obtained prediction result is more compact and coherent and has higher accuracy compared with the static Bayesian network reasoning result;
In combination with a plurality of factors for influencing the outburst risk of the underground tunneling working face, 13 variables are constructed for enabling the risk variable reasoning to be more comprehensive and accurate, but the actual testing of the outburst risk variable is asynchronous due to the fact that the specificity of two working procedures is predicted based on the fact that the coal mine is divided into a region and the working face, and therefore the data set collection is incomplete. Therefore, the two methods of parameter learning under the expectation maximization algorithm and expert reasoning decision under the fuzzy set theory are combined, and the probability distribution of the initial Bayesian network nodes is determined, so that the probability distribution has no influence on each other, and the probability distribution is more flexible in on-site forward probability reasoning risk prediction. Current bayesian reasoning generally employs one of two approaches, but is difficult to implement in the salient risk inference.
The advantages are that: according to the invention, through clearing the logic relation among 13 variables directly related to the salient risk, each cycle of the tunneling roadway is divided into independent time slices, and a salient risk prediction method based on dynamic Bayesian continuous probability reasoning is established. (1) The method completely carries out probabilistic reasoning according to the logical relation of the variables, is simple and convenient to operate, does not need weight calculation, and avoids the instability influence caused by the traditional outstanding prediction effect.
(2) Based on the probability connection of each node of the network, the connection between each variable and the risk in the same tunneling cycle is established, and the influence of the prediction result of the previous cycle on the prediction result of the current cycle is mainly considered, so that the prediction result of the outstanding risk is more continuous, stable and high in accuracy.
(3) Based on the unique advantages of probabilistic reasoning, the key variables causing the corresponding states of risks can be rapidly identified under any cycle by combining reverse diagnosis reasoning with sensitivity analysis. As the basis of the anti-outburst decision, and further as a targeted control measure, the limitation caused by uncertainty generated by subjective decision making of field experts is avoided.
Drawings
FIG. 1 shows a tunneling roadway coal and gas outburst risk prediction flow based on dynamic probabilistic reasoning in an embodiment of the invention.
Figure 2 illustrates a directed acyclic graph constructed based on risk variables in an embodiment of the invention.
FIG. 3 shows statistical analysis of historical salient accident causes of mines with test working surfaces in an embodiment of the invention.
FIG. 4 is a schematic diagram of a method for performing a deblurring calculation using an alpha weighted estimation in accordance with an embodiment of the present invention.
FIG. 5 shows an initial probability distribution diagram before test face tunneling in an embodiment of the present invention.
FIG. 6 shows a dynamic Bayesian probability distribution based on continuous update of evidence information for work surface loop correspondence in an embodiment of the present invention.
FIG. 7 shows a comparison of the static Bayes and the dynamic Bayes of the salient risk prediction results under different tunneling cycles of the test working face in the embodiment of the invention.
FIG. 8 shows the result of the test face reverse reasoning analysis calculation in the embodiment of the present invention.
FIG. 9 shows the results of the test face sensitivity analysis calculation in the embodiment of the present invention.
Detailed Description
Embodiments of the present invention will be further described with reference to the accompanying drawings:
Aiming at a tunneling roadway of a certain outburst mine, a tunneling roadway coal and gas outburst prediction flow based on dynamic probabilistic reasoning is developed, and the specific steps are as follows, as shown in fig. 1:
And (3) highlighting risk variable analysis.
Step 1.1, determining a risk variable set from the aspects of gas occurrence state, conventional drilling test parameters, coal bed conditions, geological structures and the like based on geological conditions of a mine where an actual predicted working face is combined with 'prevention and control coal and gas outburst rules', determining an internal logic relationship, and establishing a directed acyclic graph in a Bayesian network. Wherein the mining depth (X1) refers to the depth of the coal seam area where the working surface is located from the earth surface; the gas pressure (X2) and the gas content (X3) refer to the gas pressure and the gas content in the coal seam area where the working face is located; the destruction type (X4) is classified according to the degree of coal body fragmentation, and is explicitly described in the rules for preventing and treating coal and gas outburst; the geological structure complexity (X5) refers to the comprehensive geological structure condition in the range of the tunneling working face; the firmness coefficient (X6) refers to the mechanical strength of the coal body, and the larger the firmness coefficient is, the harder the coal body is represented. The cuttings quantity (X7) refers to the mass or volume of cuttings per meter during drilling a 42mm diameter borehole in the coal body; the initial drilling gas emission speed (X8) refers to the gas flow of drilling in 2min after drilling is finished by constructing a drilling hole with a designated depth in a coal seam; the drill cuttings desorption index (X9) refers to one of K1 value or Deltah 2 value. The K1 value represents the slope of a Shi Wasi desorption curve of a drilling cuttings sample during an initial period of pressure relief (5 min); the value of delta h 2 refers to the pressure difference generated by gas desorption of a drilling cutting sample in the initial period of pressure relief (2 min), and the mine is tested by delta h 2; the soft delamination thickness (X10) refers to the thickness of a less hard structural coal body in a coal seam in a tunneling cycle; the local coal thickness change (X11) refers to the condition that whether the coal layer has larger thickness change or not; the fault structure (X12) refers to whether the fault structure exists in front of the working face tunneling cycle; the flexure configuration (X13) refers to whether a flexure configuration exists in front of the face heading cycle.
Step 1.2, determining the thickness of soft layering, local coal thickness change, fault structure and buckling structure based on a salient mechanism, wherein the complexity of the geological structure is directly influenced. The mining depth is related to the quantity of drill cuttings based on stress; the gas pressure, gas content, type of failure and solidity coefficients are related to the initial rate of gas influx from the borehole, the cuttings desorption index and the soft stratification thickness to different extents. And constructing a directed acyclic graph by combining the logical relations of the variables, and dividing root nodes, intermediate nodes and leaf nodes. The variable can be properly increased or decreased and adjusted by combining the variable test condition of the actual mine. The directed acyclic graph constructed in this example is shown in FIG. 2.
And 2, dynamic Bayesian network modeling of the working face.
Step 2.1, dividing the state category of each node in the directed acyclic graph; the prominent risk nodes are difficult to divide by 2 types, meanwhile, the excessive division of the node types can cause the condition probability distribution dimension of the child nodes under the action of a plurality of nodes to be exponentially increased, and the calculation difficulty is increased, so that 2-4 optimal language terms are selected as an approximate form for measuring the variable properties, and different states of the nodes can be described according to the realization. The detailed division criteria are shown in table 1. The salient risks are divided into 3 categories, namely safety (L1), threat (L2) and salient risk (L3), and are used as reference bases for further pre-acquisition salient prevention decisions of each cycle.
Table 1 State classification of nodes
Step 2.2, continuously following 189.7m on the basis of the tunneling working face, dividing the distance of each mining cycle into 51 continuous time slices with the salient risk as a target node Y, and constructing time state transition probability. Since the mechanism of the salient risk evolution is not clear, the expert analysis determines the time transition probability, see table 2, to represent the effect of Y t-1 in the previous time slice on the target node Y t of the current time slice.
Table 2 highlights risk state transition probability partitioning
Step 2.3 for probability distribution parameter determination, the 14 nodes in the established directed acyclic graph are calculated based on a combination of parameter learning and expert decision analysis. And determining initial prior probability of the root node, conditional probability distribution parameters of the intermediate node and the target node, and establishing a dynamic Bayesian network salient prediction model.
Further, step 2.2 specifically includes:
Step 2.3.1, adopting a expectation maximization algorithm of maximum likelihood estimation, based on 475 groups of mining cycle test data sets (comprising X5, X7, X8 and X9) in the mine history samples, adopting an EM algorithm to ensure that the log likelihood of a new network is always greater than or equal to that of a previous network, obtaining CPT of a leaf node Y, and determining conditional probability distribution P (Y|X5, X7, X8 and X9) of a target node Y under the action of X5, X7, X8 and X9 nodes. Even if a small amount of partial data is missing, the result can be objectively calculated. The calculations are shown in table 3:
TABLE 3 conditional probability distribution results for leaf node Y in directed acyclic graphs based on parameter learning
Step 2.3.2 for root nodes (including X1, X2, X3, X4, X6, X10, X11, X12, X13), an initial prior probability is calculated with priority over historical incident statistics. According to the predicted mine with the working face in the example, 64 outstanding accidents occur in total according to the outstanding accident case statistics records, and then, by combining with the accident cause statistics rule analysis shown in fig. 3, the mining depth (X1) and the prior probability result of the geological structure (X11, X12X 13) can be determined by combining with the accident occurrence frequency, and the prior probability and conditional probability distribution of the rest nodes Xi (i=1, 2, …, 13) are determined by carrying out expert decision analysis flow in the step of ：P(X1＝1)＝0.3594;P(X1＝2)＝0.25;P(X1＝3)＝0.3125;P(X1＝4)＝0.0781;P(X11＝1)＝0.4844;P(X11＝2)＝0.5156;P(X12＝1)＝0.172;P(X12＝2)＝0.828;P(X13＝1)＝0.125;P(X13＝2)＝0.875; and 2.2.3 respectively. Firstly, dividing probability intervals to provide basis for expert decision analysis of variable occurrence probability. Combining with the triangle fuzzy number dividing criterion, the method has the advantage of solving the problem of ambiguity that each variable is difficult to be accurately divided into single categories; the difficulty of estimating the probability and the influence of decision uncertainty can be reduced. The probability interval division mode is shown in table 4. Wherein the i-th interval is defined as ai= (a i,mi,ui+1), 1.ltoreq.ai.ltoreq.7 and m i represents the mean.
TABLE 4 probability interval partitioning
Then, in order to avoid abnormal values or invalid data, 5 specialists with working experience for more than 10 years are selected to conduct questionnaire investigation. The expert confidence index theta is constructed, meanwhile, the influence of the expert self judging capability eta and the expert subjective confidence delta is considered, and the larger the eta is, the more believes the expert to be self judging capability. The larger δ represents the greater the reliability of the expert's judgment ability. The confidence level of each expert is calculated in such a way that θ=δ×η. As shown in table 5, δ takes the average of α and β.
TABLE 5 expert determination capability level
Each expert analyzes the possibility of the state division of each node based on the table to obtain a fuzzy possibility interval. For example, assuming P (y= 1|x 1＝1,x2 =2), it can be abbreviated asThe analysis by the j-th expert (j=1, …, 5) falls within the i-th interval (i=1, …, 5), i.e. a ij. Then the final ambiguity possibility interval/>, is calculated by equation (1)Θ j is the confidence index obtained by the j-th expert,/>Representing the data reliability of the jth expert.
In order to obtain accurate probability values, an alpha weighted estimation method (see formula (2)) is adopted to carry out fuzzy probability interval obtained from expert investigation resultsDefuzzification is performed. The method has the advantage that the information integrity after deblurring can be reserved to the maximum.For the exact values after deblurring, F α = { x|f (x) > α } represents the α level set and F (α) represents the α weighting function; average (F α) is the Average of the α level set, calculated by equation (3).
L α and u α represent the lower and upper boundaries, respectively, of the α -level set. As shown in fig. 4, the calculation can be performed by equation (4).
Typically, f (α) =1, α=0.5 is calculated from formula (5), respectively. Thus, the first and second substrates are bonded together,Calculated from (6)
To meet the normalization conditions, the method is further carried out by the formula (7)And (5) standardization. The operation is repeated, and the accurate probability distribution of the nodes can be obtained. t is the corresponding state of the node, if there are 3 states.
Taking P (x5=x|x10=3, x11=2, x12=2, x13=1) as an example, table 6 shows the final results obtained by detailed investigation analysis of each expert. The conditional probability distribution for each state combination is calculated in turn, taking X5 as an example, resulting in an overall conditional probability distribution, as shown in table 7.
Table 6 expert investigation results on P (x5=x|x10=3, x11=2, x12=2, x13=1)
/>
TABLE 7 conditional probability distribution of node X5 in directed acyclic graphs based on expert decision analysis
/>
And 2.3, constructing an initial polymorphic dynamic Bayesian network salient risk reasoning prediction model in a GeNIe Bayesian solver based on the established directed acyclic graph and the probability distribution parameters of each variable.
And 3, highlighting continuous dynamic probability reasoning analysis of risks in the tunneling process.
Further, the step 3 specifically includes:
And 3.1, when the working face prediction drilling test is not started, calculating an initial static probability result of the Y node by adopting a Bayesian network calculation principle in combination with the (8) according to initial probability distribution of each node, wherein the initial static probability result represents initial salient risk before the working face is adopted, and the nodes Xi (i=1, 2, …, n).
The result is shown in fig. 5, and it can be seen that the initial hidden danger of the node is large, the probability of L2 of the Y node is 0.43 at the maximum, and the potential outburst risk exists, so that the follow-up continuous tracking prediction is required due to the fact that the original gas of the tunneling surface is large, the coal quality is soft and the breakage is serious.
Along with the continuous test of drilling tests and geological parameters of continuous tunneling circulation, variable parameter information of X5, X7, X8 and X9 is continuously updated, and the probability distribution of the salient risk Y t of the current circulation can be obtained by adopting formulas (9) to (10) based on the probability distribution of the salient risk Y t-1 of the previous circulation and the constructed time state transition probability and combining the dynamic Bayesian network reasoning principle of the current circulation. Wherein for two adjacent time slices (loops) t-1 and t, the conditional probability distribution P (x t|xt-1) of any node x t of the t-th time slice (loop) is calculated using equation (9),I-th node representing i-th time slice,/>Representing the parent node of the ith node. Equation (10) may further calculate a joint probability distribution for any node in a dynamic bayesian network of time slices. B 0 represents the bayesian network at time t-1, and B → represents the bayesian network at time t. A connection is established between B → and B 0 by state transition probabilities. Fig. 6 shows the result of the dynamic bayesian inference probability distribution under 51 tunneling cycle information update, which shows that the probability of the salient risks of the Y node continuously fluctuates in the entire tunneling process of the working face.
To further verify the advantages of the dynamic bayesian network over the static bayesian network without considering the cycle-to-cycle correlation, the prediction results of the two methods are compared, see fig. 7. Fig. 7 (a) shows the result of the static bayesian calculation, and fig. 7 (b) shows the result of the calculation of the dynamic bayesian network. In general, in the initial tunneling stage of the working face, the state of the prominent risk is stable, the probability of L3 is close to 0, although the probability of the states of L1 and L2 shows fluctuation change, in the step (b), the high probability value of L1 is unchanged all the time, the effectiveness of gas extraction measures in the area before mining is indirectly reflected, the outburst prevention force of the working face for initial mining is on site, and mining lasting for the first 28 cycles is smooth. Compared with the method, the method has the advantages that the result of (a) is in 2,5-11 and 14-28 cycles, multiple potential risk misjudgments exist, the probability values of L1 and L2 are close, the uncertainty of risk judgments is increased, and the decision difficulty is increased for outburst prevention. As the development of the tunnel continues to progress late, the risk potential for protrusion increases, for example, during the 28-41 th cycle, in conjunction with site surveys, multiple coal seam thickening occurs in front of the face. During the tunneling to cycle 31, the drill hole is predicted to have a stuck event, and by further detecting the front coal body, the soft layer thickness is reduced but the coal thickness is not reduced, so that the partial gas accumulation can be caused. In the corresponding (a), from the 28 th cycle, the L2 gradually rises, the result is clear, and the effect of inter-cycle risk connection is reflected. However, in the (b), the problem that the probability values of the L1 and the L2 are close again occurs in the 28 th and 29 th loops, so that the prediction difficulty is increased. During the mining to cycle 40, a 1-trip drilling jet power development event was recorded. The L3 probability increases in the corresponding results of the two methods during the whole 39 th-41 th cycle, the L2 probability subsequently increases, and the alternate occurrence of the two states shows that the outstanding risk of the tunneling cycle stage needs to be considered, and the abnormal geological structure is not found in the actual field, so that the situation is presumed to be related to the fact that the local gas emission outburst prevention measures are not in place in the period. In conclusion, the two methods start from multivariate polymorphic comprehensive reasoning, so that the whole process result has reference value, and the feasibility of actual risk prediction is realized. Meanwhile, the outstanding prediction is carried out based on dynamic Bayesian reasoning, and because the relation between the front loop and the back loop is considered, the risk result is more stable and coherent, the accuracy is higher, the fluctuation of the static Bayesian reasoning among all risk states is larger, and the uncertainty of part of risk judgment still exists.
Step 3.2 reverse diagnostic analysis. The corresponding risk state probability of the target node is set to 1 in conjunction with the current actual prominent risk state, i.e., P (y=l1) =1, P (y=l2) =1, or P (y=l3) =1. Based on the calculation of the formula (11), the probability distribution of the other nodes Xi (i=1, 2, …, 13) can be obtained by reverse reasoning, and compared with the probability distribution before setting, the variables are sequentially ordered according to the probability change degree, so that the meaning of the key variable Y=L3 of the current risk state is determined:
Where P (xi=s|y=l3) (i= 5,7,8,9,10) represents the posterior probability of the state S (e.g. sj= { S1, S2, S3} = {1,2,3 }) corresponding to each node Xi in the presence of a significant risk. The probability differences P (xi=s1|y) -P (xi=sj|y) between the class of each node and the lowest class in the working face prediction process are counted, see fig. 8. For example, X5 (middle) represents P (x5=s1|y) -P (x5=s2|y), and X5 (high) represents P (x5=s1|y) -P (x5=s3|y). When the difference is smaller than 0, there is deterioration in the state of the representative variable Xi itself, so that it is likely to become a key variable, and the larger the difference, the larger the representative influence.
Step 3.3 sensitivity analysis. The sensitivity analysis method can identify the variables with larger variation amplitude and larger contribution to the fluctuation regulation and control of the risk results in the current tunneling cycle. In combination with the sensitivity analysis method, under the initial probability, the larger the SPM (X i) of the parent node is, the larger the influence on the node Y is by adopting the formula (12) to calculate. And when the observed state of the actual node isSPM (X i) of X i can be calculated by equation (13). L represents the state corresponding to leaf node Y, and x i represents the state corresponding to risk variable Xi.
The results of the sensitivity analysis were obtained before exploitation and at the cycle of maximum degree of risk prominence (t=39), as shown in fig. 9. The SPM (Xi) (i=5, 6, …, 10) values at the initial prior probability are small due to the initial lack of actual subsequent information importation, and the result is relatively balanced as shown in (a). Whereas at time t=39, in the known risk of protrusion (y=l3) state, the calculated SPM ordering is X9> X7> X8> X5> X10> X6, as shown in (b). It can be seen that the drill cuttings desorption index (X9), the drill cuttings amount (X7), the drill cuttings gas emission initial speed (X8) are three factors which are most sensitive to the circulation, and in actual operation, in order to verify the risk prediction result at the moment, the test range and the accuracy of the parameters can be sequentially re-optimized. The present risk can be completely excluded only when the predicted outcome generated by the above factors does not constitute a threat.
And 4, making a basis for anti-outburst decision-making in the tunneling cycle.
Based on the reverse diagnostic analysis in step 3.1, it is possible to determine the key variables in any working surface cycle that affect the current risk status and then take control in a targeted manner. As can be seen from fig. 7, when there is a significant threat in practice (y=l2), the variables that should be of great concern are the soft stratification thickness (X10), the drill gas flush initial velocity q (X8) and the cuttings desorption index Δh 2 (X9). When there is actually a significant risk (y=l3), the variables to be focused on are the cuttings desorption index Δh 2 (X9), the initial rate of borehole gas flooding q (X8), the soft stratification thickness (X10) and the geological formation complexity (X5). The difference in the results diagnosed for the two risk states is caused by the difference in the actual degree of prominent risk. When the drilling test indexes q and delta h 2 need to be focused, the drilling is encrypted on site, the test steps are repeated, network information is reintroduced, and the result is updated. If the risk still exists, local gas drainage measures are needed to be taken, so that the risk is reduced. Aiming at the influence of soft layering and geological structures, on one hand, the abnormal geological detection result in front of the working face is further focused, and on the other hand, necessary measures for coal seam permeability improvement and stress release are needed to be adopted, so that a favorable geological environment inducing protrusion is avoided. The diagnosis analysis can get rid of excessive dependence on the outburst prevention decision of field operators, and provides decision basis for controlling the outburst risk. The network structure and parameters constructed by the mines with different geological conditions are different, so that the diagnosis and analysis results are changed.
Based on the sensitivity analysis in the step 3.2, the variable with larger influence on the prediction result can be judged, and the higher the sensitivity of the variable is, the change of the representative variable needs to be continuously focused, so that the test precision is improved, and the salient risk is prevented from deteriorating. Fig. 9 shows that SPM (Xi) (i=5, 6, …, 10) values at the initial prior probability are small due to initial lack of actual subsequent information importation, and the result is relatively balanced, see (a). Whereas at time t=39, in the known risk of protrusion (y=l3) state, the calculated SPM ordering is X9> X7> X8> X5> X10> X6, see (b). It can be seen that the drill cuttings desorption index (X9), the drill cuttings amount (X7), the drill cuttings gas emission initial speed (X8) are three factors which are most sensitive in the circulating DBN, and in actual operation, in order to verify the risk prediction result at the moment, the test range and the accuracy of the parameters can be sequentially re-optimized. The present risk can be completely excluded only when the result predicted by the above-mentioned factors does not constitute a threat.
Claims (5)
1. A tunneling roadway coal and gas outburst prediction method based on dynamic probability reasoning is characterized by comprising the following steps:
step 1, analyzing outstanding risk variables of a tunneling roadway based on geological conditions of a mine where a working face is located, wherein the geological conditions comprise: gas occurrence state, conventional drilling test parameters, coal seam conditions and geological structures; the gas outburst risk variables of the tunneling roadway comprise: mining depth X1, gas pressure X2, gas content X3, damage type X4, geological structure complexity degree X5, firmness coefficient X6, drilling cuttings quantity X7, drilling gas emission initial speed X8, drilling cuttings desorption index X9, soft layering thickness X10, local coal thickness change X11, fault structure X12 and fold structure X13; wherein the regional prediction process variables and the working face prediction process variables are involved, the regional prediction process variables comprise X1, X2, X3, X4, X6 and X10; working face prediction process variables include X5, X7, X8 and X9; defining the degree of the prominent risk as a target variable Y, determining an inherent logic relationship between the prominent risk variables X1-X13 and the target variable Y, and establishing a directed acyclic graph in the Bayesian network, wherein the directed acyclic graph comprises 14 corresponding nodes;
step 2, dividing the state category of each node in the directed acyclic graph; dividing a tunneling working face circulation time slice, taking the prominent risk variables X1-X13 as root nodes and intermediate nodes, and taking the target variable Y as leaf nodes; establishing the connection of adjacent time slices, constructing time state transition probability, determining initial prior probability and conditional probability distribution parameters of each node, establishing a dynamic Bayesian network salient prediction model, and calculating the dynamic Bayesian network salient prediction model by adopting a mode of combining parameter learning and expert decision analysis;
Step 3, based on a Bayesian network calculation principle, when the working face prediction drilling test is not started, utilizing the initial probability distribution comprehensive reasoning of each root node of the Bayesian network to obtain the probability result of the leaf node, namely the probability distribution of the prominent risk degree of the initial working face; along with continuous test of drilling test and geological parameters of continuous tunneling circulation, continuously updating variable information of geological structure complexity degree X5, firmness coefficient X6, drilling cuttings quantity X7, drilling gas emission initial speed X8 and drilling cuttings desorption index X9, and combining probability distribution of the salient risk Yt-1 of the previous circulation and time state transition probability of construction to obtain probability distribution of the salient risk Yt of the current circulation; in the process of predicting any cycle, carrying out reverse diagnosis analysis and sensitivity analysis at any time;
Step 4, determining key variables influencing the current risk state in any working face cycle based on reverse diagnosis analysis, and then pertinently adopting control, judging variables with larger influence on a prediction result based on sensitivity analysis, wherein the higher the sensitivity of the variables is, the more frequent the variables need to pay attention to the changes, so that the testing precision is improved, and the salient risk is prevented from deteriorating;
The state categories of the nodes in the directed acyclic graph respectively comprise 2-4 state classifications, and the prominent risks are divided into 3 types: safety L1, threat L2 and outburst risk L3 are used as reference bases for further pre-acquisition outburst prevention decision of each cycle;
Defining the prominent risk variables X1, X2, X3, X4, X6, X10, X11, X12 and X13 as root nodes, and preferentially calculating initial prior probability according to the statistical frequency data of the corresponding variables in the historical accidents;
Defining the prominent risk variables X5, X7, X8 and X9 as intermediate nodes, adopting a expectation maximization algorithm of maximum likelihood estimation, and carrying out parameter learning through the historical mining cycle sample data of the mine where the prominent risk variables X5, X7, X8 and X9 are positioned to determine the conditional probability distribution P (Y|X5, X7, X8 and X9) of the target variable Y under the action of the intermediate nodes X5, X7, X8 and X9; the rest variables use expert decision analysis method;
Determining the rest target nodes Xi through probability interval division, expert investigation data collection and defuzzification in sequence, wherein i=1, 2, … and 13 are the prior probability and the conditional probability distribution; the probability interval division is combined with the triangle fuzzy number construction, so that the problem of ambiguity that the prominent risk variable is difficult to correspond to a single state is solved;
Leading in an expert confidence index theta to ensure that expert investigation results are more reliable, adopting an alpha weighted estimation method to deblur the expert investigation results, and maximally reserving the information integrity after deblurring;
And obtaining an initial prominent risk dynamic Bayesian network reasoning prediction model based on the established directed acyclic graph and the variable probability distribution parameters.
2. The tunneling roadway coal and gas outburst prediction method based on dynamic probability reasoning, according to claim 1, is characterized in that soft layering thickness, local coal thickness change, fault structure and buckling structure of a mine where a working face is located directly influence geological structure complexity based on an outburst mechanism; the mining depth is based on the relationship between stress and drill cuttings quantity; the gas pressure, the gas content, the damage type and the firmness coefficient are related to the initial gas emission speed of the drilling hole, the drill cuttings desorption index and the soft layering thickness; and constructing a directed acyclic graph by combining the logical relations of the variables, dividing root nodes, intermediate nodes and leaf nodes of the directed acyclic graph, and properly increasing or decreasing and adjusting the variables by combining the variable test conditions of actual mines.
3. The method for predicting coal and gas outburst in a tunneling roadway based on dynamic probabilistic reasoning according to claim 1, wherein risk variables X1 and X6 jointly influence a risk variable X7, risk variables X2, X3 and X6 jointly influence a risk variable X8, risk variables X2, X3 and X4 jointly influence a risk variable X9, risk variables X10, X11, X12 and X13 jointly influence a risk variable X5, and risk variables X5, X7, X8 and X9 jointly influence a target variable Y.
4. The method for predicting the coal and gas outburst of a tunneling roadway based on dynamic probabilistic reasoning as claimed in claim 1, wherein the specific steps of the reverse diagnosis and analysis are as follows:
setting the corresponding risk state probability of the target node Y to 1 in combination with the current actual prominent risk state, i.e., P (y=l1) =1, P (y=l2) =1 or P (y=l3) =1;
Based on the Bayesian network calculation principle, obtaining the inverse reasoning to obtain the probability distribution of each node Xi, wherein i=1, 2, … and 13, and comparing the probability distribution with the probability distribution before setting, and sequentially sequencing each variable according to the probability change degree, thereby determining the key variable of the current risk state.
5. The method for predicting the coal and gas outburst of a tunneling roadway based on dynamic probabilistic reasoning according to claim 1, wherein the specific steps of sensitivity analysis are as follows: and a sensitivity analysis method is adopted to identify the variable with large variation amplitude and large contribution to the fluctuation regulation and control of the risk result in the current tunneling cycle.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310710975.2A CN116882548B (en) | 2023-06-15 | 2023-06-15 | Tunneling roadway coal and gas outburst prediction method based on dynamic probability reasoning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310710975.2A CN116882548B (en) | 2023-06-15 | 2023-06-15 | Tunneling roadway coal and gas outburst prediction method based on dynamic probability reasoning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116882548A CN116882548A (en) | 2023-10-13 |
| CN116882548B true CN116882548B (en) | 2024-05-17 |
Family
ID=88268859
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310710975.2A Active CN116882548B (en) | 2023-06-15 | 2023-06-15 | Tunneling roadway coal and gas outburst prediction method based on dynamic probability reasoning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116882548B (en) |
Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106680451A (en) * | 2015-11-09 | 2017-05-17 | 河南理工大学 | Underground rapid measurement method for coal and gas outburst parameter as well as apparatus thereof |
| WO2017188858A1 (en) * | 2016-04-28 | 2017-11-02 | Schlumberger Canada Limited | Reservoir performance system |
| CN107392394A (en) * | 2017-08-20 | 2017-11-24 | 煤炭科学技术研究院有限公司 | A kind of dynamic monitoring driving face coal and gas prominent hazard prediction method |
| WO2018117890A1 (en) * | 2016-12-21 | 2018-06-28 | Schlumberger Technology Corporation | A method and a cognitive system for predicting a hydraulic fracture performance |
| CN110059963A (en) * | 2019-04-20 | 2019-07-26 | 北京交通大学 | A kind of tunnel risk evaluating method based on fuzzy polymorphism Bayesian network |
| CN111079978A (en) * | 2019-11-20 | 2020-04-28 | 辽宁工程技术大学 | Coal and gas outburst prediction method based on logistic regression and reinforcement learning |
| CN111275195A (en) * | 2020-02-13 | 2020-06-12 | 辽宁石油化工大学 | Dynamic Bayesian network modeling method based on coal gasification equipment |
| CN111311092A (en) * | 2020-02-13 | 2020-06-19 | 辽宁石油化工大学 | Coal gasification equipment dynamic risk assessment method |
| CN111582603A (en) * | 2020-05-19 | 2020-08-25 | 中煤科工集团重庆研究院有限公司 | Intelligent early warning method for coal and gas outburst based on multi-source information fusion |
| CN113689032A (en) * | 2021-08-09 | 2021-11-23 | 陕煤集团神木张家峁矿业有限公司 | Multi-sensor fusion gas concentration multi-step prediction method based on deep learning |
| CN115018334A (en) * | 2022-06-15 | 2022-09-06 | 湖南科技大学 | Fuzzy Bayesian network-based gas explosion risk assessment method |
| CN115492639A (en) * | 2022-09-20 | 2022-12-20 | 华北科技学院 | Coal roadway driving working face coal and gas outburst early warning method |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20110276514A1 (en) * | 2010-05-04 | 2011-11-10 | International Business Machines Corporation | Evaluating the quality and risk-robustness of an energy generation capacity resource plan under inherent uncertainties in energy markets and carbon regulatory regime |
| US10047679B2 (en) * | 2016-06-14 | 2018-08-14 | General Electric Company | System and method to enhance lean blowout monitoring |
| CN113536678B (en) * | 2021-07-19 | 2022-04-19 | 中国人民解放军国防科技大学 | XSS risk analysis method and device based on Bayesian network and STRIDE model |
| CN113689055B (en) * | 2021-10-22 | 2022-01-18 | 西南石油大学 | Oil-gas drilling machinery drilling speed prediction and optimization method based on Bayesian optimization |
-
2023
- 2023-06-15 CN CN202310710975.2A patent/CN116882548B/en active Active
Patent Citations (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106680451A (en) * | 2015-11-09 | 2017-05-17 | 河南理工大学 | Underground rapid measurement method for coal and gas outburst parameter as well as apparatus thereof |
| WO2017188858A1 (en) * | 2016-04-28 | 2017-11-02 | Schlumberger Canada Limited | Reservoir performance system |
| WO2018117890A1 (en) * | 2016-12-21 | 2018-06-28 | Schlumberger Technology Corporation | A method and a cognitive system for predicting a hydraulic fracture performance |
| CN107392394A (en) * | 2017-08-20 | 2017-11-24 | 煤炭科学技术研究院有限公司 | A kind of dynamic monitoring driving face coal and gas prominent hazard prediction method |
| CN110059963A (en) * | 2019-04-20 | 2019-07-26 | 北京交通大学 | A kind of tunnel risk evaluating method based on fuzzy polymorphism Bayesian network |
| CN111079978A (en) * | 2019-11-20 | 2020-04-28 | 辽宁工程技术大学 | Coal and gas outburst prediction method based on logistic regression and reinforcement learning |
| CN111275195A (en) * | 2020-02-13 | 2020-06-12 | 辽宁石油化工大学 | Dynamic Bayesian network modeling method based on coal gasification equipment |
| CN111311092A (en) * | 2020-02-13 | 2020-06-19 | 辽宁石油化工大学 | Coal gasification equipment dynamic risk assessment method |
| CN111582603A (en) * | 2020-05-19 | 2020-08-25 | 中煤科工集团重庆研究院有限公司 | Intelligent early warning method for coal and gas outburst based on multi-source information fusion |
| CN113689032A (en) * | 2021-08-09 | 2021-11-23 | 陕煤集团神木张家峁矿业有限公司 | Multi-sensor fusion gas concentration multi-step prediction method based on deep learning |
| CN115018334A (en) * | 2022-06-15 | 2022-09-06 | 湖南科技大学 | Fuzzy Bayesian network-based gas explosion risk assessment method |
| CN115492639A (en) * | 2022-09-20 | 2022-12-20 | 华北科技学院 | Coal roadway driving working face coal and gas outburst early warning method |
Non-Patent Citations (4)
| Title |
|---|
| Dynamic Bayesian network risk probability evolution for third-party damage of natural gas pipelines;Hong, Bingyuan 等;Applied Energy;20230301;第333卷;120620 * |
| Optimize the early warning time of coal and gas outburst by multi-source information fusion method during the tunneling process;Li B 等;PROCESS SAFETY AND ENVIRONMENTAL PROTECTION;20211231;839-849 * |
| 基于贝叶斯网络的煤与瓦斯突出预测研究;张克;汪云甲;;计算机工程与应用;20071011(第29期);220-221+248 * |
| 我国煤与瓦斯突出防治理论技术研究进展与展望;王恩元 等;煤炭学报;20220125;第47卷(第1期);297-322 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116882548A (en) | 2023-10-13 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Khademi Hamidi et al. | Application of fuzzy set theory to rock engineering classification systems: an illustration of the rock mass excavability index | |
| Razani et al. | A novel fuzzy inference system for predicting roof fall rate in underground coal mines | |
| US8204697B2 (en) | System and method for health assessment of downhole tools | |
| US20100042327A1 (en) | Bottom hole assembly configuration management | |
| KR101967978B1 (en) | Apparatus for predicting net penetration rate of shield tunnel boring machine and method thereof | |
| Saeidi et al. | Prediction of the rock mass diggability index by using fuzzy clustering-based, ANN and multiple regression methods | |
| Ebrahimi et al. | Estimation of shear wave velocity in an Iranian oil reservoir using machine learning methods | |
| CN110889440A (en) | Rockburst grade prediction method and system based on principal component analysis and BP neural network | |
| CN115017791A (en) | Tunnel surrounding rock grade identification method and device | |
| Zhang et al. | Cross-project prediction for rock mass using shuffled TBM big dataset and knowledge-based machine learning methods | |
| CN115115129A (en) | TBM construction speed prediction method based on weighted random forest | |
| CN116739176A (en) | Tunnel mechanized construction risk prediction method based on deep belief network | |
| CN113011094A (en) | Earth pressure balance shield machine muck improvement method based on mixed GBDT and random forest algorithm | |
| Bajolvand et al. | Optimization of controllable drilling parameters using a novel geomechanics-based workflow | |
| CN115481565A (en) | Earth pressure balance shield tunneling parameter prediction method based on LSTM and ant colony algorithm | |
| Qiu et al. | TBM tunnel surrounding rock classification method and real-time identification model based on tunneling performance | |
| CN113065188B (en) | Pile sinking process evaluation method based on machine learning, storage medium and electronic equipment | |
| Ma et al. | Landslide susceptibility assessment using the certainty factor and deep neural network | |
| CN116882548B (en) | Tunneling roadway coal and gas outburst prediction method based on dynamic probability reasoning | |
| CN104598705A (en) | Method and device for recognizing underground material layer | |
| CN117035465A (en) | Method and device for evaluating landslide susceptibility | |
| CN115358454A (en) | Coal and gas outburst prediction method based on extension-fuzzy hierarchical analysis theory | |
| Kovacevic et al. | The use of neural networks to develop CPT correlations for soils in northern Croatia | |
| Ruiz-Serna et al. | Combined artificial intelligence modeling for production forecast in a petroleum production field | |
| Ruiz et al. | Combined artificial intelligence modeling for production forecast in an oil field |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |