CN100405359C

CN100405359C - Prediction method of mine gas emission

Info

Publication number: CN100405359C
Application number: CNB2006100439620A
Authority: CN
Inventors: 刘韬; 刘亚娟; 王致杰
Original assignee: Individual
Current assignee: Liu Tao; Suzhou Vocational University
Priority date: 2006-05-17
Filing date: 2006-05-17
Publication date: 2008-07-23
Anticipated expiration: 2026-05-17
Also published as: CN1845091A

Abstract

本发明公开了一种矿井瓦斯涌出量的预测方法，其特征是，它是通过以下步骤进行预测的：构建基于距离浓度的免疫算法模型；将从数据库中获取埋藏深度、煤层深度、瓦斯含量、开采强度、邻近层距离、邻近层瓦斯含量和瓦斯涌出量的样本数据进行处理，并存入数据库中；利用基于距离浓度的免疫算法模型对数据库中各特征属性进行聚类，建立瓦斯涌出量样本数据聚类类别表；利用基于距离浓度的免疫算法模型对瓦斯涌出量样本数据聚类类别表进行挖掘，得出关联规则；将实测到的近期数据按上述关联规则预测出该工作面的瓦斯涌出量。The invention discloses a method for predicting mine gas emission, which is characterized in that it is predicted through the following steps: constructing an immune algorithm model based on distance concentration; obtaining burial depth, coal seam depth, and gas content from a database , Mining intensity, adjacent layer distance, adjacent layer gas content, and gas emission sample data are processed and stored in the database; the characteristic attributes in the database are clustered using the distance-concentration-based immune algorithm model to establish a gas influx The clustering category table of the output sample data; use the immune algorithm model based on the distance concentration to mine the clustering category table of the gas emission sample data to obtain the association rules; predict the recent data according to the above association rules to predict the work The amount of gas emitted from the surface.

Description

Prediction method of mine gas emission

技术领域 technical field

本发明涉及煤矿井下瓦斯涌出量的预测方法。The invention relates to a method for predicting gas gushing out in coal mines.

背景技术 Background technique

在煤矿开采过程中，瓦斯灾害事故频繁，瓦斯爆炸等重特大事故也时有发生。据统计，煤矿安全事故中，瓦斯爆炸事故是经济损失重大、人员伤亡最多的事故，也是造成社会影响最大的重特大事故。尤其是随着开采深度的不断增加，机械化程度的不断提高，开采强度的不断增强，瓦斯涌出量还会进一步增大，瓦斯灾害的治理越来越成为煤矿灾害防治的重点。影响瓦斯涌出量主要信息包括：埋藏深度、煤层深度、瓦斯含量、开采强度、邻近层距离、邻近层瓦斯含量等，这些数据有些是通过实测得到，有些是可以计算得到。然而在采掘过程中由于工作面处于动态变化中，而影响瓦斯涌出量的地质条件、煤层瓦斯含量和开采强度等因素处于不断变化之中，这使工作面瓦斯涌出量存在着非常大的不确定性，因此需要一种科学的方法来预测未来瓦斯涌出量，为煤矿企业高层领导制定决策提供依据。In the process of coal mining, gas disasters and accidents are frequent, and major accidents such as gas explosions also occur from time to time. According to statistics, among coal mine safety accidents, gas explosion accidents are the accidents with the largest economic losses and the most casualties, and also the major accidents that cause the greatest social impact. Especially with the continuous increase of mining depth, the continuous improvement of the degree of mechanization, and the continuous enhancement of mining intensity, the amount of gas emission will further increase, and the management of gas disasters has increasingly become the focus of coal mine disaster prevention and control. The main information affecting gas emission includes: burial depth, coal seam depth, gas content, mining intensity, distance from adjacent layers, gas content in adjacent layers, etc. Some of these data are obtained through actual measurement, and some can be calculated. However, during the mining process, due to the dynamic change of the working face, the geological conditions affecting the gas emission, the gas content of the coal seam, and the mining intensity are constantly changing, which makes the gas emission of the working face very large. Uncertainty, so a scientific method is needed to predict the amount of gas gushing in the future, so as to provide a basis for the decision-making of senior leaders of coal mine enterprises.

目前，许多煤矿建立了煤矿决策支持系统(DSS)，它包括数据库、模型库和知识库，功能是综合利用大量数据，有机组合众多模型(数学模型和数据处理模型)，通过人机交互，辅助各级决策者实现科学的决策。数据大多数是来源于煤矿企业经过长期积累的数据，数据庞大，以不同形式存在，可能在数据收集中还会造成数据失真或破坏数据完整性。这些数据库只提供了对数据的简单查询，不能有效地提取和利用有用的信息。At present, many coal mines have established a coal mine decision support system (DSS), which includes a database, a model base and a knowledge base. Decision makers at all levels realize scientific decision-making. Most of the data comes from the data accumulated by coal mining enterprises over a long period of time. The data is huge and exists in different forms, which may cause data distortion or destroy data integrity during data collection. These databases only provide simple query of data, and cannot effectively extract and utilize useful information.

数据挖掘技术是近年来非常活跃的研究领域，为人们及时准确地从庞大的数据库中获取信息提供有效的方法，特别是利用基于人工免疫的数据挖掘技术取代手工分析方法，能充分利用煤矿决策支持系统(DSS)的数据源，从大量的数据中挖掘出矿井瓦斯信息进行预测。Data mining technology is a very active research field in recent years. It provides an effective method for people to obtain information from huge databases in a timely and accurate manner. In particular, the use of data mining technology based on artificial immunity to replace manual analysis methods can make full use of coal mine decision support. The data source of the system (DSS) mines mine gas information from a large amount of data for prediction.

发明内容 Contents of the invention

本发明的目的提供一种利用人工免疫的数据挖掘技术预测矿井瓦斯涌出量的方法，以便煤矿决策者及时采取相应措施。The object of the present invention is to provide a method for predicting the amount of gas gushing out of mines using artificial immune data mining technology, so that coal mine decision makers can take corresponding measures in time.

本发明是利用原有的DSS决策支持系统上增加了数据挖掘模块进行计算和处理，来达到发明目的的。The present invention uses the original DSS decision support system to add a data mining module for calculation and processing to achieve the purpose of the invention.

本发明的数据挖掘模块是基于距离浓度的免疫算法建立的算法模型。The data mining module of the present invention is an algorithm model established based on the immune algorithm of distance concentration.

本发明具体预测过程如下：The concrete prediction process of the present invention is as follows:

1、构建基于距离浓度的免疫算法模型，算法描述如下：1. Build an immune algorithm model based on distance concentration. The algorithm is described as follows:

第一步：设解决的问题为X。将抗原和抗体分别对应于待求解的问题X和问题的一个解x_i，f(X_i)为解的适应度函数。Step 1: Let the problem to be solved be X. The antigen and the antibody correspond to the problem X to be solved and a solution _xi of the problem respectively, and f(X _i ) is the fitness function of the solution.

第二步：产生初始抗体群体。随即产生N个初始抗体，再随即产生M个抗体放入记忆库中，从记忆库中提出M个抗体加入抗体群中，构成(N+M)个初始抗体群体。Step 2: Generate an initial antibody population. Immediately generate N initial antibodies, and then immediately generate M antibodies and put them into the memory bank, and put M antibodies from the memory bank into the antibody population to form (N+M) initial antibody populations.

第三步：对初始抗体群体中各抗体进行评价，按照公式(1)、(2)、(3)计算抗体v的浓度C(x_i)；然后按照公式(4)计算抗体的期望繁殖率E(x_i)；Step 3: Evaluate each antibody in the initial antibody population, calculate the concentration C(xi ₎ of antibody v according to formulas (1), (2), and (3); then calculate the expected reproductive rate of the antibody according to formula (4) E(x _i );

${d d}_{i i} = = d d (({x x}_{i i})) = = {Σ Σ}_{j j = = 11 i i &NotEqual; &NotEqual; j j}^{n no} | | | | {x x}_{i i} - - {x x}_{j j} | | | | - - - - - - ((11))$

$d d = = {Σ Σ}_{i i = = 11}^{n no} d d (({x x}_{i i})) = = {Σ Σ}_{i i = = 11}^{n no} {Σ Σ}_{j j = = 11,, j j &NotEqual; &NotEqual; i i}^{n no} | | | | {x x}_{i i} - - {x x}_{j j} | | | | - - - - - - ((22))$

$C C (({x x}_{i i})) = = 11 - - \frac{{d d}_{i i}}{d d} = = 11 - - \frac{{Σ Σ}_{j j = = 11,, i i &NotEqual; &NotEqual; j j}^{n no} | | | | {x x}_{i i} - - {x x}_{j j} | | | |}{{Σ Σ}_{i i = = 11}^{n no} {Σ Σ}_{j j = = 11,, i i &NotEqual; &NotEqual; j j}^{n no} | | | | {x x}_{i i} - - {x x}_{j j} | | | |} - - - - - - ((33))$

显然，抗体之间的距离越大，其距离浓度越小，反之则浓度越大；Obviously, the larger the distance between antibodies, the smaller the concentration of the distance, and vice versa, the larger the concentration;

$E E. (({x x}_{i i})) = = \frac{D D. (({x x}_{i i}))}{C C (({x x}_{i i}))} - - - - - - ((44))$

式中：d_i为抗体x_i在集合X上的距离；d为所有抗体之间的距离之和；D(x_i)为抗体与抗原之间的亲和度；C(x_i)为抗体在抗体(解)空间的浓度值。In the formula: d _i is the distance of antibody _xi on the set X; d is the sum of the distances between all antibodies; D(xi ₎ is the affinity between antibody and antigen; C(xi ₎ is the antibody Concentration values in antibody (solution) space.

第四步：形成父代群体。将初始抗体群体按E(x_i)的降序排列，并取前N个个体构成父代群体；同时提前M个个体作为记忆细胞存入记忆库中。Step 4: Form a parent group. Arrange the initial antibody population in descending order of E( _xi ), and take the first N individuals to form the parent population; at the same time, M individuals in advance are stored in the memory bank as memory cells.

第五步：判断是否满足结束条件。设定最大运行世代数作为终止条件，或设定连续运行一定世代数后函数值没有变化作为终止条件。一旦条件满足则结束运算。否则继续下一步操作。Step 5: Determine whether the end condition is met. Set the maximum number of running generations as the termination condition, or set that the function value does not change after a certain number of generations of continuous operation as the termination condition. Once the condition is satisfied, the operation ends. Otherwise, continue to the next step.

第六步：抗体的增殖和分化。基于第四步的计算结果，按照抗体评价标准，将从初始抗体群体中确定的父代群体进行抗体克隆，选择实数交叉与非均匀变异方法克隆新的抗体。克隆出的抗体与原有的抗体一起构成新一代抗体群。Step 6: Proliferation and differentiation of antibodies. Based on the calculation results of the fourth step, according to the antibody evaluation standard, antibody cloning will be carried out from the parent population determined from the initial antibody population, and new antibodies will be cloned by selecting real number crossover and heterogeneous mutation methods. The cloned antibodies together with the original antibodies constitute a new generation of antibody groups.

第七步：将第六步中构成的所述新一代抗体群代替第三步、第四步中的初始抗体群体，继续执行第三步、第四步、第五步。Step 7: Replace the initial antibody population in Step 3 and Step 4 with the new-generation antibody population formed in Step 6, and continue to perform Step 3, Step 4, and Step 5.

2、从数据库中获取埋藏深度、煤层深度、瓦斯含量、开采强度、邻近层距离、邻近层瓦斯含量和瓦斯涌出量的样本数据构成一个数据集，并将数据集中的数据进行清洗和归一化处理，清洗去除不一致的数据，将清洗和归一化处理后的数据按属性值按比例缩放，使它们都落入[0，1]上，建立[0，1]数据集。2. Obtain the sample data of burial depth, coal seam depth, gas content, mining intensity, adjacent layer distance, adjacent layer gas content and gas emission from the database to form a data set, and clean and normalize the data in the data set Processing, cleaning to remove inconsistent data, scale the cleaned and normalized data according to the attribute value, so that they all fall into [0, 1], and establish a [0, 1] data set.

2、利用基于距离浓度的免疫算法模型对[0，1]数据集中各特征属性进行聚类，得到他们优化的特征属性聚类个数，再将[0，1]数据集中的每个数据属性值分别划分到各自特征属性对应的类中，建立瓦斯涌出量样本数据聚类类别表。2. Use the immune algorithm model based on distance concentration to cluster each feature attribute in the [0, 1] data set to obtain the number of their optimized feature attribute clusters, and then cluster each data attribute in the [0, 1] data set The values are divided into the corresponding classes of their respective characteristic attributes, and the gas emission sample data clustering table is established.

3、利用基于距离浓度的免疫算法模型对瓦斯涌出量样本数据聚类类别表进行挖掘，得出关联规则。设煤层藏深度为A₁、煤层深度为A₂、瓦斯含量为A₃、开采强度为A₄、邻近层距离为A₅、邻近层瓦斯含量为A₆、瓦斯涌出量为A₇，则关联规则为：3. Using the immune algorithm model based on distance concentration to mine the clustering category table of gas emission sample data, and obtain association rules. Suppose the depth of coal seam is A ₁ , the depth of coal seam is A ₂ , the gas content is A ₃ , the mining intensity is A ₄ , the distance between adjacent layers is A ₅ , the gas content in adjacent layers is A ₆ , and the gas emission is A ₇ , then The association rules are:

其中X＝{A₁，A₂，A₃，A₄，A₅，A₆}，Y＝{A₇}，S为支持度，表示满足条件X所占的百分比，C为置信度，表示满足条件X又满足条件Y的概率，S＝P(X∪Y)，C＝(Y/X)。

Among them, X={A ₁ , A ₂ , A ₃ , A ₄ , A ₅ , A ₆ }, Y={A ₇ }, S is the support degree, which means the percentage of X that satisfies the condition, and C is the confidence level, which means The probability of satisfying condition X and satisfying condition Y, S=P(X∪Y), C=(Y/X).

5、将实测到的近期数据按上述关联规则预测出采掘现场的瓦斯涌出量。5. Predict the amount of gas emission at the mining site based on the recently measured data according to the above association rules.

本发明将人工免疫数据挖掘方法引入煤矿决策支持系统(DSS)，建立了基于免疫原理的挖掘系统，充分发挥了综合业务信息集成的数据优势，为预测瓦斯涌出量提供了有力的决策依据。The invention introduces the artificial immune data mining method into the coal mine decision support system (DSS), establishes a mining system based on the immune principle, fully exerts the data advantages of comprehensive business information integration, and provides a powerful decision-making basis for predicting the amount of gas gushing out.

具体实施方式 Detailed ways

下面结合某矿的情况对本发明的技术方案作进一步的描述。The technical scheme of the present invention will be further described below in conjunction with the situation of a certain mine.

某矿现已建立了公开使用的DSS决策支持系统，在该系统的数据库中储存了以前大量的瓦斯信息数据，数据包括煤层埋藏深度、煤层深度、瓦斯含量、开采强度、邻近层距离、邻近层瓦斯含量和瓦斯涌出量。A certain mine has established a publicly available DSS decision support system. A large amount of gas information data has been stored in the database of the system. The data include coal seam burial depth, coal seam depth, gas content, mining intensity, adjacent layer distance, adjacent layer Gas content and gas emission.

实施步骤如下：The implementation steps are as follows:

1、首先构建基于距离浓度的免疫算法模型，构建方法如发明内容中预测过程步骤1第一步至第七步方法所述，在此不再重复；1. First construct an immune algorithm model based on distance concentration. The construction method is as described in the first step to the seventh step of the prediction process step 1 in the summary of the invention, and will not be repeated here;

2、从DSS数据库中获取煤层埋藏深度、煤层深度、瓦斯含量、开采强度、邻近层距离、邻近层瓦斯含量和瓦斯涌出量的样本数据，清洗去除不一致数据；为了防止具有较大值的属性相对于较小值的属性权重过大，将数据进行归一化处理；再将上述归一化处理的数据按比例缩放，使它们都落入[0，1]上，建立瓦斯涌出量数据集。现将数据集中的20个记录举列(见附表1)；2. Obtain the sample data of coal seam burial depth, coal seam depth, gas content, mining intensity, adjacent layer distance, adjacent layer gas content and gas emission from the DSS database, and clean and remove inconsistent data; in order to prevent attributes with large values The attribute weight is too large relative to the smaller value, and the data is normalized; then the above-mentioned normalized data is scaled, so that they all fall on [0, 1], and the gas emission data is established set. Now list the 20 records in the data set (see attached table 1);

3、从瓦斯涌出量数据集中随机抽取400个记录，用第1步建立的算法模型，分别对埋藏深度、煤层厚度、瓦斯含量、开采强度、邻近层间距、邻近层瓦斯等特征属性进行聚类，得到它们优化的特征属性聚类个数(见附表2)。3. Randomly select 400 records from the gas emission data set, and use the algorithm model established in the first step to aggregate the characteristic attributes such as burial depth, coal seam thickness, gas content, mining intensity, adjacent layer distance, and adjacent layer gas. Classes, get the number of clusters of their optimized feature attributes (see attached table 2).

由此将数据集中个各记录的属性值分别划分到相应的类。例如附表2中将煤层厚度被分为5类(类1-5)，其中类4的区间为[0.8650，0.9365]、类1的区间为[0.4550，0.6522]，如果煤层厚度的值为0.8929，则它应该归入类4。对于附表1中的记录，经过特征属性的聚类，形成各记录属性的类别(见附表3)，其中瓦斯涌出量设为不突现、一般、突现三种情况；In this way, the attribute values of each record in the data set are divided into corresponding classes. For example, the coal seam thickness is divided into 5 categories (category 1-5) in the attached table 2, where the interval of category 4 is [0.8650, 0.9365], and the interval of category 1 is [0.4550, 0.6522]. If the value of coal seam thickness is 0.8929 , then it should be classified as class 4. For the records in Attached Table 1, after the clustering of characteristic attributes, the category of each record attribute is formed (see Attached Table 3), in which the amount of gas emission is set to three situations: non-emergent, general, and emergent;

4、采用基于人工免疫原理的关联挖掘方法，对瓦斯涌出量数据集进行挖掘，求出关联规则，设要求解的关联规则为：“如果X是A，则Y是B”，其中，X＝{煤层埋藏深度，煤层厚度，瓦斯含量，开采强度，邻近层间距，邻近层瓦斯含量}，Y＝{瓦斯涌出量}，则A＝{A₁，A₂，A₃，A₄，A₅，A₆}，B＝{A₇}，A₁，A₂，A₃，A₄，A₅，A₆，A₇分别表示煤层埋藏深度、煤层煤层厚度、瓦斯含量、开采强度、邻近层间距、邻近层瓦斯含量、瓦斯涌出量的属性值等级。挖掘后，在给定最小支持度为0.021，最小置信度为0.41情况下，获得瓦斯涌出量关联规则如下：4. Use the association mining method based on the principle of artificial immunity to mine the gas emission data set and find the association rules. Suppose the association rules to be solved are: "If X is A, then Y is B", where X ={burial depth of coal seam, thickness of coal seam, gas content, mining intensity, distance between adjacent layers, gas content of adjacent layers}, Y={gas emission amount}, then A={A ₁ , A ₂ , A ₃ , A ₄ , A ₅ , A ₆ }, B={A ₇ }, A ₁ , A ₂ , A ₃ , A ₄ , A ₅ , A ₆ , and A ₇ represent coal seam burial depth, coal seam thickness, gas content, mining intensity, Adjacent layer spacing, adjacent layer gas content, attribute value grades of gas emission. After excavation, given the minimum support degree of 0.021 and the minimum confidence degree of 0.41, the association rules for gas emission are obtained as follows:

①煤层埋藏深度＝3∧煤层厚度＝3∧瓦斯含量＝4∧开采强度＝5∧邻近层间距＝4∧邻近层瓦斯含量＝4→瓦斯涌出量＝3(突现)；① Coal seam burial depth = 3 ∧ coal seam thickness = 3 ∧ gas content = 4 ∧ mining intensity = 5 ∧ distance between adjacent layers = 4 ∧ gas content in adjacent layers = 4 → gas emission = 3 (emergence);

②煤层埋藏深度＝2∧煤层厚度＝3∧瓦斯含量＝2∧开采强度＝3∧邻近层间距＝2∧邻近层瓦斯含量＝1→瓦斯涌出量＝1(不突现)。② Coal seam burial depth = 2∧ coal seam thickness = 3∧ gas content = 2∧ mining intensity = 3∧ distance between adjacent layers = 2∧ gas content in adjacent layers = 1 → gas emission = 1 (not emergent).

5、实测计算某一采掘现场工作面的煤层埋藏深度，煤层厚度，瓦斯含量，开采强度，邻近层间距，邻近层瓦斯含量，按照第四步得出的关联规则对该工作面瓦斯涌出量进行预测，是突现还是不突现。5. Measure and calculate the coal seam burial depth, coal seam thickness, gas content, mining intensity, distance between adjacent layers, and gas content of adjacent layers in a working face of a mining site, and the gas emission amount of the working face according to the association rules obtained in the fourth step Make predictions, emergent or not.

表1 瓦斯涌出量数据集Table 1 Gas emission data set

表2 瓦斯涌出量特征属性优化聚类个数Table 2 Number of clusters optimized for gas emission characteristic attributes

表3 瓦斯涌出量样本数据聚类类别Table 3 Clustering categories of gas emission sample data

Claims

1. the Forecasting Methodology of a [underground is characterized in that, it is predicted by following steps:

(1) structure is based on the immune algorithm model of distance concentration, and arthmetic statement is as follows:

The first step: the problem of establishing solution is X: antigen and antibody are corresponded respectively to problem X to be found the solution and problem one separate x _i, f (x _i) be the fitness function of separating;

Second step: produce initial antibodies colony: produce N initial antibodies immediately, produce M antibody more immediately and put into data base, from data base, propose M antibody and add in the antibody population, constitute (N+M) individual initial antibodies colony;

The 3rd step: each antibody in the initial antibodies colony is estimated: according to formula 1., 2., the 3. concentration C (x of calculating antibody v _i), then according to the 4. expectation breeding potential E (x of calculating antibody of formula _i);

d_{i} = d (x_{i}) = Σ_{j = 1, i &NotEqual; j}^{n} | | x_{i} - x_{j} | |

①

d = Σ_{i = 1}^{n} d (x_{i}) = Σ_{i = 1}^{n} Σ_{j = 1, j &NotEqual; i}^{n} | | x_{i} - x_{j} | |

②

C (x_{i}) = 1 - \frac{d_{i}}{d} = 1 - \frac{Σ_{j = 1, i &NotEqual; j}^{n} | | x_{i} - x_{j} | |}{Σ_{i = 1}^{n} Σ_{j = 1, i &NotEqual; j}^{n} | | x_{i} - x_{j} | |}

③

Obviously, the distance between the antibody is big more, and its distance concentration is more little, otherwise then concentration is big more;

E (x_{i}) = \frac{D (x_{i})}{C (x_{i})}

④

In the formula:

d _iBe antibody x _iIn the distance of set on the X, d be between all antibody apart from sum, D (x _i) be the affinity between antibody and the antigen, C (x _i) be the concentration value of antibody in the antibody space.

The 4th step: form parent colony: E (x is pressed in initial antibodies colony _i) descending sort, and get that top n is individual to constitute parent colony, M individuality deposits in the data base as memory cell in advance simultaneously;

The 5th step: judge whether to satisfy termination condition: set maximum operation generation number as end condition, or functional value does not change as end condition after setting continuously the certain generation number of operation,, otherwise continue next step operation in case condition satisfies then finishes computing;

The 6th step: the propagation of antibody and differentiation: based on the result of calculation in the 4th step, according to the antibody evaluation criterion, the parent colony that will determine from initial antibodies colony carries out antibody cloning, select real number to intersect and the new antibody of non-uniform mutation method clone, the antibody that clones constitutes antibody population of new generation with original antibody;

The 7th goes on foot: the antibody population described of new generation that constitutes in going on foot the 6th replaces the initial antibodies colony in the 3rd step, the 4th step, the 3rd step of continuation execution, the 4th step, the 5th step;

(2) sample data of obtaining depth of burial, the coal seam degree of depth, gas bearing capacity, mining rate, adjacent layer distance, the next layer gas content and gas emission from database constitutes a data set, and the data of data centralization are cleaned and normalized, clean and remove inconsistent data, data after cleaning and the normalized are pressed the property value bi-directional scaling, make them all fall into [0,1] on, sets up [0,1] data set;

(3) utilize immune algorithm model based on distance concentration to [0,1] each characteristic attribute of data centralization carries out cluster, the characteristic attribute cluster number that is optimized, again with [0,1] each data attribute value of data centralization is divided into respectively in the class of characteristic attribute correspondence separately, sets up gas emission sample data cluster classification table;

(4) utilize the immune algorithm model based on distance concentration that gas emission sample data cluster classification table is excavated, draw correlation rule, establishing Tibetan, the coal seam degree of depth is A ₁, the coal seam degree of depth is A ₂, gas bearing capacity is A ₃, mining rate is A ₄, adjacent layer distance is A ₅, the next layer gas content is A ₆, gas emission is A ₇, then correlation rule is:

X={A wherein ₁, A ₂, A ₃, A ₄, A ₅, A ₆, Y={A ₇, S is a support, the expression shared number percent of X that satisfies condition, and C is a degree of confidence, expression satisfy condition the again probability of Y of X that satisfies condition, S=P (X ∪ Y), C=(Y/X);

(5) the recent data that will survey dope the gas emission at digging scene by above-mentioned correlation rule.