Identification of mine water inrush source based on PCA-AWOA-ELM model
-
摘要:
矿井突水是煤矿生产过程最具威胁的灾害之一,为了保障煤矿安全生产,提高矿井突水水源识别精度,提出一种基于改进鲸鱼优化算法(AWOA)耦合极限学习机的水源识别模型。以岱庄煤矿为例,选取Na+、Ca2+、Mg2+、Cl−、SO4 2−、HCO3 −作为判别指标,基于SPSS因子分析提取评价指标主成分,6种离子间相关性较大,Ca2+和Mg2+,Ca2+、Mg2+与SO4 2−、Cl−之间的相关性均达到了0.7以上,SO4 2−和Cl−之间的相关性也达0.68,通过主成分分析提取了3个主成分,从六维空间降低到三维空间,在减少了样本指标之间信息重复的同时,也减少了极限学习机输入层数量,提高了模型对各类型数据的泛化能力。其次,引入混沌动态权重因子和精英反向机制对鲸鱼算法进行改进,改进的鲸鱼优化算法克服了极限学习机权值阈值随机取值的缺点,混沌动态权重因子、精英反向机制的引入降低了模型计算复杂度,提高了算法精度,算法速度,跳出了局部寻优。通过训练38组样本数据,优化极限学习机的权值和阈值,最终构建PCA-AWOA-ELM水源识别模型,并对10组未知的测试样本进行预测。结果表明,PCA-AWOA-ELM模型的预测精度达100%,PCA-WOA-ELM模型精度为90%,PCA-ELM、ELM模型精度为60%,PCA-AWOA-ELM模型识别精度、运行速度、稳定性均明显高于PCA-WOA-ELM模型、PCA-ELM模型和ELM模型,为矿井安全生产提供了重要保障。
-
关键词:
- 矿井突水 /
- 水源识别 /
- 主成分分析(PCA) /
- 改进鲸鱼优化算法(AWOA) /
- 极限学习机(ELM)
Abstract:Mine water inrush is one of the most threatening disasters in the coal mine production process. In order to ensure safe coal mine production and improve the accuracy of mine water inrush source identification, a water source identification model based on improved whale optimization algorithm coupling extreme learning machine is proposed. Take Daizhuang Coal Mine as an example, choose Na+, Ca2+, Mg2+, Cl−, SO42−, HCO3− as the discrimination index, analyze and extract the main components of the evaluation index based on SPSS factor analysis. The correlation between the six ions is large, Ca2+ and Mg2+, Ca2+, Mg2+ and SO42−, Cl− has reached more than 0.7, and the correlation between SO42− and Cl− has also reached 0.68. Through principal component analysis, three principal components have been extracted, from six-dimensional space to three-dimensional space, while reducing the duplication of information between sample indicators. It also reduces the number of input layers of the limit learning machine and improves the generalization ability of the model for each type of data. Secondly, Secondly, the chaotic dynamic weight factor and the elite reverse mechanism are introduced to improve the whale algorithm. The improved whale optimization algorithm overcomes the disadvantage of random value of weight threshold of limit learning machine. The introduction of chaotic dynamic weight factor and elite reverse mechanism reduces the complexity of model calculation, improves the accuracy and speed of the algorithm, and jumps out of local optimization. Through training 38 groups of sample data, optimizing the weights and thresholds of the extreme learning machine, building a PCA-AWOA-ELM water source recognition model, and predicting 10 unknown test samples. The results show that the accuracy of the PCA-AWOA-ELM model is 100%, the accuracy of the PCA-WOA-ELM model is 90%, the accuracy of the PCA-ELM and ELM model is 60%, and the recognition accuracy, running speed and stability of the PCA-AWOA-ELM model are obvious. Higher than the PCA-WOA-ELM model, PCA-ELM model and ELM model, it provides an important guarantee for safe mine production.
-
0. 引 言
中国是以煤炭能源为主的国家,煤炭在相当长的时期内仍将是我国的主要能源,在历经近一个世纪的开采后,浅部的煤炭资源已开采殆尽,华北型煤田最为突出,大部分矿井已开采至石炭系下部煤层,煤矿受岩溶水威胁尤为严重[1]。据统计,2010—2019年,全国共发生煤矿水害事故182起,死亡839人,约是世界上主要采煤国家死亡总人数的4倍,直接经济损失数十亿元,仅次于瓦斯事故造成的损失[2]。由此可见,煤矿水害是煤矿安全生产中长期存在的、需要不断解决的实际问题。快速、准确识别矿井突水水源成为当前煤矿安全研究的重要课题。
目前,众多学者通过构建高效精准的水源判别模型来指导矿井安全生产。马雷等[3]构建基于GIS的利用水温识别突水水源的模型,该法较好地适用于含水层之间存在较大的地温差(埋深差)情况下的突水水源识别。陈建平[4]基于含水层环境同位素差异研究了隆德矿区煤层识别矿井涌水来源,较好地识别了矿井水源,并估算了混合比例。王心义等[5]建立了距离判别分析模型,针对河南焦作矿区和新安矿区进行突水水源识别,该模型简单有效,具有较强的判别能力,可解决水文地质条件相似矿区的突水水源的判定。陈红江[6]建立多组逐步Bayes判别模型,张春雷[7]建立Bayes多类线性判别模型对矿井水源进行识别,计算过程简单、模型结构稳定,克服了距离判别法的缺点,但结构受到样本的限制,且结果易出现分类模糊现象。Fisher水源判别模型[8-10]建立在有限的工程实例原始数据资料基础上,模型的判别精度受原始资料数据代表性、准确性的影响。神经网络技术[11-13]、支持向量机(SVM)[14-15]、可拓识别方法[16-17]等方法均已在矿井突水水源判别中得到了广泛应用。
笔者以岱庄煤矿为例,为克服ELM模型泛化能力不足,稳定性较差和过度拟合的问题,构建了主成分分析(Principal Component Analysis,PAC)、自适应鲸鱼优化算法(Ameliorative Whale Optimization Algorithm,AWOA)和极限学习机(Extreme Learning Machine,ELM))相耦合的矿井突水水源识别模型,并与PCA-WOA-ELM模型、WOA-ELM模型、ELM模型的判别结果进行对比,结果表明,PCA-AWOA-ELM模型识别精度、运行速度、稳定性均明显高于其他三者,为矿井安全生产提供了重要保障。
1. 理论与算法
1.1 改进的鲸鱼优化算法
WOA算法的收敛精度较低且收敛速度较慢,为了克服上述缺陷,引入AWOA算法对ELM模型进行优化[18]。
在鲸鱼探索阶段(图1),加入精英个体引导机制。
$$ X\left(t+1\right)={X}_{p}\left(t\right)-{\bf{dir}}·AD $$ (1) 式中:
$ {X}_{p}\left(t\right) $ 为第t次迭代时,种群的适应度最优时的个体位置;当前种群最优个体的适应度的值大于上一代种群最优个体适应度的值时,dir为种群搜索因子,·为向量点乘,$D = \left| {{{C}}X_{{\rm{rand}}}\left( t \right) - X\left( t \right)} \right|$ ,表示个体距离随机个体Xrand的长度;t表示当前迭代次数;A和C表示系数;Xrand表示种群中随机个体的位置向量;X(t)表示鲸鱼个体(解)的位置向量;${\bf{dir}}$ 即为单位向量,反之,${\bf{dir}}$ 向量每一维的值由下式可得:$$ {{\bf{dir}}}_{j}=\left\{ \begin{aligned} &-1,&{A}_{j}·\left[{X}_{p}^{j}\left(t\right)-{X}_{p}^{j}\left(t-1\right)\right] > 0\\ &1,&其他\end{aligned} \right. $$ (2) $$ {X}_{P}^{*}={X}_{\min}+r·\left({X}_{\max}-{X}_{\min}\right) $$ (3) $$ {X}_{P}^{**}={X}_{\max}+{X}_{\min}-{X}_{p} $$ (4) 式中:
$ {X}_{\max} $ 、$ {X}_{\min} $ 为搜索空间的下界以及上界;$ r $ 为随机向量。在鲸鱼气泡网攻击阶段,引入动态混沌权重因子及收敛因子
$ \lambda $ ,具体步骤如下:$$ \omega \left(t+1\right)=4\omega \left(t\right)\left(1-\omega \left(t\right)\right),t=1,\cdots ,{t}_{\max} $$ (5) 利用
$ \omega $ 及$ \lambda $ 进行鲸鱼位置优化的公式如下:$$ X\left(t+1\right)=\omega \left(t\right){\lambda X}_{{\rm{best}}}\left(t\right)-AD $$ (6) $$ \lambda =\left({t}_{\max}-t\right)/{t}_{\max} $$ (7) 式中:
$ \omega \left(t\right) $ 为区间[0,1]的随机数。1.2 极限学习机(ELM)理论
极限学习机训练模型如图2所示,ELM算法学习步骤[19-20]如下:
1)设定隐含层神经元个数。
2)选择无限可微的隐含层神经元的激活函数,通过激活函数可以将样本映射到另一个特征空间,通过映射得到矩阵
$ {\boldsymbol{H}} $ ,$ {\boldsymbol{H}} $ 表达式如下:$$ {\boldsymbol{H}}{\boldsymbol{\beta}} ={\boldsymbol{T}} $$ (8) 式中,
$ {\boldsymbol{H}} $ 为隐含层输出矩阵;$ {\boldsymbol{T}} $ 为网络的输出矩阵;$ {\boldsymbol{\beta}} $ 为隐含层节点与输出层节点的连接权值矩阵。3)在新的特征空间下,利用最小二乘法计算输出层最优输出权重
$ {{\boldsymbol{\beta}} }^{*} $ :$$ {{\boldsymbol{\beta}} }^{*}={{\boldsymbol{H}}}^+{\boldsymbol{T}} $$ (9) 式中,
$ {{\boldsymbol{H}}}^{+} $ 为$ {\boldsymbol{H}} $ 的Moore−Penrose广义逆矩阵。2. 实践应用
2.1 研究区概况
岱庄煤矿位于济宁市区正北约4.0 km处,北起17煤层露头,南到济宁市城市规划边界线,西起济宁断层与唐口煤矿相毗邻,东北部与何岗井田接壤,东南部以八里铺断层、许厂煤矿相邻。矿井自上而下主要含水层有第四系砂砾层孔隙含水层、山西组3煤层顶、底板砂岩裂隙含水层、三灰岩溶裂隙含水层、十下灰岩溶裂隙含水层、十三灰岩溶裂隙含水层、奥陶系石灰岩岩溶裂隙含水层,近年矿井涌水量如(图3)所示。
2.2 水质特征
基于48组实测数据,采用AqQA对井田内主要充水含水层(三灰含水层(Ⅰ)、十下灰含水层(Ⅱ)、十三灰含水层(Ⅲ),奥灰含水层(Ⅳ))进行水质分析(图4)。从图4中可以看出,三灰、十下灰、十三灰、奥灰的水质类型相近,按舒卡列夫分类,属SO4 −Ca·Mg·Na型水。
2.3 PCA-AWOA-ELM水源识别模型
选取岱庄煤矿三灰(Ⅰ)、十下灰(Ⅱ)、十三灰(Ⅲ)、奥灰(Ⅳ)共48组实测水样数据(38组训练样本,10组预测样本),以Na+、Ca2+、Mg2+、Cl−、SO4 2−、HCO3 −六种离子作为判别指标,建立PCA-AWOA-ELM模型对矿井突水水源进行判别,流程图如图5所示。训练水样实测数据见表1,预测水样实测数据见表2。
表 1 岱庄煤矿水样实测数据表(训练集)Table 1. Actual measurement data of water samples in Daizhuang Coal Mine (training set)编号 各实测离子质量浓度/(mg·L−1) 水样类型 Na+ Ca2+ Mg2+ Cl− SO4 2− HCO3 − 1 387.68 209.08 62.34 68.72 1 146.29 297.28 Ⅰ 2 430.48 163.35 46.76 67.40 1 098.30 312.93 Ⅰ 3 260.00 143.00 40.80 53.50 736.00 248.00 Ⅰ 4 440.00 9.62 0.49 56.70 563.00 266.00 Ⅰ 5 285.00 251.00 41.80 71.60 901.00 399.00 Ⅰ 6 76.15 668.63 122.83 97.86 1 879.16 230.79 Ⅱ 7 76.84 675.16 123.89 87.42 1 889.60 224.27 Ⅱ 8 89.85 668.63 122.83 97.86 1 879.16 230.79 Ⅱ 9 77.80 676.91 125.74 81.77 1 873.94 234.70 Ⅱ 10 73.95 666.89 116.49 86.12 1 842.62 224.27 Ⅱ 11 89.79 688.23 121.51 94.82 1 929.28 232.09 Ⅱ 12 84.38 679.96 121.78 86.55 1 983.56 228.18 Ⅱ 13 102.78 672.99 121.25 94.38 2 004.44 219.05 Ⅱ 14 665.36 126.90 65.40 70.60 1 683.90 426.40 Ⅱ 15 148.22 620.28 113.85 78.29 1 842.62 233.39 Ⅲ 16 228.00 589.00 97.20 81.20 1 890.00 294.00 Ⅲ 17 119.36 554.07 109.36 72.20 1 659.93 207.32 Ⅲ 18 213.46 329.31 81.62 79.61 1 299.76 129.08 Ⅲ 19 127.48 638.57 121.25 96.12 1 903.18 222.96 Ⅲ 20 117.80 563.65 115.17 86.55 1 837.40 69.11 Ⅲ 21 99.16 587.17 110.15 71.33 1 704.82 236.00 Ⅲ 22 183.49 580.64 111.47 82.20 1 999.22 241.22 Ⅲ 23 218.00 601.00 114.00 82.60 1 990.00 268.00 Ⅲ 24 367.00 505.00 97.20 87.20 1 980.00 259.00 Ⅲ 25 248.00 569.00 99.20 87.20 1 870.00 284.00 Ⅲ 26 157.00 341.00 57.80 69.50 1 050.00 264.00 Ⅲ 27 549.37 173.34 53.40 67.70 1 658.63 246.30 Ⅲ 28 554.73 173.45 55.62 68.20 1 674.80 247.87 Ⅲ 29 565.70 174.24 53.26 78.29 1 484.54 265.99 Ⅳ 30 477.75 274.42 60.76 68.72 1 557.62 292.07 Ⅳ 31 363.06 201.24 133.66 68.72 1 445.91 294.68 Ⅳ 32 546.79 203.42 61.81 83.94 1 545.09 247.74 Ⅳ 33 709.76 102.80 30.11 76.98 1 053.38 791.45 Ⅳ 34 515.43 173.80 54.68 83.51 1 367.61 267.29 Ⅳ 35 376.23 291.84 83.21 91.77 1 404.15 224.27 Ⅳ 36 581.73 189.48 51.51 81.77 1 520.03 259.47 Ⅳ 37 507.84 256.40 45.80 67.80 1 753.80 170.60 Ⅳ 38 544.99 188.50 56.70 65.20 1 826.70 126.30 Ⅳ 表 2 岱庄煤矿水样实测数据表(预测集)Table 2. Measured data of water samples in Daizhuang coal mine (prediction set)编号 各实测离子质量浓度/(mg·L−1) 水样类型 Na+ Ca2+ Mg2+ Cl− SO4 2− HCO3 − 1 410.00 119.42 60.49 55.70 568.00 273.00 Ⅰ 2 623.14 108.60 53.80 65.45 1 534.80 398.40 Ⅱ 3 598.77 116.80 52.60 48.60 1 453.90 379.20 Ⅱ 4 616.12 118.30 52.40 55.80 1 563.30 391.40 Ⅱ 5 285.00 251.00 41.80 71.60 901.00 399.00 Ⅲ 6 552.65 163.54 53.45 57.36 1 654.80 268.70 Ⅲ 7 577.69 157.50 50.40 64.40 1 673.60 272.30 Ⅲ 8 315.04 266.21 78.18 60.26 1 405.14 267.78 Ⅲ 9 416.27 264.20 55.40 56.80 1 585.40 216.80 Ⅳ 10 544.82 210.65 54.20 62.36 1 758.20 194.50 Ⅳ 首先,对训练样本实测数据进行主成分分析,笔者使用SPSS.26软件实现,根据相关性系数表(表3)知,6种离子间相关性较大,Ca2+和Mg2+,Ca2+、Mg2+与SO4 2−、Cl−之间的相关性均达到了0.7以上,SO4 2−和Cl−之间的相关性也达0.68,由此看出,样本指标之间的信息出现了较大的重复,对水源判别模型的精度造成影响,故对数据进行主成分分析是十分有必要的。
表 3 水化学成分指标相关系数矩阵Table 3. Correlation coefficient matrix of water chemical composition index相关性 Na+ Ca2+ Mg2+ Cl− SO4 2− HCO3 − Na+ 1 Ca2+ −0.882 1 Mg2+ −0.72 0.874 1 Cl- −0.657 0.823 0.748 1 SO42- −0.348 0.711 0.791 0.68 1 HCO3- 0.522 −0.469 −0.406 −0.358 −0.225 1 按照累计贡献率大于85%的原则,提取3个主成分(表4),主成分与标准化原始变量间相关关系如下:
表 4 主成分总方差解释Table 4. Explanation of total variance of principal components主成分 特征根 初始特征值方差/% 累计贡献率/% 1 4.167 69.447 69.447 2 0.918 15.299 84.746 3 0.514 8.574 93.32 4 0.276 4.598 97.918 5 0.107 1.786 99.703 6 0.018 0.297 100 $$ \left\{ \begin{aligned} &{Y}_{1}=-0.838{X}_{1}+0.971{X}_{2}+0.982{X}_{3}+0.873{X}_{4}+\\ &\qquad0.769{X}_{5}-0.554{X}_{6}\\ &{Y}_{2}=0.34{X}_{1}+0.16{X}_{3}+0.16{X}_{4}+0.504{X}_{5}+0.705{X}_{6}\\ &{Y}_{3}=0.402{X}_{1}-0.152{X}_{2}+0.033{X}_{3}-0.075{X}_{4}+\\ &\qquad0.358{X}_{5}-0.441{X}_{6}\end{aligned} \right. $$ (10) 设置模型最大迭代次数100,种群规模30,输入层节点数3,输出层节点数4,隐含层神经元个数38个,网络输出状态为4,将训练好的参数及模型对并测试集进行测试,结果如图6所示。
由图6寻优收敛能力曲线比较可知,通过自适应权重的鲸鱼优化算法在迭代次数不同的情况下,寻优能力,收敛速度和寻优精度都较鲸鱼优化算法强很多,且稳定性好,因此,在ELM模型中引入AWOA优化在水源判别模型中能够以快速且准确的优势取得较好的结果。
将PCA-AWOA-ELM模型预测结果与PCA-WOA-ELM模型、PCA-ELM模型和ELM模型进行对比(图7)可知,PCA-AWOA-ELM模型的预测精度达到了100%。由图7b和图7c对比可知,鲸鱼算法对极限学习机权值和阈值的优化成效很大,改进了极限学习机随机产生权值与阈值而导致的预测性能不稳定,预测结果正确率低,随机性太强的缺点。在图7a和图7b的对比中可知,精英引导机制及动态混沌因子的引入,加强了鲸鱼优化算法的寻优能力与精度,从而更好的优化ELM模型的泛化能力及预测精度。由图8可知,改进的鲸鱼优化算法在迭代100次的情况下迭代误差极小,且在迭代63次处迭代误差达0.01,由此可知改进的鲸鱼优化算法的精度明显提高。综上所知,AWOA-ELM模型在水源识别的应用上有着泛化能力强,稳定性高,预测精度高的优势。
3. 结 论
1)运用因子分析进行主成分提取,有效地减少了各指标间的信息重复和AWOA-ELM模型输入层的个数,提高了算法的运行速度。
2)改进的鲸鱼优化算法克服了极限学习机权值阈值随机取值的缺点,对权值和阈值进行全局寻优挖掘,提高了算法的收敛速度、精度。
3)将PCA-AWOA-ELM突水水源识别模型应用于岱庄煤矿,并与PCA-WOA-ELM模型、PCA-ELM模型、ELM模型预测结果进行对比,实践证明,该模型识别精度较高,为矿井安全生产提供了重要保障。
-
表 1 岱庄煤矿水样实测数据表(训练集)
Table 1 Actual measurement data of water samples in Daizhuang Coal Mine (training set)
编号 各实测离子质量浓度/(mg·L−1) 水样类型 Na+ Ca2+ Mg2+ Cl− SO4 2− HCO3 − 1 387.68 209.08 62.34 68.72 1 146.29 297.28 Ⅰ 2 430.48 163.35 46.76 67.40 1 098.30 312.93 Ⅰ 3 260.00 143.00 40.80 53.50 736.00 248.00 Ⅰ 4 440.00 9.62 0.49 56.70 563.00 266.00 Ⅰ 5 285.00 251.00 41.80 71.60 901.00 399.00 Ⅰ 6 76.15 668.63 122.83 97.86 1 879.16 230.79 Ⅱ 7 76.84 675.16 123.89 87.42 1 889.60 224.27 Ⅱ 8 89.85 668.63 122.83 97.86 1 879.16 230.79 Ⅱ 9 77.80 676.91 125.74 81.77 1 873.94 234.70 Ⅱ 10 73.95 666.89 116.49 86.12 1 842.62 224.27 Ⅱ 11 89.79 688.23 121.51 94.82 1 929.28 232.09 Ⅱ 12 84.38 679.96 121.78 86.55 1 983.56 228.18 Ⅱ 13 102.78 672.99 121.25 94.38 2 004.44 219.05 Ⅱ 14 665.36 126.90 65.40 70.60 1 683.90 426.40 Ⅱ 15 148.22 620.28 113.85 78.29 1 842.62 233.39 Ⅲ 16 228.00 589.00 97.20 81.20 1 890.00 294.00 Ⅲ 17 119.36 554.07 109.36 72.20 1 659.93 207.32 Ⅲ 18 213.46 329.31 81.62 79.61 1 299.76 129.08 Ⅲ 19 127.48 638.57 121.25 96.12 1 903.18 222.96 Ⅲ 20 117.80 563.65 115.17 86.55 1 837.40 69.11 Ⅲ 21 99.16 587.17 110.15 71.33 1 704.82 236.00 Ⅲ 22 183.49 580.64 111.47 82.20 1 999.22 241.22 Ⅲ 23 218.00 601.00 114.00 82.60 1 990.00 268.00 Ⅲ 24 367.00 505.00 97.20 87.20 1 980.00 259.00 Ⅲ 25 248.00 569.00 99.20 87.20 1 870.00 284.00 Ⅲ 26 157.00 341.00 57.80 69.50 1 050.00 264.00 Ⅲ 27 549.37 173.34 53.40 67.70 1 658.63 246.30 Ⅲ 28 554.73 173.45 55.62 68.20 1 674.80 247.87 Ⅲ 29 565.70 174.24 53.26 78.29 1 484.54 265.99 Ⅳ 30 477.75 274.42 60.76 68.72 1 557.62 292.07 Ⅳ 31 363.06 201.24 133.66 68.72 1 445.91 294.68 Ⅳ 32 546.79 203.42 61.81 83.94 1 545.09 247.74 Ⅳ 33 709.76 102.80 30.11 76.98 1 053.38 791.45 Ⅳ 34 515.43 173.80 54.68 83.51 1 367.61 267.29 Ⅳ 35 376.23 291.84 83.21 91.77 1 404.15 224.27 Ⅳ 36 581.73 189.48 51.51 81.77 1 520.03 259.47 Ⅳ 37 507.84 256.40 45.80 67.80 1 753.80 170.60 Ⅳ 38 544.99 188.50 56.70 65.20 1 826.70 126.30 Ⅳ 表 2 岱庄煤矿水样实测数据表(预测集)
Table 2 Measured data of water samples in Daizhuang coal mine (prediction set)
编号 各实测离子质量浓度/(mg·L−1) 水样类型 Na+ Ca2+ Mg2+ Cl− SO4 2− HCO3 − 1 410.00 119.42 60.49 55.70 568.00 273.00 Ⅰ 2 623.14 108.60 53.80 65.45 1 534.80 398.40 Ⅱ 3 598.77 116.80 52.60 48.60 1 453.90 379.20 Ⅱ 4 616.12 118.30 52.40 55.80 1 563.30 391.40 Ⅱ 5 285.00 251.00 41.80 71.60 901.00 399.00 Ⅲ 6 552.65 163.54 53.45 57.36 1 654.80 268.70 Ⅲ 7 577.69 157.50 50.40 64.40 1 673.60 272.30 Ⅲ 8 315.04 266.21 78.18 60.26 1 405.14 267.78 Ⅲ 9 416.27 264.20 55.40 56.80 1 585.40 216.80 Ⅳ 10 544.82 210.65 54.20 62.36 1 758.20 194.50 Ⅳ 表 3 水化学成分指标相关系数矩阵
Table 3 Correlation coefficient matrix of water chemical composition index
相关性 Na+ Ca2+ Mg2+ Cl− SO4 2− HCO3 − Na+ 1 Ca2+ −0.882 1 Mg2+ −0.72 0.874 1 Cl- −0.657 0.823 0.748 1 SO42- −0.348 0.711 0.791 0.68 1 HCO3- 0.522 −0.469 −0.406 −0.358 −0.225 1 表 4 主成分总方差解释
Table 4 Explanation of total variance of principal components
主成分 特征根 初始特征值方差/% 累计贡献率/% 1 4.167 69.447 69.447 2 0.918 15.299 84.746 3 0.514 8.574 93.32 4 0.276 4.598 97.918 5 0.107 1.786 99.703 6 0.018 0.297 100 -
[1] 施龙青. 底板突水机理研究综述[J]. 山东科技大学学报(自然科学版),2009,28(3):17−23. doi: 10.16452/j.cnki.sdkjzk.2009.03.005 SHI Longqing. Summary of Research on Mechanism of Water-inrush from Seam Floor[J]. Journal of Shandong University of Science and Technology(Natural Science),2009,28(3):17−23. doi: 10.16452/j.cnki.sdkjzk.2009.03.005
[2] 董书宁,王 皓,周振方. 我国煤矿水害防治工作现状及发展趋势[J]. 劳动保护,2020(8):58−60. doi: 10.3969/j.issn.1000-4335.2020.08.027 DONG Shuning,WANG Hao,ZHOU Zhenfang. Experts introduce domestic coal mine water hazard prevention work[J]. Labour Protection,2020(8):58−60. doi: 10.3969/j.issn.1000-4335.2020.08.027
[3] 马 雷,钱家忠,赵卫东. 基于GIS和水质水温的矿井突水水源快速判别[J]. 煤田地质与勘探,2014,42(2):49−53. doi: 10.3969/j.issn.1001-1986.2014.02.010 MA Lei,QIAN Jiazhong,ZHAO Weidong. An approach for quickly identifying water-inrush source of mine based on GIS and groundwater chemistry and temperature[J]. Coal Geology & Exploration,2014,42(2):49−53. doi: 10.3969/j.issn.1001-1986.2014.02.010
[4] 陈建平,潘光义,吴 丽,等. 基于环境同位素和水化学特征识别矿井涌水来源[J]. 环境化学,2018,37(6):1410−1420. CHEN Jianping,PAN Guangyi,WU Li,et al. Identifying the source of the groundwater based on the characteristics of environmental isotopes and water chemistry[J]. Environmental Chemistry,2018,37(6):1410−1420.
[5] 王心义,徐 涛,黄 丹. 距离判别法在相似矿区突水水源识别中的应用[J]. 煤炭学报,2011,36(8):1354−1358. doi: 10.13225/j.cnki.jccs.2011.08.028 WANG Xinyi,XU Tao,HUANG Dan. Application of distance discriminance in identifying water inrush resource in similar coalmine[J]. Journal of China Coal Society,2011,36(8):1354−1358. doi: 10.13225/j.cnki.jccs.2011.08.028
[6] 陈红江,李夕兵,刘爱华. 矿井突水水源判别的多组逐步Bayes判别方法研究[J]. 岩土力学,2009,30(12):3655−3659. doi: 10.3969/j.issn.1000-7598.2009.12.016 CHEN Hongjiang,LI Xibing,LIU Aihua. Studies of water source determination method of mine water inrush based on Bayes' multi-group stepwise discriminant analysis theory[J]. Rock and Soil Mechanics,2009,30(12):3655−3659. doi: 10.3969/j.issn.1000-7598.2009.12.016
[7] 张春雷,钱家忠,赵卫东,等. Bayes方法在矿井突水水源判别中的应用[J]. 煤田地质与勘探,2010,38(4):34−37. doi: 10.3969/j.issn.1001-1986.2010.04.008 ZHANG Chunlei,QIAN Jiazhong,ZHAO Weidong,et al. The application of bayesian approach to discrimination of mine water-inrush source[J]. Coal Geology & Exploration,2010,38(4):34−37. doi: 10.3969/j.issn.1001-1986.2010.04.008
[8] 鲁金涛,李夕兵,宫凤强,等. 基于主成分分析与Fisher判别分析法的矿井突水水源识别方法[J]. 中国安全科学学报,2012,22(7):109−115. doi: 10.3969/j.issn.1003-3033.2012.07.018 LU Jintao,LI Xibing,GONG Fengqiang,et al. Recognizing of mine water inrush sources based on principal components analysis and fisher discrimination analysis method[J]. China Safety Science Journal,2012,22(7):109−115. doi: 10.3969/j.issn.1003-3033.2012.07.018
[9] 陈红江,李夕兵,刘爱华,等. 用Fisher判别法确定矿井突水水源[J]. 中南大学学报(自然科学版),2009,40(4):1114−1120. CHEN Hongjiang,LI Xibing,LIU Aihua,et al. Identifying of mine water inrush sources by Fisher discriminant analysis method[J]. Journal of Central South University(Science and Technology),2009,40(4):1114−1120.
[10] 刘 鑫,陈陆望,林曼利,等. 采动影响下矿井突水水源Fisher判别与地下水补给关系反演[J]. 水文地质工程地质,2013,40(4):36−43. doi: 10.16030/j.cnki.issn.1000-3665.2013.04.021 LIU Xin,CHEN Luwang,LIN Manli,et al. Fisher discriminant analysis for coal mining inrush water source under mining-induced disturbance and inversion of groundwater recharge relation[J]. Hydrogeology & Engineering Geology,2013,40(4):36−43. doi: 10.16030/j.cnki.issn.1000-3665.2013.04.021
[11] 李垣志,牛国庆,刘慧玲. 改进的GA-BP神经网络在矿井突水水源判别中的应用[J]. 中国安全生产科学技术,2016,12(7):77−81. doi: 10.11731/j.issn.1673-193x.2016.07.014 LI Yuanzhi,NIU Guoqing,LIU Huiling. Application of improved GA- BP neural network on identification of water inrush source in mine[J]. Journal of Safety Science and Technology,2016,12(7):77−81. doi: 10.11731/j.issn.1673-193x.2016.07.014
[12] 钱家忠,吕 纯,赵卫东,等. Elman与BP神经网络在矿井水源判别中的应用[J]. 系统工程理论与实践,2010,30(1):145−150. QIAN Jiazhong,LYU Chun,ZHAO Weidong,et al. Comparison of application on Elman and BP neural networks in discriminating water bursting source of coal mine[J]. Systems Engineering-Theory & Practice,2010,30(1):145−150.
[13] 吴 岩,余智超. 神经网络在矿井突水水源判别中的应用[J]. 工矿自动化,2011,37(10):60−62. WU Yan,YU Zhichao. Application of neural network in water source distinguishing of mine water inrush[J]. Industry and Mine Automation,2011,37(10):60−62.
[14] 邵良杉,李相辰. 基于MIV-PSO-SVM模型的矿井突水水源识别[J]. 煤炭科学技术,2018,46(8):183−190. doi: 10.13199/j.cnki.cst.2018.08.030 SHAO Liangshan,LI Xiangchen. Indentification of mine water inrush source based on MIV-PSO-SVM[J]. Coal Science and Technology,2018,46(8):183−190. doi: 10.13199/j.cnki.cst.2018.08.030
[15] 邵良杉,李印超,徐 波. 矿井突水水源识别的RS-LSSVM模型[J]. 安全与环境学报,2017,17(5):1730−1734. SHAO Liangshan,LI Yinchao,XU Bo. RS-LSSVM model for identifying and determinating the mining water inrush origin[J]. Journal of Safety and Environment,2017,17(5):1730−1734.
[16] 张瑞钢,钱家忠,马 雷,等. 可拓识别方法在矿井突水水源判别中的应用[J]. 煤炭学报,2009,34(1):33−38. doi: 10.3321/j.issn:0253-9993.2009.01.007 ZHANG Ruigang,QIAN Jiazhong,MA Lei,et al. Application of extension identification method in mine water-bursting source discrimination[J]. Journal of China Coal Society,2009,34(1):33−38. doi: 10.3321/j.issn:0253-9993.2009.01.007
[17] 冯 琳,李凤莲,张雪英,等. 改进可拓识别方法及其在突水水源判别中的应用[J]. 工矿自动化,2015,41(2):61−65. doi: 10.13272/j.issn.1671-251x.2015.02.017 FENG Lin,LI Fenglian,ZHANG Xueying,et al. Improved extension identification method and its application to water bursting source discrimination[J]. Industry and Mine Automation,2015,41(2):61−65. doi: 10.13272/j.issn.1671-251x.2015.02.017
[18] 张广炎. 基于鲸鱼算法优化极限学习机的热电偶非线性补偿方法[D]. 湘潭: 湘潭大学, 2019. ZHANG Guangyan. Nonlinear compensation method of thermocouple based on extreme learning machine optimized by whale algorithm[D]. Xiangtan: Xiangtan University, 2019.
[19] 任海峰, 严由吉, 吴青海. 基于SAPSO-ELM的瓦斯涌出量分源预测及应用[J/OL]. 煤田地质与勘探, 2021, 49(2): 102−109. REN Haifeng, YAN Youji, WU Qinghai. SAPSO-ELM gas emission prediction model and application based on split-source prediction[J/OL]. Coal Geology & Exploration, 2021, 49(2): 102−109.
[20] 董东林,陈昱吟,倪林根,等. 基于WOA-ELM算法的矿井突水水源快速判别模型[J]. 煤炭学报,2021,46(3):984−993. doi: 10.13225/j.cnki.jccs.2020.1827 DONG Donglin,CHEN Yuyin,NI Lingen,et al. Fast discriminant model of mine water inrush source based on WOA-ELM algorithm[J]. Journal of China Coal Society,2021,46(3):984−993. doi: 10.13225/j.cnki.jccs.2020.1827
-
期刊类型引用(5)
1. 李双慧,朱宏军,朱开鹏,黄选明. 不连沟井田岩溶水水化学特征及其演化规律研究. 能源与环保. 2024(02): 155-162 . 百度学术
2. 王皓,孙钧青,曾一凡,尚宏波,王甜甜,乔伟. 蒙陕接壤区煤层顶板涌水水源智能判别方法. 煤田地质与勘探. 2024(04): 76-88 . 百度学术
3. 肖观红,鲁海峰. 基于PCA-GA-RF的矿井突水水源快速识别模型. 煤矿安全. 2024(06): 184-191 . 百度学术
4. 周全超,刘玲,李继升,殷裁云. 基于多种解析法和数值法求解巨厚砂岩水文地质参数. 煤炭科学技术. 2024(S1): 174-182 . 本站查看
5. 曹思文,刘源,武亚遵,王雪奇. 基于多元统计和混合模型的矿井涌水构成比例判别. 煤炭工程. 2024(12): 133-139 . 百度学术
其他类型引用(1)