Applying feature extraction of acoustic emission and machine learning for coal failure forecasting
-
摘要: 同步采集了煤樣單軸壓縮破壞過程的聲發射全波形數據和應力數據，提取了聲發射梅爾倒譜系數作為樣本特征，定義煤樣當前受力與其峰值載荷的比值為煤樣的應力狀態并將其作為樣本標簽，利用機器學習方法構建了煤樣破壞狀態的預測模型。結果表明：梅爾倒譜系數可以很好地表征煤樣的破壞狀態，該參量在煤樣達到受力峰值80%后表現出明顯突增或突降或先增加然后突降的規律，機器學習能夠利用該樣本特征建立煤樣破壞狀態預測模型進而預測煤樣的危險狀態，利用五折交叉驗證方法評價模型的預測準確度達到88.61%，模型預測效果和穩定性良好；進一步討論了不同重要度的梅爾倒譜系數組合作為樣本特征對于模型預測效果的影響，發現樣本特征中含有重要度高的特征和關鍵特征是模型預測準確度高的關鍵。這可為進一步完善煤巖動力災害預測預警提供借鑒。Abstract: Recently, with increasing mining scale, intensity, and depth, the geological and mining conditions in coal mines are becoming more complicated; therefore, it has resulted in a more difficult situation of coal mine dynamic hazards, including rockburst, coal and gas outburst etc. Dynamic hazards are now posing a serious threat to the safety of coal mining. The precise forecasting of dynamic hazards is significant to their effective control. The acoustic emission (AE) monitoring technique is an effective geophysical monitoring and early warning method which can effectively reveal the characteristics and laws of coal and rock failure under loading. It has been successfully applied in the laboratory and engineering fields. To deeply analyze the characteristics of AE signals in the process of coal-rock damage and failure, thus, to help realize the precise monitoring and early warning of coal mine dynamic hazards, this study first conducted a uniaxial compression test on coal samples in the laboratory, and at the meantime, synchronously collected the full waveform data of AE and the loading data in the entire process of coal failure. Subsequently, using the feature extraction technique in the field of automatic speech recognition, this study extracted the Mel-frequency cepstral coefficient (MFCC) of AE and used it as the sample feature; the stress state of the coal sample was defined as the ratio of the current load the sample bore to its peak load and was employed as the sample label; a model for coal failure state forecasting was established by adopting machine learning methodology. Finally, the model’s forecasting accuracy was evaluated using the five-fold cross-validation method; the influence of different MFCC combinations as sample features on the forecasting accuracy of the model was discussed. The results show that MFCC can well characterize the failure state of coal samples. This parameter behaves in regular variation with increasing loading and shows the law of an obvious sudden increase or sudden decrease or increase followed by a sudden decrease when the loading exceeds 80% of the coal sample’s peak load. The established model can be well used to forecast coal failure state. The accuracy (ACC), true positive rate (TPR), true negative rate (TNR), and area under the curve (AUC) of the model forecasting reach 88.61%, 72.34%, 93.16%, and 0.93, respectively. Machine learning methodology can fully use MFCC features of AE and can identify essential sample features that are difficult to identify with the human eyes. Significant and key features included in the samples are the keys to the high forecasting accuracy of the model. TPR, TNR, and AUC of the model forecasting would be significantly influenced if crucial features were excluded from the samples. Adding features with low importance to the samples has little influence on the forecasting result of the model. This study’s results can provide a reference for further improving the prediction and early warning of coal and rock dynamic hazards.
-
圖 2 煤樣破壞狀態預測模型構建流程圖
Figure 2. Flow chart for the building of coal failure forecasting model
圖 3 實驗系統示意圖及聲發射探頭布置圖
Figure 3. Schematic diagram of experimental system and layout of acoustic emission sensors
圖 4 4-1煤樣MFCC與應力隨時間的變化. (a) MFCC-1；(b) MFCC-2；(c) MFCC-3；(d) MFCC-4；(e) MFCC-5；(f) MFCC-6；(g) MFCC-7；(h) MFCC-8；(i) MFCC-9；(j) MFCC-10；(k) MFCC-11；(l) MFCC-12
Figure 4. Variation of MFCC and stress of coal sample No. 4-1 with increasing time: (a) MFCC-1; (b) MFCC-2; (c) MFCC-3; (d) MFCC-4; (e) MFCC-5; (f) MFCC-6; (g) MFCC-7; (h) MFCC-8; (i) MFCC-9; (j) MFCC-10; (k) MFCC-11; (l) MFCC-12
圖 5 不同煤樣在不同加載速率下MFCC-1的變化. (a) No. 1-2, 3 μm·s–1；(b) No. 2-1, 7 μm·s–1；(c) No. 3-1, 11 μm·s–1；(d) No. 4-2, 15 μm·s–1；(e) No. 5-1, 20 μm·s–1
Figure 5. Variation of MFCC-1 of different coal samples under different loading rates: (a) No. 1-2, 3 μm·s–1; (b) No. 2-1, 7 μm·s–1; (c) No. 3-1, 11 μm·s–1; (d) No. 4-2, 15 μm·s–1; (e) No. 5-1, 20 μm·s–1
圖 6 不同煤樣在不同加載速率下MFCC-2的變化. (a) No. 1-2, 3 μm·s–1；(b) No. 2-1, 7 μm·s–1；(c) No. 3-1, 11 μm·s–1；(d) No. 4-2, 15 μm·s–1；(e) No. 5-1, 20 μm·s–1
Figure 6. Variation of MFCC-2 of different coal samples under different loading rates: (a) No. 1-2, 3 μm·s–1; (b) No. 2-1, 7 μm·s–1; (c) No. 3-1, 11 μm·s–1; (d) No. 4-2, 15 μm·s–1; (e) No. 5-1, 20 μm·s–1
圖 8 煤樣破壞狀態預測模型的MFCC特征重要度
Figure 8. Importance of each MFCC parameter of the forecasting model
圖 9 不同樣本特征下煤樣破壞狀態預測模型的ACC、TNR、TPR和AUC
Figure 9. ACC, TNR, TPR, and AUC of the forecasting model under different MFCC combinations
表 1 Light GBM算法參數設置
Table 1. Parameter setting of Light GBM algorithm
Parameter Value Role Learning_rate 0.1 Control model training speed Objective Binary Assigned learning tasks n_estimators 20 Control the number of training sessions Early_stopping_rounds 50 Control the maximum number of training sessions Max_depth 3 Set the depth of the decision tree Num_leaves 8 Adjust the complexity of the decision tree Subsample 0.8 Control the sampling ratio per tree Colsample_bytree 0.8 Control the proportion of features used per tree Min_child_weight 2 Control the weights of the minimum leaf nodes Reg_alpha 0 Prevent overfitting Reg_lambda 1 Prevent overfitting Verbose 100 Output model results 
下載: 導出CSV
表 2 煤樣聲發射樣本統計
Table 2. Statistics of acoustic emission samples of each coal sample
Coal samples Number of acoustic emission fragments Safety sample Dangerous sample Total number of samples 1-1 11032 3365 14397 1-2 11762 2066 13828 2-1 4173 1211 5394 2-2 4013 725 4738 3-1 2728 813 3541 3-2 2014 602 2616 4-1 1362 687 2049 4-2 2178 486 2664 5-1 1347 344 1691 5-2 1277 305 1582 Total 41886 10604 52490
下載: 導出CSV
表 4 煤樣加載實驗方案
Table 4. Coal sample loading scheme
Coal samples Sample size Sample number Loading rate/(μm·s?1) Kuangou Mine ?50 mm×100 mm 1-1，1-2 3 2-1，2-2 7 3-1，3-2 11 4-1，4-2 15 5-1，5-2 20
下載: 導出CSV
表 5 五折交叉驗證方法分組情況
Table 5. Grouping of the five-fold cross validation
Fold number Training set Validation set Fold-1 1-2, 2-1, 2-2, 3-1, 4-1, 4-2, 5-1, 5-2 1-1, 3-2 Fold-2 1-1, 2-1, 2-2, 3-1, 3-2, 4-2, 5-1, 5-2 1-2, 4-1 Fold-3 1-1, 1-2, 2-2, 3-1, 3-2, 4-1, 5-1, 5-2 2-1, 4-2 Fold-4 1-1, 1-2, 2-1, 3-1, 3-2, 4-1, 4-2, 5-2 2-2, 5-1 Fold-5 1-1, 1-2, 2-1, 2-2, 3-2, 4-1, 4-2, 5-1 3-1, 5-2
下載: 導出CSV
表 6 基于五折交叉驗證的模型ACC、TPR、TNR和AUC結果
Table 6. Results showing the ACC, TPR, TNR, and AUC of the forecasting based on five-fold cross validation
Fold number TN FP FN TP ACC/% TPR/% TNR/% AUC Fold-1 10213 223 1193 1980 89.60 62.40 97.86 0.97 Fold-2 10224 273 198 2006 96.29 91.02 97.40 0.99 Fold-3 4777 303 484 883 87.79 64.59 94.04 0.90 Fold-4 3668 354 499 621 83.41 55.45 91.20 0.88 Fold-5 2732 471 105 789 85.94 88.26 85.30 0.92 Average value 88.61 72.34 93.16 0.93 Note: TN is true negative and means the number of negative predicted as negative classes; TP is true positive and means the number of positive predicted as positive classes; FN is false negative and means the number of positive predicted as negative classes; FP is false positive and means the number of negative predicted as positive classes; ACC is the accuracy; TPR is the true positive rate; TNR is the true negative rate; AUC is the area under the ROC curve.
下載: 導出CSV
表 7 23種MFCC特征組合
Table 7. 23 MFCC feature combinations
Serial number MFCC combination Serial number MFCC combination 1 [2] 13 [12,3,1,10,8,11,4,7,5,9,6] 2 [2,12] 14 [3,1,10,8,11,4,7,5,9,6] 3 [2,12,3] 15 [1,10,8,11,4,7,5,9,6] 4 [2,12,3,1] 16 [10,8,11,4,7,5,9,6] 5 [2,12,3,1,10] 17 [8,11,4,7,5,9,6] 6 [2,12,3,1,10,8] 18 [11,4,7,5,9,6] 7 [2,12,3,1,10,8,11] 19 [4,7,5,9,6] 8 [2,12,3,1,10,8,11,4] 20 [7,5,9,6] 9 [2,12,3,1,10,8,11,4,7] 21 [5,9,6] 10 [2,12,3,1,10,8,11,4,7,5] 22 [9,6] 11 [2,12,3,1,10,8,11,4,7,5,9] 23 [6] 12 [2,12,3,1,10,8,11,4,7,5,9,6]
下載: 導出CSV
www.77susu.com<span id="fpn9h"><noframes id="fpn9h"><span id="fpn9h"></span> <span id="fpn9h"><noframes id="fpn9h"> <th id="fpn9h"></th> <strike id="fpn9h"><noframes id="fpn9h"><strike id="fpn9h"></strike> <th id="fpn9h"><noframes id="fpn9h"> <span id="fpn9h"><video id="fpn9h"></video></span> <ruby id="fpn9h"></ruby> <strike id="fpn9h"><noframes id="fpn9h"><span id="fpn9h"></span> -
參考文獻
[1] Li C H, Bu L, Wei X M, et al. Current status and future trends of deep mining safety mechanism and disaster prevention and control. Chin J Eng, 2017, 39(8): 1129李長洪, 卜磊, 魏曉明, 等. 深部開采安全機理及災害防控現狀與態勢分析. 工程科學學報, 2017, 39(8):1129 [2] Li X B, Gong F Q. Research progress and prospect of deep mining rock mechanics based on coupled static-dynamic loading testing. J China Coal Soc, 2021, 46(3): 846李夕兵, 宮鳳強. 基于動靜組合加載力學試驗的深部開采巖石力學研究進展與展望. 煤炭學報, 2021, 46(3):846 [3] Liu Z Q, Song Z Y, Ji H G, et al. Construction mode and key technology of mining shaft engineering for deep mineral resources. J China Coal Soc, 2021, 46(3): 826劉志強, 宋朝陽, 紀洪廣, 等. 深部礦產資源開采礦井建設模式及其關鍵技術. 煤炭學報, 2021, 46(3):826 [4] Qi Q X, Pan Y S, Shu L Y, et al. Theory and technical framework of prevention and control with different sources in multi-scales for coal and rock dynamic disasters in deep mining of coal mines. J China Coal Soc, 2018, 43(7): 1801齊慶新, 潘一山, 舒龍勇, 等. 煤礦深部開采煤巖動力災害多尺度分源防控理論與技術架構. 煤炭學報, 2018, 43(7):1801 [5] He X Q, Dou L M, Mu Z L, et al. Continuous monitoring and warning theory and technology of rock burst dynamic disaster of coal. J China Coal Soc, 2014, 39(8): 1485何學秋, 竇林名, 牟宗龍, 等. 煤巖沖擊動力災害連續監測預警理論與技術. 煤炭學報, 2014, 39(8):1485 [6] Li Z L, He S Q, Song D Z, et al. Microseismic temporal-spatial precursory characteristics and early warning method of rockburst in steeply inclined and extremely thick coal seam. Energies, 2021, 14(4): 1186 doi: 10.3390/en14041186 [7] Lou Q, He X Q, Song D Z, et al. Time-frequency characteristics of acoustic-electric signals induced by coal fracture under uniaxial compression based on full-waveform. Chin J Eng, 2019, 41(7): 874婁全, 何學秋, 宋大釗, 等. 基于全波形的煤樣單軸壓縮破壞聲電時頻特征. 工程科學學報, 2019, 41(7):874 [8] Guo J Y, Zhang Y Z. Analysis on acoustic emission characteristics of coal under uniaxial compression. Coal Technol, 2021, 40(4): 129郭敬遠, 張玉柱. 煤單軸壓縮破壞過程聲發射特征分析. 煤炭技術, 2021, 40(4):129 [9] Liu J H, Zhao L, Song S M, et al. Ultrasonic velocity and acoustic emission properties of concrete eroded by sulfate and its damage mechanism. Chin J Eng, 2016, 38(8): 1075劉娟紅, 趙力, 宋少民, 等. 混凝土硫酸鹽腐蝕損傷的聲波與聲發射變化特征及機理. 工程科學學報, 2016, 38(8):1075 [10] Zhang J, Chai M Y, Xiang J H, et al. Fatigue damage evaluation of 316LN stainless steel using acoustic emission monitoring. Chin J Eng, 2018, 40(4): 461張進, 柴孟瑜, 項靖海, 等. 基于聲發射監測的316LN不銹鋼的疲勞損傷評價. 工程科學學報, 2018, 40(4):461 [11] Jin P J, Wang E Y, Song D Z. Study on correlation of acoustic emission and plastic strain based on coal-rock damage theory. Geomech Eng, 2017, 12(4): 627 doi: 10.12989/gae.2017.12.4.627 [12] Deng X B, Liu Y Z, Xing K, et al. Analysis based on AE space-time evolution characteristics for stage division of whole stress-strain curve of rock. Chin J Rock Mech Eng, 2018, 37(Suppl 2): 4086鄧緒彪, 劉遠征, 邢礦, 等. 基于聲發射時空演化的巖石全應力-應變曲線階段特征分析. 巖石力學與工程學報, 2018, 37(Suppl 2): 4086 [13] Ren J X, Jing S, Zhang K. Study on failure mechanism and acoustic emission characteristics of outburst proneness coal rock under dynamic and static loading. Coal Sci Technol, 2021, 49(3): 57任建喜, 景帥, 張琨. 沖擊傾向性煤巖動靜載下破壞機理及聲發射特性研究. 煤炭科學技術, 2021, 49(3):57 [14] Ji H G, Mu N N, Zhang Y Z. Analysis on precursory characteristics of coupled acoustic emission and pressure for rock burst events. J China Coal Soc, 2013, 38(Suppl 1): 1紀洪廣, 穆楠楠, 張月征. 沖擊地壓事件AE與壓力耦合前兆特征分析. 煤炭學報, 2013, 38(Suppl 1): 1 [15] Chelali F Z, Djeradi A. Text dependant speaker recognition using MFCC, LPC and DWT. Int J Speech Technol, 2017, 20(3): 725 doi: 10.1007/s10772-017-9441-1 [16] Deshwal D, Sangwan P, Kumar D. Feature extraction methods in language identification: A survey. Wireless Pers Commun, 2019, 107(4): 2071 doi: 10.1007/s11277-019-06373-3 [17] Mei Q P, Gül M, Boay M. Indirect health monitoring of bridges using Mel-frequency cepstral coefficients and principal component analysis. Mech Syst Signal Process, 2019, 119: 523 doi: 10.1016/j.ymssp.2018.10.006 [18] Jiang Y, Yu M J, Zhang M Q. Spark discharge identification of electrostatic dust removal based on the back-propagation neural network. Inf Control, 2019, 48(6): 754江鶯, 俞銘津, 張夢琦. 基于BP神經網絡的電除塵火花放電識別. 信息與控制, 2019, 48(6):754 [19] Wang H L, Song D Z, Li Z L, et al. Acoustic emission characteristics of coal failure using automatic speech recognition methodology analysis. Int J Rock Mech Min Sci, 2020, 136: 104472 doi: 10.1016/j.ijrmms.2020.104472 [20] Pei Y Y, Yang X B, Chuan J P, et al. Time series prediction of microseismic energy level based on feature extraction of onedimensional convolutional neural network. Chin J Eng, 2021, 43(7): 1003裴艷宇, 楊小彬, 傳金平, 等. 一維卷積神經網絡特征提取下微震能級時序預測. 工程科學學報, 2021, 43(7):1003 [21] He Z X, Peng P G, Liao Z Q. An automatic identification and classification method of complex microseismic signals in mines based on Mel-frequency cepstral coefficients. J Saf Sci Technol, 2018, 14(12): 41何正祥, 彭平安, 廖智勤. 基于梅爾倒譜系數的礦山復雜微震信號自動識別分類方法. 中國安全生產科學技術, 2018, 14(12):41 [22] Xie T, Zheng X D, Zhang Y. Seismic facies analysis based on linear prediction cepstrum coefficients. Chin J Geophys, 2016, 59(11): 4266 doi: 10.6038/cjg20161127解滔, 鄭曉東, 張?. 基于線性預測倒譜系數的地震相分析. 地球物理學報, 2016, 59(11):4266 doi: 10.6038/cjg20161127 [23] Chen R H, Huang H M, Chai H M. Study on the discrimination of seismic waveform signals between earthquake and explosion events by convolutional neural network. Prog Geophys, 2018, 33(4): 1331 doi: 10.6038/pg2018BB0326陳潤航, 黃漢明, 柴慧敏. 地震和爆破事件源波形信號的卷積神經網絡分類研究. 地球物理學進展, 2018, 33(4):1331 doi: 10.6038/pg2018BB0326 [24] Davis S, Mermelstein P. Comparison of parametric representations for monosyllabic word recognition in continuously spoken sentences. IEEE Trans Acoust Speech Signal Process, 1980, 28(4): 357 doi: 10.1109/TASSP.1980.1163420 [25] Wu M T, Chen Q S, Qi C C. Slope safety, stability evaluation, and protective measures based on machine learning. Chin J Eng, 2022, 44(2): 180武夢婷, 陳秋松, 齊沖沖. 基于機器學習的邊坡安全穩定性評價及防護措施. 工程科學學報, 2022, 44(2):180 [26] Li B. Practical Application of Machine Learning. Beijing: Posts & Telecom Press, 2017李博. 機器學習實踐應用. 北京: 人民郵電出版社, 2017 [27] Yu D C, Zhao W F, Nie K, et al. Visibility forecast model based on LightGBM algorithm. J Comput Appl, 2021, 41(4): 1035余東昌, 趙文芳, 聶凱, 等. 基于LightGBM算法的能見度預測模型. 計算機應用, 2021, 41(4):1035 [28] Zhou Z H. Machine Learning. Beijing: Tsinghua University Press, 2016周志華. 機器學習. 北京: 清華大學出版社, 2016 [29] Jung Y. Multiple predicting K-fold cross-validation for model selection. J Nonparametric Stat, 2018, 30(1): 197 doi: 10.1080/10485252.2017.1404598 [30] Rodriguez J D, Perez A, Lozano J A. Sensitivity analysis of k-fold cross validation in prediction error estimation. IEEE Trans Pattern Anal Mach Intell, 2010, 32(3): 569 doi: 10.1109/TPAMI.2009.187 [31] Qi C C, Fourie A, Du X H, et al. Prediction of open stope hangingwall stability using random forests. Nat Hazards, 2018, 92(2): 1179 doi: 10.1007/s11069-018-3246-7 [32] Liu H L, Wu S, Wei G Y, et al. Click-through rate prediction model based on a deep neural network. Chin J Eng,https://doi.org/10.13374/j.issn2095-9389.2021.03.23.002劉弘歷, 武森, 魏桂英, 等. 基于深度神經網絡的點擊率預測模型. 工程科學學報,https://doi.org/10.13374/j.issn2095-9389.2021.03.23.002 [33] Xue Y G, Bai C H, Qiu D H, et al. Predicting rockburst with database using particle swarm optimization and extreme learning machine. Tunn Undergr Space Technol, 2020, 98: 103287 doi: 10.1016/j.tust.2020.103287 -
點擊查看大圖
計量
- 文章訪問數: 514
- HTML全文瀏覽量: 153
- PDF下載量: 82
- 被引次數: 0
