程久龙, 赵家宏, 董毅, 等. 基于LBA-BP的矿井瞬变电磁法岩层富水性的定量预测[J]. 煤炭学报, 2020, 45(1): 330-337. DOI: 10.13225/j.cnki.jccs.YG19.1508
引用本文: 程久龙, 赵家宏, 董毅, 等. 基于LBA-BP的矿井瞬变电磁法岩层富水性的定量预测[J]. 煤炭学报, 2020, 45(1): 330-337. DOI: 10.13225/j.cnki.jccs.YG19.1508
CHENG Jiulong, ZHAO Jiahong, DONG Yi, et al. Quantitative prediction of water abundance in rock mass by transient electro-magnetic method with LBA-BP neural network[J]. Journal of China Coal Society, 2020, 45(1): 330-337. DOI: 10.13225/j.cnki.jccs.YG19.1508
Citation: CHENG Jiulong, ZHAO Jiahong, DONG Yi, et al. Quantitative prediction of water abundance in rock mass by transient electro-magnetic method with LBA-BP neural network[J]. Journal of China Coal Society, 2020, 45(1): 330-337. DOI: 10.13225/j.cnki.jccs.YG19.1508

基于LBA-BP的矿井瞬变电磁法岩层富水性的定量预测

Quantitative prediction of water abundance in rock mass by transient electro-magnetic method with LBA-BP neural network

  • 摘要: 作为巷道超前探测主要方法的矿井瞬变电磁法,目前仅能通过视电阻率来定性评价掘进工作面前方岩层富水性相对强弱。为了实现岩层富水性的定量预测,依据阿尔奇公式建立掘进工作面前方岩层富水性不均匀的数学模型,通过全空间三维时域有限差分(FDTD)数值模拟结果提取预测所用特征,根据所选特征与岩层富水性相关的地质参数确定富水性预测神经网络的结构。采用具有Lévy飞行特征的蝙蝠算法优化BP神经网络(LBA-BP)进行了富水性预测的仿真测试和现场试验。在仿真测试中,为接近实际情况,在数值模拟结果中加入了不同程度的噪声,对比测试结果发现尽管预测误差随噪声的增大而明显波动,但岩层孔隙度和含水饱和度预测的均方误差均不超过1%,预测准确度处在一个较高的水平。在现场试验中,根据现场地质资料划分富水性评价分级标准,由预测结果和视电阻率解释结果可以看出,LBA-BP预测方法对巷道前方岩层富水性的预测准确度要明显优于常规视电阻率解释方法。研究结果表明:感应电动势、采样时间、视电阻率、探测距离和对数坐标下的感应电动势衰减速率可作为岩层富水性预测所用特征,LBA-BP方法结合相应特征可以实现岩层富水性定量预测,提高了矿井瞬变电磁法对岩层富水性的解释精度。

     

    Abstract: The existence of water in the rock stratum in front of the roadway seriously threatens the safe driving of the mine roadway. Mine Transient Electromagnetic Method is the main method in advanced detection of roadway. Current- ly,apparent resistivity is the only way to qualitatively evaluate the water abundance of rock stratum in front of road- way. In order to realize a quantitative prediction of water abundance in rock stratum,the first step is to establish the mathematical model of the uneven water abundance of rock stratum in front of the roadway based on Archie formula. Then numerical simulation can be carried out by 3D finite difference time domain (FDTD) and its results can be used to extract the features needed in prediction. According to the selected features and the geological parameters related to the water abundance of rock stratum,the water abundance can be determined to predict the structure of neural net- work. The BP neural network was optimized by the bat algorithm based on Lévy flight improvement,and the simulation test and field test of water abundance prediction were carried out by using the LBA-BP method. In the simulation test, in order to approach the actual situation,different degrees of noise were added to the numerical simulation results. By comparing the test results,it was found that although the prediction error fluctuated significantly with the increase of noise,the mean square error of rock porosity and water saturation prediction was less than 1% ,and the prediction ac- curacy was at a high level. In the field test,according to the classification criteria of water abundance evaluation based on the field geological data,it can be seen from the prediction results and the apparent resistivity interpretation results that the accuracy of the LBA-BP prediction method is significantly better than that of the conventional apparent resis- tivity interpretation method in predicting the water abundance of stratum in front of the roadway. The results show that the induced electromotive force,sampling time,apparent resistivity,detection distance and induced electromotive force attenuation rate in logarithmic co-ordinates can be used as the characteristics of water abundance prediction. The LBA- BP method realizes the near quantitative or quantitative prediction of water abundance in rock stratum,and improves the interpretation accuracy of the advance detection data for water abundance in rock stratum.

     

/

返回文章
返回