Scholarlink

<p>The Transient electromagnetic method is widely used in coal mine goaf exploration. However, data inversion is a complex problem requiring nonlinear solution equations, and the electromagnetic response of double-layer goafs is complex, leading to low accuracy in identifying their types and ranges. A three-dimensional model conforming to the geological characteristics of the study area is constructed, and the transient electromagnetic induction electromotive force data at each measurement point are obtained via numerical simulations. Based on the smoke ring inversion and depth correction of the data, apparent resistivity and its gradient along the apparent depth, as well as the logarithm of induced electromotive force and its gradient along the time plane, are selected as the sample parameters, which are sensitive to the electrical reflection and abnormal characterization of goafs. Principal component analysis of the multilayer depth parameters yields the standard values of each parameter at the depth of the target layer. Training samples for different goaf types are established, and the backpropagation (BP) neural network algorithm is used to train the samples to construct a neural network model of each goaf type. The training samples are back-estimated, and the type of goaf in the entire area is predicted. Results show that the back-estimation prediction accuracy of the No.2 coal seam samples is 84.7%, and the prediction accuracy of the entire area is 84.3%. Meanwhile, the back-estimation prediction accuracy of the No.3 coal seam samples is 94%, and the prediction accuracy of the entire region is 87.87%. The proposed method is used to identify the type of double-layer goaf in coal mines, and the distribution characteristics of the goaf type in the entire area are rapidly obtained. The back-estimation prediction accuracies of the No. 2 and 3 coal seam samples are 91.1% and 93.3%, respectively. Discrimination results are confirmed via drilling. Findings show that the proposed BP neural network discriminant model based on the transient electromagnetic multiparameters realizes the rapid and accurate identification of double-layer goaf types in coal mines and avoids multiple solutions and low recognition caused by the inversion of the transient electromagnetic data.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Neural Network–Based Prediction Method for Double-layer Goaf Types in Coal Mines Based on Multiparameters of Transient Electromagnetic Method

  • 期刊论文

  • 发表日期：

    2025年8月11日

查看全文
  • Yi Dong,
  • Haijun Xie,
  • Nan Zhu,
  • Lingling Lu

摘要

The Transient electromagnetic method is widely used in coal mine goaf exploration. However, data inversion is a complex problem requiring nonlinear solution equations, and the electromagnetic response of double-layer goafs is complex, leading to low accuracy in identifying their types and ranges. A three-dimensional model conforming to the geological characteristics of the study area is constructed, and the transient electromagnetic induction electromotive force data at each measurement point are obtained via numerical simulations. Based on the smoke ring inversion and depth correction of the data, apparent resistivity and its gradient along the apparent depth, as well as the logarithm of induced electromotive force and its gradient along the time plane, are selected as the sample parameters, which are sensitive to the electrical reflection and abnormal characterization of goafs. Principal component analysis of the multilayer depth parameters yields the standard values of each parameter at the depth of the target layer. Training samples for different goaf types are established, and the backpropagation (BP) neural network algorithm is used to train the samples to construct a neural network model of each goaf type. The training samples are back-estimated, and the type of goaf in the entire area is predicted. Results show that the back-estimation prediction accuracy of the No.2 coal seam samples is 84.7%, and the prediction accuracy of the entire area is 84.3%. Meanwhile, the back-estimation prediction accuracy of the No.3 coal seam samples is 94%, and the prediction accuracy of the entire region is 87.87%. The proposed method is used to identify the type of double-layer goaf in coal mines, and the distribution characteristics of the goaf type in the entire area are rapidly obtained. The back-estimation prediction accuracies of the No. 2 and 3 coal seam samples are 91.1% and 93.3%, respectively. Discrimination results are confirmed via drilling. Findings show that the proposed BP neural network discriminant model based on the transient electromagnetic multiparameters realizes the rapid and accurate identification of double-layer goaf types in coal mines and avoids multiple solutions and low recognition caused by the inversion of the transient electromagnetic data.