Characterisation for spatial distribution of mining-induced stress through deep learning algorithm on SHM data

ABSTRACT

The study of mining-induced stress is essential to ensure the safety production of coalmine. Due to the limited number of monitoring points and local monitoring area, the perception of structure status is insufficient. This study aims to present a deep learning (DL) model to derive the stress distribution characteristics of the overall coalmine roof. First, the framework of spatial deduction model termed as transferring convolutional neural network (TCNN) is presented, where the convolutional neural network is transferred on different datasets. According to this framework, the spatial correlations of structural mechanical responses at different heights above roadway roof are learned through numerical simulation. Subsequently, the learned results are transferred to monitoring data to derive the actual state of the overall roof. In order to verify the reliability of the TCNN model, the stress sensor is installed in the derived plane to collect the actual data, and two indicators are adopted to evaluate the reasonability of deduction results. Experimental results indicated that 92.25% features of mining-induced stress distribution are captured by the TCNN model and the deduction error is 2.037 MPa. Therefore, the presented model is reliable to obtain the overall mechanical state of the coalmine roof, and it is supposed to promote the application of DL in underground construction.

KEYWORDS:

• Underground construction
• deep learning
• monitoring
• coalmine
• mining-induced stress

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work is supported by the National Natural Science Foundation of China [grant number U1806226], [grant number 51991392]; Key deployment projects of Chinese Academy of Sciences [grant number ZDRW-ZS-2021-3]; Project for Research Assistant of Chinese Academy of Sciences [grant number E2294102].