In the heart of China’s eastern mining regions, a silent battle is waging beneath the surface. Vertical shafts, the lifelines of mines, are succumbing to tilt deformations and breakages, threatening both safety and productivity. Enter Hui Liu, a researcher from the College of Resource and Environmental Engineering at Anhui University, who is leading the charge against these subterranean challenges with an innovative deep learning approach.
Liu and his team have been grappling with the perplexing problem of shaft tilting in deep vertical shafts, particularly those traversing thick water-bearing loose layers. Their latest study, published in Meitan xuebao, or the Journal of Coal Science and Engineering, offers a beacon of hope for the mining industry.
The team’s focus was a deep vertical shaft in Lunan, plunging 800 meters into the earth. By monitoring the shaft’s tilting and deformation, they uncovered crucial insights. “The wellbore tilt mainly occurs in the loose layer,” Liu explains, “and the tilt value decreases linearly from shallow to deep, with a maximum of 352 mm.” The bedrock layer, while more stable, still showed deformations up to 88 mm. The culprits? Mining-induced deformation propagation in the loose layer and changes in the aquifer’s seepage hydrophobicity.
But Liu’s team didn’t stop at diagnosis. They harnessed the power of deep learning to predict these deformations, employing four different models: recurrent neural network (RNN), long short-term memory network (LSTM), gated recurrent unit (GRU), and one-dimensional convolutional neural network (1DCNN). The results were impressive. The models showed a strong correlation with measured values, with Spearman correlation coefficients ranging from 0.867 to 0.978. The 1DCNN model, in particular, proved to be the most accurate, with a mean absolute error (EMA) within 0.003–0.009 m and a root mean square error (ERMS) within 0.004–0.011 m.
The implications for the energy sector are profound. Accurate prediction of wellbore deformations can significantly enhance mine safety and productivity. It can also inform better grouting repair and management strategies, as evidenced by the successful application of Liu’s model in a wellbore grouting repair project.
But Liu’s work isn’t just about fixing problems; it’s about preventing them. By understanding the spatial and temporal change characteristics of shaft tilting and the main influencing factors, mining companies can proactively manage their shafts, avoiding costly repairs and dangerous situations.
As Liu puts it, “The wellbore deformation prediction model based on deep learning has good prediction ability.” And with further refinement, it could revolutionize the way we approach mine safety and management.
The research published in Meitan xuebao, or the Journal of Coal Science and Engineering, is a testament to the power of deep learning in tackling real-world problems. It’s a call to action for the energy sector to embrace these technologies, to learn from Liu’s work, and to shape a safer, more productive future for mining. The battle beneath the surface is far from over, but with innovators like Liu leading the charge, the industry is in good hands.
