Prediction of instability of formwork concrete pier based on big data machine learning for secondary mining without coal pillar mining.

This study addresses the problem of excessive damage to flexible formwork concrete pier columns caused by secondary mining in pillarless coal mining, with a focus on the 1315 working face of Zhaozhuang Coal Mine and the 23,107 working face of Xiegou Coal Mine. Through field research, numerical simulation, theoretical analysis, big data machine learning, and field testing, the stress migration patterns and destabilization mechanisms of flexible formwork concrete pier columns under secondary mining conditions were investigated. The findings revealed that stress concentration in the columns during mining could lead to excessive damage, compromising safety. A Gaussian process regression (GPR)-based stress prediction model was developed (optimal kernel: ARD-Rational-Quadratic-Kernel, with MSE = 1.3463, RMSE = 1.1603, MAE = 0.6138, and MAPE = 0.4041), demonstrating significantly higher accuracy than linear regression models (error reduced by 1-2 orders of magnitude) and BP neural networks (MSE = 2.0962). The model further indicated that the damage extent of the columns followed a two-stage pattern with increasing distance from the mining face: initial near-linear growth, followed by a stabilized rate of increase. Field tests confirmed that reinforcing the flexible pier columns with Z6 concrete reinforcing agent ensured safe mining operations, validating the practical applicability of the prediction model.