2025地能科学与工程国际学术会议

21 / 2025-11-18 11:53:00

面向绿色安全开采的倾斜厚大矿体地表移动智能预测系统与应用

智能预测系统、后续充填采矿、PCA-GA-BP神经网络、地表移动与变形、概率积分法

An Intelligent Prediction System for Surface Movement and Deformation in Subsequent Filling Mining of Inclined, Thick Ore Bodies

Abstract：Ensuring green and safe mining is a core demand in the contemporary development of Earth's energy resources. Subsequent filling mining, as a typical green mining technique, still induces surface movement and deformation, posing threats to surface infrastructure and ecological security. Accurate prediction of these deformations is therefore crucial. Traditional methods relying solely on the Probability Integral Method (PIM) often face challenges of low accuracy and poor adaptability when geological conditions change or measured data is scarce.
To address this challenge, this study develops an Intelligent Prediction System using MATLAB App Designer. The core innovation lies in the integration of the theoretical framework of PIM with a hybrid PCA-GA-BP Neural Network. Principal Component Analysis (PCA) optimizes the input parameters, a Genetic Algorithm (GA) optimizes the initial weights and thresholds of the Backpropagation (BP) neural network, ultimately forming a powerful parameter inversion and prediction model.
This system features three main functional modules: initial parameter setting, expected parameter calculation (powered by the PCA-GA-BP network), and visualization of prediction results. It enables the intelligent and automated prediction of surface subsidence, inclination, curvature, horizontal movement, and deformation.
Industrial field validation at the Daxin Beidoushan Phosphate Mine demonstrated the system's high reliability. The predicted maximum settlement values for the strike and inclined main sections were -38.37 mm and -39.73 mm, respectively. The system achieved an overall prediction accuracy of 90.56%, as validated by ROC curve analysis (AUC = 0.9056). Comparative analysis with field measurements confirmed the system's effectiveness, and site-specific correction coefficients were proposed to further enhance prediction precision for practical engineering applications.
This research provides a robust, intelligent tool for predicting mining-induced subsidence, directly supporting the strategic goals of "clean, low-carbon, safe, and efficient" mining. It holds significant value for promoting green mining practices, preventing geological disasters, and ensuring the sustainable development of Earth's energy resources.

Keywords: Intelligent Prediction System, Subsequent Filling Mining, PCA-GA-BP Neural Network, Surface Movement and Deformation, Probability Integral Method.