Identification Technology and Application of Filling Media in Deep Goaf Areas for Mine Water Hazard Prevention and Control
Abstract
Accurate identification of filling media in goaf areas is essential for assessing water hazards and ensuring water environment safety in deep mines. Traditional natural source magnetotelluric methods rely on single parameters, which often lead to ambiguous inversion results, making it challenging to precisely determine the specific fill materials in goaf areas. To address this common issue in mine water hazard detection, this study focuses on the complex water environments of deep mines and conducts specialized technical research on single-component natural source magnetotellurics. Through a series of experiments covering four typical filling modes (gas, water, collapsed rock, and mixed fill), we employed Fourier series methods for frequency domain decomposition of waveform data that contain physical property information. An Random Forest classification model was developed using the direct current component, fundamental component, second harmonic, and third harmonic, achieving a classification accuracy of 85% and indicating good discriminative capability of these frequency domain features. Sobol global sensitivity analysis was then performed to quantify the relative contributions of these components to variations in the physical property index. A recognition model for physical property index of filling media in goaf areas compatible with mine water environments was developed. To validate the physical property index model, field experiments in four goaf scenarios and complex working faces effectively addressed industry challenges, providing reliable methods for detecting old mine water hazards.