Abstract: Accurately detecting the structure of top coal is crucial for achieving intelligent fully-mechanized caving mining. However, the complex structure of coal seams in China, often consisting of multiple layers of separated gangue, can significantly impact the effectiveness of fully-mechanized caving mining. To address this challenge, this paper proposes a precise detection method for top coal structure based on multi-frequency radar fusion. This method aims to improve the detection accuracy of the shallow part within the radar antenna’s detection depth and interpret the structural information inside the top coal. The main research steps are as follows: Firstly, preprocessing and spatial alignment of radar data of different frequencies are carried out to establish the spatial correspondence between radar data of different frequencies. Secondly, a sliding window and wavelet transform weighted fusion method is employed to process the multi-frequency radar data. The time-varying weight value of each frequency signal is determined based on the energy proportion of each segment wavelet signal in the window. Additionally, an edge detection algorithm is introduced to enhance the fusion efficiency of the wavelet transform to the radar data, thereby achieving effective fusion of the multi-frequency radar data. Finally, taking into account the differences in dielectric constants between coal, gangue, and rock, as well as the attenuation characteristics of electromagnetic wave propagation, an internal echo intensity model of top coal is established. The interface information between coal, gangue, and rock inside the top coal is calculated using a stratified identification method, enabling the inverse interpretation of the internal structure of top coal, including the thickness of top coal, the thickness and number of gangue layers, and the spacing between gangue layers. The test results demonstrate that the proposed method can effectively integrate radar data of different frequencies and accurately detect the internal structural characteristics of top coal. Moreover, the detection errors of the thickness of top coal, the thickness of gangue, and the spacing between gangue layers are all less than 10%. This method enables the accurate detection of top coal structure, providing theoretical and technical support for the intelligent fully-mechanized caving mining.

Key words: fully-mechanized caving mining, multi-frequency radar fusion, top coal structure, detection