One of the most significant challenges for automating coal mining machinery is the shearer horizon control, of which the fundamental problem focuses on the coal-rock interface recognition (CIR). This paper proposes two novel CIR methods as well as the radar-based CIR scheme in underground mining. The low-frequency ultra-wideband (UWB) electromagnetic pulse radars are mounted on the shield plates of hydraulic supports to estimate the air and coal layer thickness above their antenna. First, the modified spatial smooth processing (MSSP) based covariance matrix is built to de-coherent the coherent echoes. Then, the lower–upper (LU) decomposition is incorporated into the conventional matrix propagator to improve the robustness and efficiently estimate the noise subspace. The Multiple Signal Classification (MUSIC) algorithm is applied to estimate the interface information. To further improve the robustness and accuracy, the alternative iteration process is given by using the oblique projection and the maximum likelihood (ML) method. Moreover, a priori knowledge of the emitted signal is not required anymore. The effectiveness and efficiency of two proposed algorithms are verified by simulation and experimental analyses.
 

coal-rock interface,matrix propagation,maximum likelihood,radar signal processing