全尾砂胶结充填法，在金属矿山的开采中被广泛应用。充填料浆在管输过程中的堵管、爆管现象，给充填管道带来很大的安全隐患，是制约充填技术发展和应用的瓶颈。为了探索矿山充填管道堵塞的可视化检测方法，获取矿山充填管道内壁结块的几何形状与相对位置的三维信息，利用COMSOL软件建立多层ERT传感阵列的三维模型，采用有限元方法求解ERT正问题，阐明了ERT三维敏感场内电势与电流密度的变化趋势。在此基础上，以灵敏度为媒介分析ERT的“软场”特性，揭示了三维敏感场内检测灵敏度的分布规律。采用Gauss-Newton算法求解ERT的逆问题，对5种堵塞仿真模型进行三维图像重建，仿真结果表明正则化Gauss-Newton一步算法在重建图像质量上与正则化Gauss-Newton迭代算法差别极小，但在重建时间上优于正则化Gauss-Newton迭代算法，更适合三维图像重建。Gauss-Newton一步算法中选择Laplace先验知识优于Noser先验知识。最后，根据仿真模型，采用某金属矿尾砂制作尺寸不同的4种结块试件，养护14 h后放入充满质量分数为74%全尾砂充填料浆的竖直管道中的5个不同位置，模拟充填管道内壁5种不同的结块位置及堵塞程度。利用自主研发的48电极3D-ERT成像系统，采用Gauss-Newton一步算法，对以上5种堵塞情况进行三维成像实验。实验结果表明：实测结果与仿真结果一致，重建图像可以准确反映出管道内壁结块的几何形状和所处位置等三维信息。

Full tailing sand cementation filling method has been widely used in metal mines. However, the safety and routine operations of filling pipeline are significantly threated by plugging and bursting of the filling slurry in the pipe transmission process, which is a great obstacle restricting the development and application of filling technology. In the paper, the three-dimensional model of multi-layer ERT sensing array is constructed based on COMSOL software for the purpose of exploring the visualization detection method of mine filling pipe blockage, obtaining the three-dimensional information of geometry and relative position of the lump inside the mine filling pipe. Moreover, the ERT positive problem is solved by using finite element method. The trend of potential and current density in the three-dimensional ERT sensitive field are revealed. On this basis, the “soft field” characteristics of ERT were analyzed in terms of sensitivity, and the distribution law of detection sensitivity in the three-dimensional sensitive field is provided. The inverse problem of ERT is solved by Gauss-Newton algorithm, and the three-dimensional image reconstruction is performed for five blockage simulation models. The simulation results show that the difference between the regularized Gauss-Newton one-step algorithm and the regularized Gauss-Newton iterative algorithm in terms of reconstructed image quality is minor. The reconstruction time of the regularized Gauss-Newton one-step algorithm is, however, better than the regularized Gauss-Newton iterative algorithm, which is more suitable for 3D image reconstruction. The Laplace prior knowledge selected in the Gauss-Newton one-step algorithm is better than the Noser prior knowledge. Finally, according to the simulation model, four kinds of agglomerated specimens with different sizes were made using the metal mine tailing sand. After 14 hours of curing, the specimens are put into five different positions in a vertical pipe filled by full tailing backfill with 74% solid content to simulate different agglomerated positions and clogging degrees inside the filling pipe. Then the self-developed 48-electrode 3D-ERT imaging system and the Gauss-Newton one-step algorithm are utilized to perform 3D imaging experiments on the above five blockage cases. The experimental results show that the measured results are consistent with the simulation results, and the reconstructed images can accurately reflect the three-dimensional information of the geometry and location of the clumps inside the pipeline.