针对金属矿山开采过程中的地表沉陷和废料堆积的问题，一般采用尾砂充填法将尾砂废料、水泥、水等混合后填充至采空区。但在充填管道的输送过程中，由于充填料浆的胶结性或管道的不合理使用，会导致充填管道的堵塞和爆管现象，严重危害金属矿山的安全生产。因此，充填管道的堵塞检测及堵塞预警对金属矿山的安全生产意义重大。由于传统的检测方式存在成本高、非可视化、实时性差的问题，因此本研究将电阻层析成像（Electrical Resistance Tomography，简称ERT）技术应用在金属矿山充填管道的堵塞检测和堵塞预警中。本文主要进行了以下工作：

针对传统图像重建算法耗时多、质量低、无法满足金属矿山充填管道堵塞检测需求的问题。在总结现有图像重建算法的基础上，本文提出基于梯度加速的对称交替方向乘子算法（Gradient Acceleration-Symmetric Alternating Direction Multiplier Method，简称GA-SADMM）用于图像重建。通过仿真设置14种不同的流型从直观和数值分析的角度对比GA-SADMM算法与其他图像重建算法的综合性能。实验结果表明，GA-SADMM算法的平均相对误差为0.2322，平均相关系数为0.8345，在保证图像重建质量的同时极大减小了图像重建耗时。

在硬件资源充足的条件下，为了改善灵敏度矩阵的病态程度，进一步提高图像重建的质量和减少成像耗时。本文在现有有限元法的基础上，提出了一种基于电压先验的自适应网格剖分方法。首先，通过输入归一化的测量电压到MobileNetV3网络中，对成像区域进行预测。其次，根据预测结果对成像区域的剖分单元重构并修改剖分模型。最后，调用GA-SADMM算法进行成像。对比优化后的GA-SADMM算法和原GA-SADMM算法的图像重建结果，求解的平均耗时减少了21.3%，平均相关系数提高了19.3%，平均相对误差减小了25.4%，有效的提高图像重建的质量和效率。

为解决传统的检测方式存在成本高、非可视化、实时性差的问题。本文设计开发了基于ERT的金属矿山充填管道堵塞检测系统。系统包括以ERT并行采集无线传感节点为基础搭建的多节点动态实验平台；以组态王设计的上位机监控界面；以MATLAB实现图像重建算法、色彩聚类堵塞信息计算及堵塞风险预警功能。

通过对不同浓度料浆和养护14小时堵塞结块的电导率进行分析，从理论上说明ERT技术对于金属矿山充填管道堵塞检测的可行性。以不同的堵塞结块分别进行静态和动态测量，并根据图像重建结果进行多级风险预警。实验结果表明，各图像重建结果均能反映实际的结块位置大小，且多级风险预警结果准确。

本研究针对金属矿山的充填管道堵塞检测及预警问题，设计了完整的堵塞检测装置，并提出了重建质量高且耗时少的图像重建算法。实验结果表明，ERT技术在金属矿山充填管道堵塞检测中具有积极的学术意义和广泛的应用前景。

In response to the problems of surface subsidence and waste accumulation during the mining process of metal mines, the tailings filling method is generally used to mix tailings waste, cement, water, etc. and fill them into the goaf. However, during the transportation process of the filling pipeline, due to the cementation of the filling slurry or the unreasonable use of the pipeline, it can lead to blockage and explosion of the filling pipeline, seriously endangering the safety production of metal mines. Therefore, the detection and warning of blockage in filling pipelines are of great significance for the safety production of metal mines. Due to the high cost, non visualization, and poor real-time performance of traditional detection methods, this study applies Electrical Resistance Tomography (ERT) technology to the detection and warning of blockage in metal mine filling pipelines. The main work of this article is as follows:

Aiming at the problems of traditional image reconstruction algorithms that are time-consuming, low-quality, and unable to meet the needs of metal mine filling pipeline blockage detection. On the basis of summarizing existing image reconstruction algorithms, this paper proposes the Gradient Acceleration Symmetric Alternating Direction Multiplier Method (GA-SADMM) for image reconstruction based on gradient acceleration. Compare the comprehensive performance of the GA-SADMM algorithm with other image reconstruction algorithms from the perspective of visual and numerical analysis by simulating 14 different flow patterns. The experimental results show that the average relative error of the GA-SADMM algorithm is 0.2322, and the average correlation coefficient is 0.8345, which greatly reduces the image reconstruction time while ensuring the quality of image reconstruction.

Under sufficient hardware resources, in order to improve the pathological degree of the sensitivity matrix, further improve the quality of image reconstruction, and reduce imaging time. This article proposes an adaptive mesh generation method based on voltage prior, based on existing finite element methods. Firstly, the normalized measurement voltage is input into the MobileNetV3 network to predict the imaging area. Secondly, based on the predicted results, reconstruct the segmentation units of the imaging area and modify the segmentation model. Finally, call GA-SADMM algorithm for imaging. Comparing the image reconstruction results of the GA-SADMM algorithm after segmentation with the original GA-SADMM algorithm, the average solving time was reduced by 21.3%, the average correlation coefficient was increased by 19.3%, and the average relative error was reduced by 25.4%, effectively improving the quality and efficiency of image reconstruction.

To solve the problems of high cost, non visualization, and poor real-time performance in traditional detection methods. This article designs and develops a metal mine filling pipeline blockage detection system based on ERT. The system includes a multi node dynamic experimental platform based on ERT parallel acquisition of wireless sensor nodes; The upper computer monitoring interface designed with KingView; Realize image reconstruction algorithms, color clustering blockage information calculation, and blockage risk warning functions using MATLAB.

The feasibility of ERT technology for detecting blockage in metal mine filling pipelines is theoretically demonstrated by analyzing the conductivity of slurry with different concentrations and blockage after 14 hours of maintenance. Perform static and dynamic measurements with different blockages and clumps, and conduct multi-level risk warnings based on image reconstruction results. The experimental results show that the reconstruction results of each image can reflect the actual size of the agglomeration location, and the multi-level risk warning results are accurate.

This study focuses on the detection and warning of filling pipeline blockage in metal mines. A complete blockage detection device has been designed, and an image reconstruction algorithm with high reconstruction quality and less time consumption has been proposed. The experimental results indicate that ERT technology has positive academic significance and broad application prospects in the detection of filling pipeline blockage in metal mines.