关键词: 矿山充填, 电容层析成像, 图像重建, 极限学习机, 卷积神经网络

Abstract: In the process of mine filling,the pipe of agglomeration and inclusions of waste rock filling pulp causes blocking pipe or tube and so on serious safety accidents,restricted the application and development of mine filling technology,so in a timely manner to the pipe blockage agglomerate and detect the location and size of waste rock has great significance to the security and stability of mine filling.Based on the technology of Electrical Capacitance Tomography (ECT),the detection method of mine filling pipeline was studied.Aiming at the problems of poor imaging quality and low accuracy of the traditional ECT reconstruction algorithm,a new ECT image reconstruction method based on extreme learning machine and convolutional neural network was proposed,which was suitable for the detection of solid-liquid two-phase flow in filling pipes.The image reconstruction network consists of two parts:a hidden layer feedforward neural network and an image prediction network.The mapping relationship between capacitance data and dielectric constant value was established by using extreme learning machine,and the image was reconstructed by image prediction network.Through the simulation experiment of filling pipe,it is proved that this method can not only effectively reduce the artifact and deformation of reconstructed image,improve the reconstruction accuracy,but also have a good reconstruction effect for the complex situation that may occur in filling pipe.The ECT image reconstruction method proposed in this study has a good effect on the detection of clogging and caking and waste rock existing in the mine filling pipeline,which can improve the application and promotion of ECT technology in the filling pipeline detection.