Abstract: Metal interference has always been one of the key factors affecting the reliability of transient electromagnetic (TEM) data in mines. Not only does it cause misjudgment in data interpretation, but it also can lead to ineffective signal. It is crucial to comprehensively study the TEM response characteristics under the influence of metal environment. Thus, it is necessary to adopt the 3D vector finite-element (FE) method to realize a TEM forward modeling. Using refined tetrahedral grids for the spatial discretization of the computational domain, utilizing the second-order backward Euler scheme to discretize the governing equation, and applying the direct solver MUMPS for the solution of the linear equations, so that the high-precision numerical solutions can be obtained. After verifying the accuracy of the algorithm, comparing the results of mine TEM responses in metal interference environments, it proves that the unstructured FE method can get good results for the metal-containing full-space models even with significant resistivity contrast and large permeability parameters. To analyze the TEM responses under a metal interference environment, typical block-shaped metal and I-beam models that are suitable for actual sizes are designed, the EM diffusion profiles are displayed for different models with refined tetrahedral mesh to fit interface, and the influences of the distance between the metal object and receiving coil and the size of the metal on the response are discussed. The numerical results of unstructured FE method indicate that the metal in the roadway distorts the EM fields, the electromotive force (EMF) curve generates abnormally high values in the late time. The larger the metal size is, or the closer the metal is from the coils, the larger the amplitude of the EMF and the influence area, even submerging the signal of the target water bearing geological body. For the block metal and the continuous I-beam having similar response characteristics, its influence can be weakened by adjusting their distance.