Abstract: The water inrush hazards have become one of the major threats in tunnel construction, and the 3D localization of water inrush hazard sources (water-bearing structure) in front of the tunnel face has become a key problem to be solved urgently. For the above problem, the method of direct current (DC) resistivity 3D inversion tomography advanced detection is proposed. Firstly, 3D whole space stereoscopic observation mode of DC resistivity advanced detection is established. Due to serious flanking interference near survey line, an interference recognition and removal method is proposed, which realizes the measurement of 3D effective information in front of the tunnel face. By analyzing the advantages of linear and nonlinear inversion methods, a joint inversion tomography method which combines the improved genetic algorithms (GA) and the least-squares method is proposed. By use of this method, the dependence on the initial model is eliminated, while the efficiency of inversion is guaranteed. Then, numerical experiments for inversion tomography focusing on the faults, fractures and karst caves and other typical water-bearing structures are carried out systematically. And, the 3D tomography characteristics and geological interpretation criterion about typical water-bearing structures are analyzed and gained. An inversion map from observed data to 3D tomography and then to geological model is realized. Finally, the proposed method is applied in model tests and practical projects. Accurate 3D localization about water-bearing structures is realized, and it proves that this method has good application prospect.