Nondestructive testing of substation grounding grids is an issue that has increasing importance. The traditional EMI method transforms the condition of the undergrounding conductors to the surficial induced electric signal in the sensing coil. However, The EMI signals excited by multiple coexisted faults combining with other unknown noises surrounding the substation often cause the failure of detection. Therefore, the observed EMI signals rather complex and cannot be used directly. To address this problem, the separation of individual signatures from the mixture is posed as an SCBSS problem. To extract the induced signal, an EEMD-based EMI method is proposed. The desired signal is then reconstructed to visualize the structure of the grounding grids by a virtual instrument that consists of DAQ and digital signal processing modules. The numerical simulation and practical experiments are employed. The results show the proposed method can be used to effectively detect the topological structure of grounding grid in real substations' electromagnetic environment.INDEX TERMS Grounding grid, electromagnetic induction (EMI), ensemble empirical mode decomposition (EEMD), single channel blind source separation (SCBSS), nondestructive testing (NDT).