Abstract: Insulated gate bipolar transistor(IGBT) is the main component of inverter. Its stability and reliability have great influence on the normal operation of the inverter. To detect the common single-IGBT and double-IGBTs open-circuit fault of inverter, a fault diagnostics method based on empirical mode decomposition(EMD) and back propagation neural network(BPNN) is proposed. Firstly, the waveform characteristics of single-IGBT and double-IGBTs open-circuit fault of inverter are analyzed, and the three-phase current outputs of the inverter are decomposed by EMD. On this basis, the fault eigenvector is constructed and then trained by BPNN to realize the fault diagnosis. Secondly, to solve the mode aliasing in EMD, the diagnostics method is optimized by using complementary ensemble empirical mode decomposition(CEEMD). Finally, the proposed diagnostics method is verified on a low-power testbed, and the experimental results show that the diagnostics method based on CEEMD-BPNN can diagnose the faulty IGBTs effectively.