An Investigation into Noise Source Separation and Blind Identification Method for Electric Drive System Based on Single-Channel Noise Sources

During the analysis of noise source characteristics in electric vehicles, signal processing methods are required for the noise excitation source of the electric drive system (EDS) to obtain its time–frequency domain signal features. This paper proposes a single-channel noise source separation and identification method. Firstly, the single-channel noise source separation and blind identification methods, the complete ensemble EMD with adaptive noise (CEEMDAN) and the improved CEEMDAN (ICEEMDAN), have been established based on the empirical mode decomposition (EMD) algorithm. Comparative simulation and analysis are conducted by using similarity coefficients and residual errors as parameters, also with independent component analysis (ICA) method. Finally, acoustic noise data collection and processing for the EDS under multiple operating conditions are performed, in which the obtained data from steady-state conditions are analyzed using the proposed ICEEMDDAN methods both with ICA. The results show that multiple independent noise signals can be effectively obtained after using the proposed methods. Furthermore, by evaluating the sound pressure level of each independent sound source in the time–frequency domain, the significant contributions of secondary meshing noise and switching frequency noise in the analyzed operating conditions are determined.