Underdetermined blind source separation with adaptive chirp mode decomposition for compound rolling bearing fault signals

Abstract The challenge of underdetermined blind source separation (UBSS) is on the need to separate multiple fault signals from a single-channel signal. Conventional separation algorithms are restricted by the lack of sufficient information to extract multiple source signal characteristics. To tackle the challenge, this paper introduces a novel method that combines sparse denoising-aided Adaptive Chirp Mode Decomposition (ACMD) with Fast Independent Component Analysis (FastICA). The underdetermined problem is transformed into a well-determined one by using decomposition techniques. Firstly, enhanced sparsity of GMC is used to denoise the collected signal and further enhance fault features. Next, the denoised signal is decomposed into multiple Intrinsic Mode Functions (IMFs) using ACMD. Subsequently, the optimal IMFs are selected and then constructed a new observation signal, which is then used as the input for FastICA. At the same time, Singular Value Decomposition (SVD) is performed on each decomposed IMF to obtain the corresponding singular values, which are determined the number of signal sources and the separation layers of the FastICA method by the maximum ratio between adjacent singular values. Ultimately, the constructed observation signal is separated. This method is applied to single-channel UBSS of compound faults in rolling bearings. Numerical simulations and experimental studies demonstrate the method's effectiveness.