An UBSS Method for Signals with Non-Uniform Energy Distribution of Varied Frequency Bins
Naixin Chen, Chunli Zhu, Lei Chen · 2023
Coupled signals sensing from electromechanical systems have significant impacts on the intelligent fault diagnosis application, which is generally formulated as an undertermined blind source separation (UBSS) issue. However, it is especially challenging when dealing with signals of non-uniform energy distribution of different frequency bins. In this work, we proposed an UBSS framework with an adaptive optimal frequency bin selection approach, for improving the signal sparsity and estimation accuracy of the mixing matrix. Simulation results show that the proposed method achieves average accuracy of 82.40%, 86.91 % and 87.19% with the signal-to-noise ratio (SNR) set as 10, 15 and 30$\text{dB}$of the tested case, respectively. This work has a good potential on reducing the intelligent state monitoring system's false alarm rate.