Conventional ICA algorithms are sensitive to initial values, prone to local optima, and lack robustness in practice. To address these issues, we incorporate quantum-behaved particle swarm optimization into independent component analysis and propose a blind source separation algorithm called QPSO-ICA. The demixing matrix is encoded with quantum bit probability amplitudes; superposition is used to preserve population diversity, and quantum rotation gates drive the state update. In each iteration, a valid demixing matrix is obtained through measurement collapse and orthogonalization, with a fitness function that maximizes non-Gaussianity while penalizing correlations among the extracted components.
On simulated signals (5 Hz sine, 3 Hz square, and 7 Hz sawtooth mixed with 30 dB Gaussian noise), QPSO-ICA achieves an average correlation coefficient of 0.9783, on par with FastICA (0.9799), but with smaller variance across repeated runs. On real rolling-bearing data (ball, inner-race, and outer-race faults), the average correlation coefficient reaches 0.9098, higher than 0.8887 for FastICA, and the envelope spectrum peaks of the separated signals agree with the theoretical fault characteristic frequencies. The results on both simulated and measured data confirm the effectiveness of the proposed method and its practical value for breach detection.
 

blind source separation; independent component analysis; quantum-behaved particle swarm optimization; bearing fault diagnosis