Feature Intelligent Generation Based on Quantum Computing and Variational Independent Component Analysis for Nonstationary Weak Fault Diagnosis

Aiming at reinforcing the significance of non-stationary weak fault signal buried in strong random noise for fault diagnosis of rotating machinery, this article presents a novel QVFIG method of feature intelligent generation by enhancing the fault signal-noise-ratio in time domain, time-frequency domain, Hilbert space, and spatial domain. We construct an intelligent diagnosis model based on the proposed novel variational blind source separation method, variational mode decomposition, time-frequency distribution of Cohen's class, quantum computing, and generative adversarial networks. Extensive experiments on three test rigs verify the proposed QVFIG method has robust fault feature representation and discriminative ability, which can provide the clear proof for preventive maintenance decision-making.