Bearing Non-Uniform Loading Condition Monitoring Based on Dual-Channel Fusion Improved DenseNet Network

Misalignment or unbalanced loading of machine tool spindle bearings often results in skewed bearing operation, which makes the more susceptible to failure. In addition, due weak impact signal a single feature information cannot accurately characterize operation status bearing. To address above problems, this paper proposes method monitor uneven running state load based on dual-channel fusion improved dense connection (DenseNet) network. First, original is pre-processed by overlapping sampling method, and experimental data are obtained frequency-domain time-frequency-domain algorithms; then processed input into 1D-DenseNet 2D-DenseNet models respectively for extraction; features fused concat splicing output belongs each category The probability distribution used operating bearings. Finally, validity algorithm model verified using Case Western Reserve University public rolling set, an bench designed built verification operation. comparative analysis shows that can extract comprehensively finally achieve 100% recognition accuracy.