Incipient Fault Monitoring Method Based on Quadratic Kernel Entropy Component Analysis and Its Application

Aiming at the characteristics that the incipient faults of mechanical equipment are small and easy to be submerged by noise, this paper proposes a quadratic kernel entropy component analysis algorithm based on kernel entropy component analysis, multi-scale sample entropy, and a comprehensive index combining SPE statistics and
$T^{2}$
statistics, and constructs a data-driven early fault monitoring model. Firstly, the kernel entropy component analysis is used to complete the dimensionality reduction of the data and the initial feature extraction. Secondly, the multi-scale sample entropy is used to completely characterize the complexity of the data samples and the non-linear specificity of the data samples to zoom in on the early micro-failure and carry out the secondary feature extraction. Finally, the comprehensive index combining SPE statistics and
$T^{2}$
statistics is used to comprehensively monitor the incipient fault. The model is validated by public datasets and has been verified to have good early fault monitoring capability in engineering examples.