In the realm of blind deconvolution (BD), accurately estimating transfer functions is often impractical due to the complexity and noise associated with machinery systems, rendering it an ill-posed problem.
The use of sparsity indexes as optimization objective functions in BD helps combat the challenges posed by non-stationary and periodic fault characteristics prevalent in machinery systems.
Kurtosis serves as a valuable statistical tool in the optimization of blind deconvolution, as increased kurtosis values indicate a significant deviation from a standard normal distribution, highlighting cyclic impulses in vibration signals.
The methodology section details the development of a time domain quadratic convolutional filter, showcasing its superiority in extracting cyclic features compared to conventional neural networks.
#blind-deconvolution #noise-estimation #quadratic-neural-networks #machine-learning #signal-processing
Collection
[
|
...
]