Sparse and low-rank enhanced dictionary learning for aero-engine gear fault diagnosis

Gearfaultsare one of the most frequent inducements of aero-engine failures. Therefore, it is extremely critical to perform timely and reliable aero-engine gear fault diag nosis. However, it is a challenge to extract weak gear fault features from the vibration signals that are submerged by complex interference from multiple sources. In this paper, a dual reweighted sparse and low-rank enhanced dictionary learning (Re²wSLDL) method is pro posed to incorporate sparse and low-rank fault priors into dictionary learning. Specifically, the reweighted l₁-norm is used as the sparse regularization and the reweighted nuclear norm is used as the low-rank regularization, so that the fault dictionary is learned under the syner gistic constraints of one-dimensional global sparsity and two-dimensional global low-rank property. Furthermore, an iterative solver combining variable splitting and alternating opti mization is developed. Finally, simulation studies and application cases verify the effective ness and superiority of the proposed method in aero-engine gear fault diagnosis.