Sparsity-constrained Ramanujan subspace feature tracking method for compound fault diagnosis

To solve the problems of insufficient feature extraction accuracy in compound fault diagnosis caused by the strong coupling of multi-fault features and obvious differentiation of intensity characteristics, a novel multi-fault feature extraction method is proposed in this paper, named sparsity-constrained Ramanujan subspace feature tracking (SRSFT). By developing a novel sparse Ramanujan sequence, the proposed method establishes a sparse subspace with inherent periodicity. This framework effectively maps the signal into a domain where fault features are more readily separable, thereby achieving a marked enhancement in both decomposition efficiency and compound fault feature extraction. Furthermore, the proposed enhanced period estimation technique improves the accuracy of signal weak periods detection while reinforcing the stability of feature extraction performance. Notably, the SRSFT method demonstrates exceptional capability in separating and extracting compound fault features by comprehensively leveraging the periodicity and sparsity of fault features. Simulation and experimental results consistently demonstrate that the proposed method exhibits superior performance in multi-period impulse separation and extraction, offering an effective and robust solution for compound fault diagnosis applications.