LAUSR

Swarm decomposition (SWD) is an emerging signal decomposition method and has been applied in the fault diagnosis of rotating machinery. However, the performance of SWD is highly dependent on the user-defined parameter. In this article, an adaptive swarm decomposition (ASWD) method guided by spectral characteristic information scanner (SCIS) is proposed to automatically decompose the vibration signal into a set of subcomponents. The proposed method can not only adaptively extract the weak fault-related component from the signal contaminated by strong noise but also avoid the problem of the user-defined parameter in the original SWD. First, the estimation approach of center frequencies (CFs) in the original SWD is thoroughly analyzed to explore the main factor influencing the division of frequency bands. Then, a novel adaptive SCIS motivated by the convergence tendency of variational model is established to reveal spectrum structure information of the input signal and thus detects the target CFs simultaneously without any prior knowledge. Subsequently, the proposed method incorporates the SCIS, thereby effectively implementing the adaptive division of frequency bands with no requirement of any predefined parameter. The numerical simulation and two experimental cases are used to verify the feasibility and superiority of the proposed ASWD by comparison with some prevalent signal processing methods.