Masked Time–Frequency Representation and Enhanced Priors Enabled Weak Electromechanical Signal Decomposition Algorithm

Electromechanical equipment typically consists of numerous parts, including motors, reducers, and loads, which contribute to the complexity of signals sampled from the equipment. Performing signal decomposition calculations to extract highly state-correlated yet weak components holds significant engineering application value, as these components are crucial for accurate equipment state analysis and evaluation. In this article, we propose a masked time–frequency representation (MTFR) enabled adaptive nonlinear chirp signal decomposition algorithm, abbreviated as MACMD. This algorithm primarily consists of three calculation steps, forming a multilayered framework for signal decomposition. The first step employs a novel method for signal preprocessing, which involves an MTFR to limit the calculation range of component signals (CMs), ensuring the extraction of significant and relevant components. The rest two steps, based on the original adaptive nonlinear chirp signal decomposition (ACMD) algorithm, incorporate smooth and sparse priors of the instantaneous frequency (IF) of the CM, which allows for more robust noise resistance and a more accurate representation of the extracted components. The constructed framework effectively integrates inner product-based algorithms, adaptive decomposition algorithms, and signal sparse prior information, achieving an effective deconstruction of the signal’s time–frequency structure. Furthermore, the decomposed CMs possess more significant frequencies/orders with physical meanings than the original signal, and the proposed algorithm provides an effective application tool for complex equipment condition monitoring. Finally, the effectiveness of the algorithm is validated through simulation and actual signal calculation.