Fault Identification of Mechanical Shock Signals Using Improved EEMD–Multi-Scale Analysis

Precise identification of mechanical shock liabilities in turning machinery is crucial for ensuring consistent operation and preventing early faults. This paper proposes an Improved Collaborative Empirical Mode Decomposition (IEEMD) combined with Multi-Scale Analysis (MSA) to enhance the accuracy and robustness mechanical shock-based fault identification. The planned framework first decomposes complex shock signals into intrinsic mode functions (IMFs) using an augmented noise-aided EEMD with adaptive noise scaling and ensemble averaging to suppress mode mixing and improve frequency resolution. Multi-scale statistical features such as root mean square (RMS), kurtosis, and energy randomness are then extracted across multiple analysis scales to detect transient fault signature. Four case studies, including bearing fault, rotor imbalance, stator inter-turn faults, and attached bearing and rotor fault, were conducted to validate the method. Numerical results show that the proposed IEEMD-MSA approach attains an average fault identification exactness of 95.7%, advances signal-to-noise ratio by up to 18.5%, reduces mode mixing by 20.3%, and enhances feature firmness by 15.2% compared to outdated EEMD-based methods. These results establish strong adaptability, high sympathy, and reliability of the proposed method for composite mechanical fault diagnosis under noisy operating conditions.