Optimized Variational Mode Decomposition Algorithm Based on Pre-Processing Algorithm

Existing optimization algorithms for Variational Mode Decomposition (VMD) are all based on multiple iterations, which pose a problem due to their large computational load, making them unsuitable for real-time fault diagnosis of engineering signals. This paper proposes an optimized VMD algorithm based on a preprocessing algorithm, which avoids iterative computation and thus improves computational efficiency. Firstly, the signal is Fourier transformed, and the spectrum is segmented based on the peak frequency of the signal, resulting in the number of decomposition layers after optimization. Secondly, based on the correlation between the quadratic penalty term and the signal noise, the optimal solution under different signal noises and different penalty terms is calculated, and the signal root mean square is used to describe the noise situation of the signal, constructing a penalty term optimization function based on signal root mean square. Finally, the optimization algorithm is validated using a diesel engine cylinder head simulation signal. The results show that the optimization algorithm can significantly reduce the computational load while ensuring the decomposition accuracy of the sine signal and avoiding mode mixing, with the computation time accounting for only 0.01% of the total time, providing a feasible optimization strategy for engineering applications.

Keywords: Signal processing of reliability test, Variational mode decomposition, Fault diagnosis, Fault feature extraction, VMD, Preprocessing algorithm, Spectral segmentation.