Denoising Algorithm for Acoustic Emission Signals of CFRP Bogie Frame Damage Based on SSA-ICEEMDAN-NLM

To address the challenge of noisy acoustic emission (AE) signals generated during the damage evolution of carbon fiber composite train bogie frames, this paper proposes a novel denoising algorithm. This method combines Sparrow Search Algorithm-optimized Improved Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (ICEEMDAN) with Non-Local Means (NLM) filtering. First, SSA is employed to optimize key ICEEMDAN parameters—specifically, the noise amplitude weight and the number of additions—using minimum envelope entropy as the fitness function. This approach resolves the empirical and suboptimal parameter selection issues inherent in traditional methods. Subsequently, the intrinsic mode functions (IMFs) obtained from the SSA-optimized ICEEMDAN decomposition are analyzed. Noise-dominated IMFs are identified and discarded based on sample entropy and spectral characteristics, while low-correlation components are further separated using correlation coefficients. Finally, the NLM algorithm is applied to denoise the selected effective IMFs. The denoised signal is reconstructed from these processed IMF components.Simulation and real-world AE signal analysis demonstrate that compared to ICEEMDAN-Wavelet Threshold and Variational Mode Decomposition-Wavelet Packet Threshold methods, the SSA-ICEEMDAN-NLM algorithm achieves superior denoising performance in both Signal-to-Noise Ratio (SNR) and Root Mean Square Error (RMSE) metrics. The proposed algorithm more effectively preserves critical signal features, providing a reliable foundation for subsequent signal analysis and damage identification in bogie frames.