Structural health monitoring in a heavy noise environment: A Hybrid ICEEMDAN-FDD Approach for Signal Processing

This paper presents a robust approach for analyzing noisy signals in structural health monitoring (SHM) via improved complete ensemble empirical mode decomposition with adaptive noise (ICEEMDAN) combined with frequency domain decomposition (FDD). The method's effectiveness is validated through two case studies: a numerical model of the ASCE-AISC benchmark structure and an experimental study of the Bahman Bridge at Ferdowsi University of Mashhad. In the ASCE benchmark study, the approach is tested on small-magnitude damage scenarios under varying levels of Gaussian noise. ICEEMDAN extracts intrinsic mode functions (IMFs) from the signals, which are then processed via FDD to estimate the modal parameters. The Bahman Bridge study applies the method to ambient vibration data, where a sensor malfunction introduces significant noise into the output. The noisy signal is analyzed via the proposed approach, and the results are compared with those from healthy sensors. The findings highlight the robustness and versatility of the ICEEMDAN–FDD method over conventional techniques, such as EMD–FDD and FDD for SHM, across both numerical and experimental settings. This approach provides a promising solution for SHM under noisy conditions, enabling accurate modal identification.