Title: Damage detection for long-span bridges through support vector machine, wavelet transform, and multivariate empirical mode decomposition

Authors: Rouzbeh Doroudi; Seyed Hossein Hosseini Lavassani; Mohsen Shahrouzi

Addresses: Civil Engineering Department, Faculty of Engineering, Kharazmi University, Tehran, Iran ' Civil Engineering Department, Faculty of Engineering, Kharazmi University, Tehran, Iran ' Civil Engineering Department, Faculty of Engineering, Kharazmi University, Tehran, Iran

Abstract: Structural health monitoring (SHM) is crucial for long-span bridges, yet the challenge lies in handling vast data and precise feature selection. In this study to address this, observer-teacher-learner-based optimisation is used to fine-tune support vector machine parameters. The wavelet transform (WT) is employed to effectively filter noise and unwanted frequencies from structural acceleration responses. Further, acceleration responses are decomposed into intrinsic mode functions (IMFs) using multivariate empirical mode decomposition (MEMD). These IMFs are then utilised for classification with statistical features as required data for damage identification through optimal SVM. Data from Tianjin Yonghe Bridge in China validates the model's ability to accurately detect damage, its location, and severity. This approach demonstrates a practical method for damage detection in large-scale bridges by extracting statistical time-domain features through WT and MEMD, presenting a valuable contribution to the SHM field.

Keywords: structural health monitoring; SHM; statistical time-domain features; bridge health assessment; damage classification; damage severity; feature extraction.

DOI: 10.1504/IJSTRUCTE.2024.138127

International Journal of Structural Engineering, 2024 Vol.14 No.2, pp.164 - 185

Received: 08 Apr 2023
Accepted: 23 Oct 2023
Published online: 29 Apr 2024 *

Full-text access for editors Full-text access for subscribers Purchase this article Comment on this article