Structural health monitoring of harbor caissons using support vector machine and principal component analysis. (October 2021)

• Record Type: Journal Article
• Authors: Bolourani, Anahita; Bitaraf, Maryam; Nekouvaght Tak, Ala
• Is Part Of: Structures. Volume 33(2021)
• Page Start: 4501
• Page End: 4513
• Publication Date: 2021-10
• Subjects: Structural health monitoring -- Harbor caissons -- Machine learning -- Support vector machine (SVM) -- Principal component analysis (PCA) -- Finite element (FE) modeling -- Autoregressive (AR) model; Structural engineering -- Periodicals; 624.1
• DOI: 10.1016/j.istruc.2021.07.032
• Languages: English
• ISSNs: 2352-0124

Abstract

Abstract: Harbor caissons are regarded as one of the most critical infrastructures, and any disruption in their operation would have dramatic economic consequences. The need for a health monitoring system for these structures is thus evident. The primary aim of this study is to develop a structural health monitoring (SHM) system to effectively detect damage in harbor caissons. To this end, through performing dynamic analysis, acceleration signals are extracted from the locations in the model corresponding to the presumed accelerometer's placement in an actual structure. Three levels of white Gaussian noises are added to the original signal to simulate the ambient noise. Using the acquired signals, the damage-sensitive features in both frequency and time domains are determined in the structure. In this study, leveraging Principal Component Analysis (PCA) and Support Vector Machine (SVM), the damage-sensitive features associated with the damaged and undamaged structure are reduced in dimensionality and classified. Hence, the introduced system can assess the state of the structure based on the input signal from the accelerometer and detect possible damage and its severity. The effectiveness of the proposed system in detecting damage is shown using the numerical model of a real harbor caisson.