Detection of subsurface voids in concrete-filled steel tubular (CFST) structure using percussion approach. (30th November 2020)

• Record Type: Journal Article
• Authors: Chen, Dongdong; Montano, Victor; Huo, Linsheng; Fan, Shuli; Song, Gangbing
• Is Part Of: Construction & building materials. Volume 262(2021)
• Publication Date: 2020-11-30
• Subjects: Concrete-filled steel tubular structures (CFST) -- Void detection -- Support vector machine (SVM) -- Machine learning -- Percussion; Building materials -- Periodicals; 624.18
• DOI: 10.1016/j.conbuildmat.2020.119761
• Languages: English
• ISSNs: 0950-0618

Abstract

Highlights: A novel percussion technique to detect subsurface voids is proposed. Machine learning algorithms are applied in the percussion method. Power spectrum density is used to extract features for the SVM model. The average prediction accuracy of SVM model is 94.17%. The prediction accuracy of SVM is 3.94% higher than that of the decision tree. Abstract: Concrete-filled steel tubular (CFST) structures are essential load-bearing components in many civil engineering structures. Subsurface voids between the contacting surface of the concrete and steel in a CFST structure reduce the load-bearing capacity of the CFST structure. This paper presents a novel, non-destructive, percussion-based approach to detect subsurface voids in CFST structures. In our approach, we exploit the contrasting sound produced by the percussion of surfaces with and without subsurface voids. Percussive acoustic signals in non-void and void zones are recorded. By analyzing the power spectrum density (PSD) of the percussion sound, nine features can be extracted. Two specimens (A and B) were fabricated in our experiment. The features of the sound signal extracted from the specimen A are used as the database for training and testing the support vector machine (SVM) model. Then, the trained SVM is applied to specimen B to determine whether or not a void between the concrete core and the outer steel tube exists. The experimental results show that the prediction precision is 94.17%. Therefore, the percussion-based approach is a simple, efficient, and accurate method to detect the void defects.