A novel electromechanical impedance-based method for non-destructive evaluation of concrete fiber content. (10th October 2022)

• Record Type: Journal Article
• Authors: Yang, Ziqian; Gao, Weihang; Chen, Lin; Yuan, Cheng; Chen, Qingjun; Kong, Qingzhao
• Is Part Of: Construction & building materials. Volume 351(2022)
• Publication Date: 2022-10-10
• Subjects: Electromechanical impedance -- Fiber content -- Fuzzy entropy -- Gradient boosting decision tree algorithm -- Grid search cross-validation -- Spherical smart aggregate; Building materials -- Periodicals; 624.18
• DOI: 10.1016/j.conbuildmat.2022.128972
• Languages: English
• ISSNs: 0950-0618

Abstract

Highlights: This study proposes a novel EMI-based non-destructive method for evaluation of fibre content inside SFC structures. A concept of FuzzyEn is introduced for fusion of conductance and susceptance to construct sample features. The proposed method improves the limitations of the current methods that normally apply to near-surface area evaluation. Abstract: The uneven distribution of fibres in concrete is a key factor that affects the mechanical properties of the components/structures. Many approaches have been developed to detect the fiber content of concrete. However, most of the existing methods are expensive, destructive, and mainly suitable for near-surface areas. To evaluate fibre content inside the steel-fiber concrete (SFC), this study proposes an electromechanical impedance (EMI)-based non-destructive method using embeddable sensing technology. The proposed method employs the embeddable sensors to sample EMI siganls, the concept of fuzzy entropy (FuzzyEn) to extract the signal features and an ensemble machine learning algorithm known as the gradient boosting decision tree (GBDT) to classify different labels. To evaluate the fiber content inside SFC, firstly, lead zirconate titanate (PZT)-based smart spherical aggregates (SSAs) are embedded into the host components/structures with different fiber contents, and a series of admittance (reciprocal of impedance) spectra of the SSAs are sampled as the dataset. Subsequently, the conductance and susceptence (real and imaginary parts of the admittance spectra) are calculated using their corresponding FuzzyEn values as two sample features. Finally, with the help of the grid search cross-validation method, the GBDT classifier is trained to classify different labels corresponding to the fiber contents. An experimental study was conducted to validate the feasibility of the proposed method, and the results confirmed the ability of the method to evaluate fiber contents inside SFC structures. The embeddable sensing technology assisted method can significantly improve the limitations of the current methods that normally apply to near-surface area evaluation.