Hilbert-Huang Transform Application in Acoustic-Vibration Signal Detection of Voids in Concrete-Filled Steel Pipes

Literature Overview

The paper by Han Xi, Xiang Li, Zhong Li, Yang Jin, and Chen Hui, published in the Journal of Chongqing Jiaotong University (Natural Science Edition) in 2013, presents a non-destructive evaluation (NDE) methodology for detecting voids between the steel pipe and concrete infill in concrete-filled steel tube (CFST) members. The research was funded under the Chongqing Science and Technology Key Project (CSCT2009AB6135). The core innovation lies in the application of the Hilbert-Huang Transform (HHT) to transient impact acoustic-vibration signals, followed by improved K-means clustering for binary pattern recognition. This work is particularly relevant to engineers involved in the quality control of CFST structures, where internal voids compromise load transfer efficiency and structural integrity.

Core Technical Approach

The methodology proceeds through three principal stages: signal acquisition under transient impact loading, feature extraction via HHT decomposition, and classification through clustering analysis. The Hilbert-Huang Transform comprises two sub-procedures: Empirical Mode Decomposition (EMD) and Hilbert spectral analysis. EMD adaptively decomposes the non-stationary, nonlinear acoustic-vibration signal into intrinsic mode functions (IMFs), each representing a distinct oscillatory component. The Hilbert spectrum then maps instantaneous frequency and amplitude information for each IMF, enabling extraction of time-frequency localized features.

Signal Acquisition and Feature Extraction

The transient impact is applied to the surface of the CFST member, generating stress waves that propagate through the steel pipe wall, the concrete core, and any void interfaces. Voids create impedance discontinuities that reflect and scatter waves, producing characteristic echoes in the received signal. The instantaneous amplitude features extracted from the HHT decomposition serve as the input vector for the subsequent classification step. This approach overcomes the limitations of traditional Fourier-based methods, which assume stationarity and cannot adequately represent the transient, multi-frequency nature of impact response signals in heterogeneous composite structures.

Improved K-Means Clustering for Pattern Recognition

The improved K-means algorithm addresses the well-known sensitivity of standard K-means to initial centroid selection. In the context of void detection, the algorithm classifies the extracted feature samples into two clusters: one representing sound (fully filled) sections and another representing defective (void-containing) sections. The improvement likely involves iterative centroid reinitialization or fuzzy membership weighting to enhance convergence reliability. The binary classification output provides a practical engineering decision: whether a given section of the CFST member contains voids requiring remediation.

Technical Parameters and Comparison with Alternative Methods

Engineering Practice Integration

From a quality control perspective, this method offers a non-invasive, in-situ inspection capability for CFST members after construction. In large-scale infrastructure projects such as bridges, towers, and offshore platforms where CFST members are extensively used, the ability to detect internal voids without destructive testing represents a significant advancement. The method aligns with the principles outlined in relevant Chinese standards such as GB 50935-2014 (Code for Design of Concrete-Filled Steel Tubular Structures) and JGJ/T 191-2010 (Technical Specification for Concrete-Filled Steel Tubular Structures), which mandate quality verification of concrete infill density.

The practical implementation requires calibration with known-good and known-defective specimens to establish reference feature patterns. Engineers should consider the influence of impact energy level, sensor coupling quality, and environmental noise on signal fidelity. A systematic approach following the PDCA cycle—Planning the inspection protocol, Doing the signal acquisition and analysis, Checking results against acceptance criteria, and Acting on identified defects—ensures methodological rigor.

Key Reflections and Insights

The selection of instantaneous amplitude as the primary feature is judicious, as it directly reflects the energy distribution of the impact response, which is fundamentally altered by the presence of voids. However, one must consider whether additional features such as instantaneous frequency centroids or energy ratios between specific IMF bands might improve classification accuracy, particularly for small or partially filled voids. The study demonstrates feasibility, but field validation under varying concrete grades, steel pipe wall thicknesses, and void geometries remains essential for widespread engineering adoption.

This work exemplifies the convergence of signal processing theory with structural engineering practice. For metallurgists and welding engineers involved in CFST fabrication, understanding the acoustic signature of internal defects provides complementary insight to traditional NDE methods such as ultrasonic testing (UT) and radiographic testing (RT), which are more commonly applied to weld joints and pipe walls. The HHT-based approach extends quality assurance to the post-assembly stage, covering the concrete infill quality that is otherwise difficult to verify in-situ.

In conclusion, the HHT-based void detection methodology presented in this paper represents a practical advancement in the non-destructive evaluation of concrete-filled steel tube members, offering engineers a data-driven, adaptive approach to internal defect identification that complements existing inspection protocols and contributes to the overall reliability of CFST structures in critical infrastructure applications.