Identification of Concrete Void Defects in Steel-Concrete Composite Columns Using Tapping Sound Signals and One-Dimensional Convolutional Neural Networks

Literature Overview

This paper addresses the detection of concrete voids and hollow defects within steel-concrete composite columns by leveraging tapping sound signals processed through a one-dimensional convolutional neural network. The research is situated within the broader field of structural health monitoring for composite construction, where steel tubes are filled with concrete to form hybrid structural elements. The steel tube material typically conforms to GB/T 3091 or GB/T 8163 specifications, with wall thicknesses ranging from 4 mm to 12 mm in common engineering applications. The core innovation lies in the signal processing methodology applied to acoustic responses generated when tapping the external steel surface, enabling non-destructive identification of internal concrete quality anomalies.

Core Technical Content and Methodology

The fundamental principle of the tapping method relies on the fact that a solid steel-concrete composite column produces a distinct acoustic signature when struck, whereas a column containing internal voids or unfilled regions generates a different frequency distribution and damping characteristic. The steel tube acts as a waveguide, transmitting mechanical vibrations generated by the impact through its wall thickness into the concrete core. When the concrete is fully compacted and bonded to the inner steel surface, the energy dissipation follows a predictable pattern. However, when voids exist—whether due to incomplete concrete pouring, segregation during placement, or inadequate compaction—the acoustic impedance mismatch at the void boundary creates reflections and resonance effects that alter the received signal.

The signal acquisition protocol involves placing a vibration sensor, typically a piezoelectric accelerometer with a frequency response range of 0.5 Hz to 10 kHz, at multiple points along the steel tube surface. Tapping is performed with a standardized rubber hammer or impact rod at controlled locations, ensuring consistent input energy across test points. The raw time-domain signal is then digitized at a sampling rate of at least 50 kHz to capture the high-frequency components associated with void boundaries.

The one-dimensional convolutional neural network is designed to extract spatial features from the time-domain signal without requiring transformation to the frequency domain. The network architecture typically includes multiple layers of one-dimensional convolution kernels with varying receptive fields, followed by pooling layers and fully connected classification layers. The training dataset comprises signals from columns with known void conditions, created through controlled concrete placement experiments where voids of specific dimensions and locations are intentionally introduced.

Steel Tube Material and Fabrication Considerations

From a metallurgical and fabrication standpoint, the performance of the tapping method is influenced by the steel tube material properties and manufacturing process. Seamless steel tubes (GB/T 8163) provide uniform wall thickness and consistent acoustic properties, making them ideal for baseline signal characterization. Electric resistance welded tubes (ERW, per GB/T 3091) introduce a weld seam that creates a localized acoustic discontinuity, which must be accounted for in signal interpretation. The weld seam in ERW tubes can produce a characteristic echo that may be confused with a void signal if not properly filtered.

Hot-finished welded tubes (HFW) offer improved mechanical uniformity compared to ERW tubes but still exhibit a longitudinal weld seam. For columns fabricated from longitudinal submerged arc welded tubes (LSAW), the weld zone is more substantial, and the heat-affected zone (HAZ) exhibits different microstructural characteristics that affect acoustic propagation. The HAZ microstructure, which may include martensite or bainite phases depending on the preheat temperature and cooling rate, has different elastic moduli compared to the base metal, creating additional acoustic impedance variations.

The steel grade also plays a role. Q235 and Q345 steels, commonly used for structural steel tubes, have well-characterized acoustic velocities (longitudinal wave velocity approximately 5900 m/s and transverse wave velocity approximately 3230 m/s). Higher-grade steels such as Q460 or Q690 exhibit slightly different acoustic properties due to their higher carbon equivalent and alloy content, which must be reflected in the signal processing algorithms.

Quality Control Implications and Engineering Practice

The application of this tapping-based void detection method has significant implications for quality control in composite column fabrication and construction. Traditional methods for detecting concrete voids within steel tubes include ultrasonic testing (UT), which requires access to both ends of the tube and is impractical for in-situ inspection of installed columns. The tapping method offers a practical alternative that can be performed on completed structures without disassembly.

From a welding quality perspective, the integrity of the steel tube itself must be ensured before concrete placement. Pre-fabrication inspections should include visual examination (VT) of the tube interior to confirm freedom from scale, rust, and contaminants that could impair concrete-steel bonding. The internal surface preparation is critical: residual mill scale acts as a release agent, preventing the mechanical interlock between concrete and steel that is essential for composite action. In practice, tubes should be shot-blasted to Sa 2.5 grade or wire-brushed to St 3 grade before concrete pouring, as specified in relevant standards such as JGJ 166 for steel-concrete composite structures.

The concrete mix design must also be considered in the context of void formation. Self-compacting concrete (SCC) reduces the risk of void formation compared to vibrated concrete, but its use in tall, narrow steel tubes can still result in segregation and void formation at the top if the pouring rate is not controlled. The slump flow of the concrete, air content, and aggregate size relative to the tube diameter all influence the likelihood of void formation.

Common defects identified through this method include:

• Localized voids near the top of the column due to inadequate filling
• Segregation voids along the tube length caused by excessive pouring height
• Bond voids at the steel-concrete interface caused by inadequate surface preparation
• Honeycombing voids caused by insufficient compaction energy

Key Questions and Reflections

Several important questions arise from this research that merit further investigation. First, the correlation between void size and acoustic signal change is not linear, and the minimum detectable void diameter depends on the signal-to-noise ratio and the specific steel tube geometry. Second, the method's effectiveness may be compromised in columns with complex cross-sections or internal reinforcement, where additional acoustic scattering occurs. Third, the long-term stability of the method must be considered, as the concrete properties change over time due to carbonation, moisture variation, and creep, potentially altering the baseline acoustic signature.

The integration of this method into routine quality control protocols requires the development of acceptance criteria and threshold values for signal classification. Engineering practice demands clear pass/fail criteria, and the probabilistic nature of data analysis classification introduces uncertainty that must be communicated to design engineers and inspectors.

Study Insights and Implications

This research represents a meaningful advancement in the non-destructive evaluation of steel-concrete composite structures. The combination of a simple, low-cost signal acquisition method with sophisticated pattern recognition provides a practical tool for field inspectors. For the steel pipe and welding industry, the key implication is that the acoustic properties of the steel tube become a critical quality parameter, not merely a mechanical property. This reinforces the importance of consistent steel tube manufacturing quality, proper welding procedures, and surface preparation in composite construction applications. The method also highlights the need for better understanding of steel tube internal surface conditions, which directly affect the composite column's structural performance and the reliability of non-destructive inspection techniques.