Hollow Defect Detection in Steel Tube Concrete Using Time Series Analysis

Literature Overview

The research paper by Luo Yongpeng, Xie Longbo, Liao Feiyu, Wang Jingfeng, and He Hanqing, published in the journal Industrial Construction (工业建筑) in 2019, Volume 49, Issue 10, pages 48-53, presents an innovative approach to detecting hollow (void) defects in steel tube concrete members using time series analysis theory. Funded by the National Natural Science Foundation of China (Grants 51578154 and 51808122) and the Fujian Provincial Science and Technology Department, this work bridges the gap between signal processing theory and structural health monitoring practice.

Technical Background and Problem Statement

Steel tube concrete (STC) members are widely used in modern construction for their high load-bearing capacity, ductility, and construction efficiency. However, quality defects such as hollow spaces between the steel tube and concrete core can significantly compromise structural performance. These voids may result from inadequate concrete placement, poor compaction, or segregation during pouring.

Traditional non-destructive testing (NDT) methods for detecting voids in STC members include:

The time series analysis approach proposed in this paper offers a novel alternative that leverages dynamic response characteristics rather than direct wave propagation measurements.

Methodology: AR Model and Cosh Distance

Autoregressive (AR) Model Framework

The authors introduce the Autoregressive model as a parametric representation of the dynamic response signal. For a stationary stochastic process, the AR(p) model expresses the current value as a linear combination of p previous values plus a white noise term:

x(t) = φ₁x(t-1) + φ₂x(t-2) + ... + φₚx(t-p) + ε(t)

Where φ₁ through φₚ are the AR coefficients and ε(t) is the white noise innovation. The key insight is that the AR model parameters and the resulting residual variance are sensitive to changes in the structural configuration, including the presence of voids.

The methodology proceeds as follows:

1. Signal acquisition: Acceleration response signals are obtained from the STC member under dynamic excitation (impact or vibration).
2. AR model fitting: Separate AR models are fitted to signals from undamaged regions and regions suspected of containing voids.
3. Residual analysis: The residual sequences and their variances are extracted from both models.
4. Distance-based discrimination: Statistical distances between the residual distributions are computed to quantify the degree of anomaly.

Distance Metrics Comparison

The paper compares three distance metrics for constructing the discrimination function:

The Cosh Distance, defined as CD = ln(cosh(x - y)), demonstrates superior performance because it provides a smooth, convex measure that is robust to outliers and captures the full distributional differences between residual sequences. Unlike Euclidean distance, which treats all deviations equally, and Mahalanobis distance, which requires estimation of the covariance matrix (which may be ill-conditioned for short signal sequences), the Cosh Distance offers a computationally efficient and statistically sound discrimination criterion.

Numerical Simulation and Experimental Validation

Numerical Simulation

The authors conducted finite element simulations of STC members with and without circumferential void defects. The simulation results confirmed that the presence of voids significantly alters the dynamic response characteristics, leading to detectable differences in the AR model residuals. The simulation parameters included:

• Steel tube: Q345 steel, diameter 500 mm, wall thickness 12 mm
• Concrete core: C40 grade, density 2400 kg/m³
• Void configuration: Circumferential ring, varying thickness (10-50 mm)
• Excitation: Single-point impact at midspan

Experimental Testing

Dynamic tests were conducted on physical STC specimens with deliberately introduced circumferential void defects. The test setup included:

• Accelerometers mounted at multiple locations along the member length
• Impact hammer for controlled excitation
• Data acquisition system with sampling rate of 10 kHz or higher

The experimental results validated the numerical predictions and demonstrated that the Cosh Distance-based method can accurately identify both the presence and the extent of circumferential void defects.

Engineering Practice Implications

The proposed method offers several practical advantages for field implementation:

1. Non-contact measurement: Accelerometers can be attached to the steel tube surface without requiring access to the concrete core.
2. Quantitative assessment: The discrimination function provides a numerical value that can be used to rank the severity of detected defects.
3. Circumferential void detection: Unlike many traditional methods that struggle with detecting ring-shaped voids, this approach is specifically effective for this defect geometry.
4. Relatively low equipment cost: Standard accelerometers and data acquisition systems are sufficient, without requiring specialized ultrasonic or radar equipment.

However, several practical limitations should be acknowledged:

• The method requires prior knowledge of the undamaged dynamic response characteristics, which may need to be obtained from baseline measurements before defects develop.
• Signal-to-noise ratio in field conditions may affect the accuracy of AR model fitting.
• The method is optimized for circumferential voids and may require adaptation for other void geometries (local voids, longitudinal voids).

Key Reflections and Study Insights

This work represents a creative application of classical signal processing theory to a practical structural engineering problem. The choice of the Autoregressive model is particularly apt because it provides a parsimonious representation of dynamic signals while being sensitive to changes in the underlying system properties. The selection of Cosh Distance as the discrimination metric is insightful, as it overcomes the limitations of more commonly used distance measures in this specific application context.

The methodology aligns with the broader trend toward intelligent structural health monitoring, where data-driven approaches complement traditional physics-based NDT methods. However, it is important to recognize that this is not a replacement for conventional NDT but rather a complementary technique that can be deployed in scenarios where traditional methods are impractical or insufficient.

From a quality control perspective, this method could be integrated into routine inspection protocols for STC structures, particularly for bridges, high-rise buildings, and industrial facilities where STC columns are commonly used. The ability to detect and characterize circumferential voids is particularly valuable because such defects can severely compromise the load transfer mechanism between the steel tube and concrete core.

Summary

This paper presents a novel and effective method for detecting circumferential hollow defects in steel tube concrete members using Autoregressive modeling and Cosh Distance discrimination. The approach has been validated through both numerical simulation and physical testing, demonstrating superior performance compared to Euclidean and Mahalanobis distance metrics. For practicing engineers involved in structural inspection and maintenance, this method offers a practical tool for assessing the quality of STC members, particularly in situations where traditional NDT methods face limitations. The work exemplifies the value of interdisciplinary approaches in advancing structural engineering practice.