Bridge damage detection using PCA-DWT with limited sensors

Structural damage detection is crucial for ensuring the safety of the bridge, yet traditional methods often rely on many sensors which limits their practical application due to high costs and complexity. A new approach combining Principal Component Analysis (PCA) and Discrete Wavelet Transform (DWT) offers a promising solution, enabling accurate damage localization even with a limited number of sensors.

 

A research team from Jinan University, Dongguan University of Technology, and Guangzhou Polytechnic University recently introduced this method, detailing its theoretical foundation, validation through simulation and experiment, and its effectiveness under various damage scenarios. The study showed how the method can reduce costs while improving detection accuracy, supporting broader implementation in real-world bridge health monitoring.

 

The team published their findings in Lifeline Emergency and Safety on Dec 10, 2025.

 

“Our method aims to extract damage-sensitive index from bridge displacement responses under moving loads, using only a limited sensor. By integrating PCA and DWT, we can effectively extract structural dynamic components and accurately locate single and multiple damage”, said Zhenhua Nie, corresponding author of the paper, professor in the College of Mechanics and Construction Engineering at Jinan University.

 

The team presents a clear methodology in which PCA is applied to displacement data to obtain principal components containing both modal shapes and dynamic responses. The first dynamic component is then separated using a moving average filter and analyzed by DWT to extract wavelet approximation coefficients. These coefficients are reconstructed into a signal, and a damage index (DIDC) is defined to identify and localize damage.

 

“When damage occurs, the DIDC shows a distinct peak at the damage location, even with two sensors. It is because PCA can effectively reduce the impact of the environment and extracts the dominant dynamic component, while DWT enhances sensitivity to local damage”, explained Nie.

 

The method was validated through numerical simulation of a simply supported beam bridge and experimental testing on a steel beam bridge under moving vehicle loads. Results demonstrated accurate damage localization for single and multiple damage scenarios under different damage severities, sensor numbers, and noise levels. The approach also showed robustness against Gaussian noise.

 

A simulation under different sensor configurations confirmed that the method remains effective even with reduced sensor numbers. “Our results indicate that damage can be reliably detected with only two sensors, which significantly lowers the cost and complexity of structural health monitoring systems”, again said Nie.

 

The research team expects the method to support more accessible and efficient bridge practices. “This PCA-DWT method is well-suited for real engineering applications where sensor deployment is limited. We plan to further optimize the algorithm for real-time monitoring and integrate it with sensor networks for large infrastructure”, said Nie.

 

Other contributors include Xiaojie Zheng, Honghong Li, and Hongwei Ma from Jinan University, Guangzhou Polytechnic University, and Dongguan University of Technology.

 

This work was supported by the National Natural Science Foundation of China under Grant No. 52178289.

---

---