Depth detection of guardrail posts based on guided waves and sparse Bayesian learning

Guardrails can prevent vehicles from crashing into the opposite lane, thereby reducing the risk of fatalities. However, the burial depth may be insufficient when the underlying soil is hard. The detection of the burial depth of guardrail posts mainly relies on the impact-echo method. According to the time-of-flight of the bottom echo, the burial depth is calculated based on wave velocity. However, since the frequency of guided waves cannot be precisely controlled by hammer impact and guided wave velocity is related to the frequency, the impact echo method suffers from insufficient accuracy for detecting burial depth. Therefore, this paper utilizes transducers to replace the impact hammer to excite guided waves. An automatic identification method is developed based on the Sparse Bayesian learning to compensate for the dispersion of guided waves and to recognize the bottom echo. The burial depth is determined based on a Bayesian model that utilizes multiple transducer pairs without prior knowledge of material properties and wave velocities. The detection accuracy is examined for different burial depths in simulations, experiments, and field tests. Simulation results show a mean relative error of 3.23% and experimental results show a mean relative error of 8.43%. The proposed method promotes the maintenance of highway infrastructures.