Abstract: To ensure the efficiency of post-tensioning tendons in bridge structures during the long-term service lifetime, it is critical to evaluate the grouting quality in post-tensioning tendon ducts. Due to the concealed positions and complex geometries, it is not an easy task to accurately detect the grouting defects in the tendon ducts. To this end, this study proposes a grouting compactness evaluation method based on deep learning of ultrasonic signals. Firstly, piezoelectric ceramic (PZT) transducers are embedded to generate and collect ultrasonic signals propagating along the tendon ducts. Then, wavelet packet transform is applied to obtaining the multi-scale time-frequency features of ultrasonic signals for different grouting cases. Finally, a convolutional neural network deep learning model is established to extract grouting defects related time-frequency features of the signals and to classify different grouting cases. Through finite element numerical simulation of the post-tensioning tendon ducts, the distribution of wave field in different grouting cases with partial grouting and cavity defects was investigated. The simulation results show that the proposed method is feasible and efficient for grouting compactness evaluation in post-tensioning tendon ducts.
Keywords: post-tensioning tendon duct; grouting compactness; ultrasonic signal; wavelet packet transform; deep learning
合肥工业大学学报自科版