Description
Aging infrastructures, particularly concrete bridges, in most developed countries are growing, directly impacting their safety and longevity. Therefore, early detection of damages such as wire breakage is not only vital; it is urgent to prevent catastrophic structural failures such as the Polcevera bridge in Genova, Italy. Prestressing cable degradation in concrete bridges poses a significant risk to structural integrity. Therefore, effective monitoring and early detection of prestressing cable degradation are essential to securing the safety and stability of the bridges.
In response to this challenge, a novel approach for wire breakage detection, using Mel-frequency cepstrum coefficients (MFCCs) and a multilayer perceptron (MLP) is proposed. To validate the effectiveness of the proposed models, experimental data from two Italian bridges were collected. This dataset was used for training and testing the models. To address the limited real-world data, data augmentation techniques, including MixUp, time-shifting, and polarity inversion, were employed, leading to enhance the models' robustness and prevent overfitting. MFCCs extract features from acoustic emission signals and are classified by MLP.
The results demonstrate effective damage detection and classification, underscoring the MLP's potential for real-time bridge monitoring. The MFCCs-MLP combination advances wire breakage detection, while data augmentation addresses data scarcity challenges. This method holds promise as a robust, broadly applicable model for bridge monitoring, enhancing infrastructure safety and durability.
