Condition assessment of prestressed hollow-core slab bridges based on 1D CNN using acoustic emission technique

To address the issues of low data processing efficiency, high computational complexity, and insufficient accuracy in damage detection of prestressed hollow slab bridges based on acoustic emission monitoring, this paper proposes a prestressed hollow slab bridge condition assessment method combining 1D convolutional neural networks (CNN) and acoustic emission technology. Firstly, to process AE data accurately and in real time, a 1D CNN model adapted to the AE time domain signal is built through the design of different network structure models. Secondly, the whole process from damage expansion to structural failure is recorded and analysed by static loading test of the scaled-down structure of the investigated prestressed hollow-core slab beam to provide the complete database for the 1D CNN model. Finally, in order to further improve the accuracy, transfer learning (TL) is proposed to apply the data extended actual bridge data to the 1D CNN model. Experimental results indicate that the proposed method can efficiently and accurately assess the condition of prestressed hollow slab bridges and precisely classify the stages of the damage process.