LAUSR

Abstract Acoustic emission (AE) is a promising technology for structural health monitoring to reduce costs and improve performance but hard to interpret. This paper puts forward an advanced deep neural network (DNN) method to dig information between AE parameters and performance of the structures since raw data obtained by the AE structural monitoring system is nearly useless without reinterpretation. Firstly, this study carried out a series of AE monitoring to catch crack signals of the prestressed concrete samples of cycle loads under three-point bending tests. The AE parameters analyzed include A mean , RA , LE , LQ , A max , EA , V e , and AK . The combination of them gives a comprehensive structural performance evaluation by the DNN model. The results show a good correlation between AE parameters with the crack activities of different loads. A mean , RA , LE are the critical parameters in detecting whether the structure meets the design strength requirement. The combination of all these parameters gives an overall assessment of the structures. Finally, the method applies to the monitoring of two real-world bridges and gets comprehensive evaluations for all the monitoring points. This paper not only addresses a systematic experiment and a comprehensive parameter analysis method to realize data interpretation, but also practical applications in real-world bridges.