PSI - Issue 64

Paul Winkler et al. / Procedia Structural Integrity 64 (2024) 1264–1270 Author name / Structural Integrity Procedia 00 (2019) 000 – 000

1268

5

4. Structure-borne sound detection during crack formation The cracks, which form due to the increasing loads locally release elastic energy scattered throughout the whole structure by high frequency content deformation waves. This acoustic emission (AE) is measured with a system supplied by QASS GmbH (2024) using two piezo ceramic sensors at a sampling rate of 1.6 MHz. In the anticipated region of crack formation, they are connected to the concrete surface using M5 brass threaded dowels such that AE waves are properly transferred. The low noise laboratory environment allowed for identifying distinct structure borne signals with frequencies up to several 100 kHz. A big challenge lies in choosing the proper preamplification, since the expected signal amplitudes span a wide range, which is unknown beforehand. Signals can be identified in several ways. Especially under noisy conditions it is useful to extract the typical high frequency contributions, which stand out to regular background signals. In the context of SHM, the next step after identifying AE signals is localizing their source and thereby potential structural damage. This can be done by means of time difference of arrival (TDoA) methods, which are used e.g. in GPS systems but require a precise way of determining the signal onset. A direct option is a triggering system with a predefined threshold amplitude. However, this can give very unreliable estimates for signals with low signal to noise ratios or flat initial increases. In Boniface et al. (2020) it is shown, that the Akaike criterion (AIC), which uses the change in signal variance, is much more robust in determining AE signal onset. For a signal s of N data points, on point k it is defined as: (1) A threshold is still used to roughly identify the time of a signal and subsequently the AIC function is computed in an interval of 3 ms around the first point of exceedance. The time of arrival is then set to the time where the function reaches the minimum value. This estimation is done for both sensors and since their signals are synchronized in one device, one can match compatible individual signals by determining the TDoA. Each time difference corresponds to a difference in distance from the source to the sensor locations. This difference is per definition constant on hyperbolic curves, which have their focal points at the sensor positions. They have a distance of 56 cm, which corresponds to a maximum time shift, which was measured with artificial signals to be about 170 mus. Larger time shifts will not be considered valid signals, since they cannot originate from the same source. The structure is divided into spatial zones corresponding to a specific interval of the arrival time difference and each event is assigned to one zone. The distribution of event numbers for loading steps at 60 kN and 120 kN are shown comparatively in Fig. 4a and 4b.

Fig. 4a: Distribution of event numbers for 60kN

Fig. 4b: Distribution of event numbers for 120kN