PSI - Issue 64

F.-B. Cartiaux et al. / Procedia Structural Integrity 64 (2024) 285–292 Cartiaux / Structural Integrity Procedia 00 (2019) 000 – 000

291

7

an unsupervised approach, and enables us to distinguish, from among all the responses recorded on a structure, classes of events that are more or less critical with regard to the selected descriptors. Continuous monitoring of a structure then enables us to observe the evolution of these different classes over time (number of individuals in each class, predominant descriptors, average individual and standard deviation, etc.). To define these descriptors, we propose to make extensive use of the continuous wavelet transform, with algorithms to extract the structure's modal parameters (modal frequency, modal damping, modal deformation). A study was carried out on the wavelet and its time-frequency resolution, as well as on the ridge extraction algorithm, to build dynamic descriptors robust to ambient stresses. The first stage is to reorganize each dynamic signal considering a single record, supposed representative of one vehicle. For each signal, the reconstruction of CWT was done (Fig. 7a and Fig. 7b). Then collecting all information from the two bridges, (AV and AM), to reduce the dimension of our {Xi} vectors, we use a principal component analysis (PCA), which will enable us to better visualize our data set and distinguish, between the {Xi} vectors, the most extreme signals to analyze them. For the outcoming research, we immediately detect a significative difference between the two bridges during one month of measurements (November 2023) as released on Fig. 7 (c).

Downstream Deck Upstream Deck

Acceleration (g)

Frequency (Hz)

Time (s)

Time (s)

First Principal Component

Second Principal Component

Fig. 7. (a, left) Single dynamic signal and (b, center) its CWT solution; (c, right) Results on the two Jules Verne bridges during November 2023 (PCA, second principal components, versus the primary one) 5. Acoustic emission campaign The acoustic emission (AE) campaign was realized by a team of the laboratory of acoustics at University of Le Mans within one day, July 5, 2022. The aim was to gather data from an on-site concrete structure under real traffic conditions to apply findings already made on laboratory tests in a previous action (Mandal et al., 2022). The latter corresponds to laboratory experiments, where the applied load (strain) is controlled and can be known during the performed mechanical tests. In the case of the on-site measurements the load is exerted by moving vehicles through their wheels, hence not controlled. We take therefore advantage of the data collected by the strain sensors deployed on the bridge to correlate them with the acoustic emission data. The typical AE data and variation in strain due to the passage of a vehicle is shown in Fig. 8 (a). It should be noted that the AE sensors were attached near to the second group of strain sensors; hence the appearance of AE hits is well synchronized with the strain data of the second group. Since the AE data recorded during the passage of a vehicle is well contained within the timespan of the recorded strain, a pass wise analysis of AE data was performed. Fig. 8 (b) shows the evolution of the average frequency (AF) as a function of RA-value (Mandal et al., 2022). This type of presentation, which is often used for crack type identification, has been analyzed in the case of known truck loads. A higher RA is found to be linked with heavy truck loads, whereas a higher AF is observed for light trucks. It is well known that tensile micro cracks are linked with high AF and low RA, whereas low AF and high RA indicate shear cracks (Mandal et al., 2022). This analysis shows that the effects of a heavy or light vehicle passing over a concrete surface can be different and well identified, and that links with micro-mechanisms within the concrete (tensile/shear) are possible. In this way, data relating to strain measurements can be reinforced with acoustic emission to better quantify and differentiate the effects of applied stresses on the micro-structure of the materials involved (Mandal et al., 2024).