PSI - Issue 62

Israel Alejandro Hernández-González et al. / Procedia Structural Integrity 62 (2024) 879–886 Hernández-González et al./ Structural Integrity Procedia 00 (2019) 000 – 000

885

7

Fig. 4. (a) Mode shapes of the Méndez-Núñez Bridge estimated by MTL-DNN, and (b) MAC matrix between the mode shapes estimates by CoV-SSI and the predictions of the MTL-DNN.

3.1. Continuous modal parameter identification from MTL-DNN predictions.

Fig. 5. Tracking of the resonant frequencies of the Méndez-Núñez Bridge from September 27th until October 17th, 2023. Filled and open dots represent the estimates by MTL-DNN and CoV-SSI, respectively.

In this section, the acceleration data collected from September 27 th until October 17 th , 2023, is processed to assess the effectiveness of the proposed MTL-DNN approach for continuous modal identification. This dataset comprises a total of 924 records, each one containing 30 minutes acceleration records acquired continuously. The architecture of this network is defined with the same hyper-parameters and the same identical architecture. In this case, the first two 30-min acquisitions (360,000 samples) were considered in the training dataset. The subsequent automated post processing involves the extraction of the complex mode shapes and modal characteristics using the ITD approach. Finally, to track the 14 physical modes throughout the whole monitoring period, a simple modal tracking algorithm