PSI - Issue 64

Philipp Kähler et al. / Procedia Structural Integrity 64 (2024) 1248–1255 Kähler / Petryna / Structural Integrity Procedia 00 (2019) 000 – 000

1255

8

(a)

(b)

Fig. 7. (a) Time evolution of the ensemble mean values of the mass matrix entries, (b) result of detected system changes, darker blue indicates bigger changes in the mass, red numbers indicate poor localization of the added mass values. 4. Discussion and conclusions An SHM approach with a combination of two ensemble-based KF for the state and model parameter estimation was introduced in the present contribution along with a continuous load monitoring process using CNNs. It was shown on a representative example that the cluster structure of the CNNs can identify all relevant load parameters quite well provided a sufficient measurement data is available. Furthermore, a KF update scheme was developed and thoroughly tested on a laboratory structure. The proposed approach can estimate the state and model parameters even under uncertain starting conditions. Both approaches, the data driven load monitoring and the data assimilation process using ensemble-based KF, complement each other in order to develop a DT for SHM. Acknowledgements The financial support of the German Research Foundation (DFG) within the priority program SPP 2388 100+, Project LEMOTRA (No. 501728141) is gratefully acknowledged. References Askari, M., Li, J., Samali, B., 2016. Application of Kalman Filtering Methods to Online Real-Time Structural Identification: A Comparison Study. International Journal of Structural Stability and Dynamics, No. 6, DOI: 10.1142/S0219455415500169 Avci O., A delja er, O., iran az, ., ussein, ., Ga ouj, ., Inman, ., . A review of vi ration ased damage detection in civil structures From traditional methods to achine earning and eep earning applications. echanical stems and ignal rocessing. No. 7, OI . /j. mssp. . 7 77 urgers, G., van eeuwen, . J., vensen, G., 99 . On the Anal sis cheme in the nsem le alman Filter. onthl eather eview. No. , OI . 75/ 5 9 ( 99 ) < 7 9 A IT > . .CO; hrendorfer, artin, 7. A review of issues in ensem le ased alman filtering. eteorologische Zeitschrift. No. , OI . 7/ 9 9 / 7/ 5 Goodfellow, I., engio, Y., Courville, A., . eep earning. The IT ress. I N 97 5 Grieves, ., 5. igital Twin anufacturing xcellence through Virtual Factor eplication. hite aper amill, T. ., hita er, J. ., . nsem le ata Assimilation without ertur ed O servations. onthl eather eview. No. 7, OI . 75/ 5 9 ( ) < 9 A O> . .CO; Lan, Z., Yin, B., Wang, T., Zuo, G., 2017. A non-intrusive load identification method based on Convolutional Neural Network. IEEE Conference on Energy Internet and Energy System Integration (EI2) 2017 aes, ., Gillijns, ., om aert, G., . A smoothing algorithm for joint input state estimation in structural d namics. echanical stems and ignal rocessing No. 9 , 9 9, OI . /j. mssp. 7. . 7 chillings, C., tuart, A. ., . Anal sis of the nsem le alman Filter for inverse pro lems. IA Journal on Numerical Anal sis. No. 55, OI . 7/ 5959X Xie, L., Zhou, L., Wan, C., Tang, H., Xue, S., 2018. Parameter Identification for Structural Health Monitoring with Extended Kalman Filter considering integration and noise effect. Journal of Applied Sciences No. 12, DOI: 10.3390/app8122480