EUVIMEDEUVIMED entdecken
Open Access · CC BY 4.0

AI-Driven Digital Twin Framework for Adaptive and Real-Time Structural Health Monitoring of Offshore Marine Structures

James Riffat, Kourosh Nazari, Seyed Reza Samaei

Artificial Intelligence for Sustainable Cities · 25.03.2026 · 1 · 105-122

DOI: 10.65582/aifsc.2026.006 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Artificial Intelligence for Sustainable Cities, online erschienen am 25.03.2026. Verfasst von James Riffat und 2 weiteren Autor:innen. Die Arbeit wurde bisher 10-mal zitiert (Crossref) und nennt 37 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY 4.0).

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Reliable structural health monitoring of offshore jack-up platforms is challenging due to harsh environments, measurement uncertainty, and gradual damage over time. This study proposes an integrated framework combining digital twin technology with a machine learning–based damage identification pipeline for adaptive assessment of jack-up legs. A simplified model using a ten-element Euler–Bernoulli beam represents a 124 m leg with a fixed base. Synthetic datasets were generated by introducing random stiffness reductions in elements 3–8, covering single and multiple damage scenarios with severities of 5–20%. To improve realism, environmental variability and measurement uncertainty were incorporated through temperature variations between −10°C and +30°C together with multiple levels of simulated sensor noise. The signal processing workflow involved detrending, band-pass filtering, Fast Fourier Transform analysis, and adaptive peak detection to extract modal features, including natural frequencies and spectral entropy indicators. These features were used to train a four-layer multilayer perceptron implemented in . Model performance was evaluated using five-fold stratified cross-validation. The classifier achieved an accuracy of 41.0%, a macro-F1 score of 40.1%, and a ROC AUC of 0.8155 on the test dataset, indicating reliable discrimination between healthy and damaged structural states despite environmental variability and measurement noise. In parallel, an adaptive digital twin updating procedure was implemented to refine the numerical model using modal frequency discrepancies. This updating process reduced the root mean square error of frequency prediction from 0.0596 Hz to 0.0554 Hz, corresponding to a 6.98% improvement in predictive consistency between the numerical model and the simulated structural response. The results demonstrate that coupling machine learning based damage classification with digital twin model updating provides a practical pathway toward adaptive monitoring of offshore structures.

Zitieren

APA

Riffat, J., Nazari, K., & Samaei, S. R. (2026). AI-Driven Digital Twin Framework for Adaptive and Real-Time Structural Health Monitoring of Offshore Marine Structures. Artificial Intelligence for Sustainable Cities, 1, 105-122. https://doi.org/10.65582/aifsc.2026.006

Vancouver

Riffat J, Nazari K, Samaei SR. AI-Driven Digital Twin Framework for Adaptive and Real-Time Structural Health Monitoring of Offshore Marine Structures. Artificial Intelligence for Sustainable Cities. 2026;1:105-122. doi:10.65582/aifsc.2026.006

BibTeX
@article{Riffat2026,
  author = {Riffat, James and Nazari, Kourosh and Samaei, Seyed Reza},
  title = {AI-Driven Digital Twin Framework for Adaptive and Real-Time Structural Health Monitoring of Offshore Marine Structures},
  journal = {Artificial Intelligence for Sustainable Cities},
  year = {2026},
  volume = {1},
  pages = {105--122},
  doi = {10.65582/aifsc.2026.006}
}
Kostenlos starten →