Title:
AI Based Surrogate Model for Digital Twins for Structural Health Monitoring
Author(s):
Cervenka
Publication:
Web Session
Volume:
ws_S24_Cervenka.pdf
Issue:
Appears on pages(s):
Keywords:
DOI:
Date:
3/23/2024
Abstract:
Artificial Intelligence (AI) using Artificial Neural Networks (ANN) is applied for real time fast response surrogate models in the digital twin concept for structural health monitoring. Digital twin is a modern concept, in which a digital replica of a real product or structure is developed, and the virtual twin i.e., numerical model, closely communicates, and exchanges data with the real structure. The digital twin method is used for the assessments of safety, durability and reliability of reinforced concrete structures. The surrogate model based on ANN is used in the digital twin approach for two purposes. The deep learning of the ANN surrogate model is using the sample data generated by sensitivity studies using the virtual twin, i.e. the numerical model based on nonlinear finite element analysis using the software ATENA (www.cervenka.cz/products/atena). First an ANN model is used in the calibration phase of the virtual twin. Once the virtual twin is calibrated to realistically simulate the real structural behaviour, it is applied for the physically informed deep learning of the ANN model for the fast response surrogate model to provide critical safety information for structural health bridge monitoring. The paper will present the development of efficient and accurate ANN based surrogate model and the physically informed deep machine learning methods. The presented approach is applied to two pilot bridges in Czech Republic as a part of a research project supported by Czech Technology Agency and Ministry of Transport CK03000023 “Digital twin for in-creased reliability and sustainability of concrete bridges”.