Resumen de AI-Powered Models: Catalysts for Digital Twin Advancementsin Operations & Maintenance
A strategy for the development of accurate yet dynamic and time-evolving engineering models is presented in this work. These models are deemed the backbone of the Digital Twins (DTs), representing the core intelligence that empowers them to fulfil their objectives acrossdifferent domains and industries. Models form the dynamic heart that allows diagnostics and damage assessment, predictive maintenance, and overall operational enhancement, with data,analytics and visualization serving as enablers of the models’ efficacy.Artificial intelligence (AI)-based models and specifically, physics-informed neural networks (PINNs,) emerge as an optimal model architecture owing to their ability to integrate the learning capacity and high performance of data-driven models such as neural networks (NNs), while incorporating the physics theory and constraints into the learning process. This integration effectively captures complex physical behaviours, thereby enhancing model precision and interpretability.The challenge related to deploying a suitable model architecture and gathering sufficient data in quality and quantity to accurately perform its training is overcome in the present research. A model-assisted training approach is employed, capitalizing on generative models which assume a pivotal role in training the predictive models tailored to handle the DT functions.
