A novel DDDAS architecture combining advanced sensing and simulation technologies for effective real-time structural health monitoring

A still open question is how to design smart and autonomous mechanical structures able to perform online monitoring of their integrity and take anticipated actions during service before downtime of failure occur. To address this question, we present in this Chapter some innovative physics-guided numerical techniques for effective data assimilation and control. These techniques are integrated into a global feedback loop where the mechanical structure in service continuously and dynamically interacts with a digital twin which is updated and enriched on-the-fly for further command synthesis, as per the Dynamic Data Driven Applications Systems (DDDAS) paradigm. The overall strategy is unified around the concept of modified Constitutive Relation Error (mCRE) for enhanced robustness. The strategy also involves multidisciplinary numerical tools, such as Kalman filtering, adaptive modeling, deep learning, or Model Predictive Control, in order to accommodate real-time, accuracy, and portability constraints. The interest of the DDDAS approach is illustrated on several applications with evolutive mechanical systems, and with various sources of data obtained with advanced experimental devices.