Project Details
Projekt
Automated data-driven damage detection
Applicants
Privatdozent Dr. Stefan Bosse; Professorin Dr. Carmen Gräßle; Professor Dr. Dirk A. Lorenz
Subject Area
Measurement Systems
Image and Language Processing, Computer Graphics and Visualisation, Human Computer Interaction, Ubiquitous and Wearable Computing
Image and Language Processing, Computer Graphics and Visualisation, Human Computer Interaction, Ubiquitous and Wearable Computing
Term
since 2020
Project identifier
Deutsche Forschungsgemeinschaft (DFG) - Project number 418311604
In this subproject, we develop robust data-driven approaches for automated damage detection in composite materials using guided ultrasonic waves suitable for off-line and on-line distributed material- integrated diagnosis systems. For this, we take intercrossing paths of model-based, model-assisted and model-free methods with the objective of fusing the methods in a joint framework. As model-based we consider methods utilizing the underlying physical model equations (analytical models involving differential equations), their discretization together with suitable time stepping schemes as well as reduced model equations using projection-based intrusive model reduction. As model-assisted we consider methods utilizing information from the physical, discretized or reduced model equations together with information from measurement data. These include Bayesian inversion methods, data assimilation techniques and physics informed neural networks (PINNs). As model-free we consider methods which are based purely on data and free of physical model equations such as inference approaches, neural networks and machine learning techniques learning input-output relations. Our research strategy consists in developing a framework fusing these three different methodical paths with the aim to achieve a thorough understanding of the wave propagation, its interaction with damages and to ultimately use this framework to detect damages. On this way, we aim to build junctions between the approaches in order to exploit the respective advantages of the methods.
DFG Programme
Research Units
