Project Details
Projekt
Automatic data-driven modeling and H2/H-infinity- norm-based dimension reduction of process-oriented and cooperative systems for SHM condition analysis with methods of system identification and machine learning on exposed structures - Phase two
Applicant
Professor Dr.-Ing. Armin Lenzen
Subject Area
Structural Engineering, Building Informatics and Construction Operation
Term
since 2022
Project identifier
Deutsche Forschungsgemeinschaft (DFG) - Project number 501664543
The SPP2388 deals methodically and experimentally with the condition assessment and condition estimation, in particular of bridge structures, in the context of the lifespan of a mechanical structure. The methodology is known as Structural Health Monitoring (SHM). The central element is a digital twin, as an image of reality, which records all structure-specific information via the combination of data acquisition, data assimilation, system space and model-oriented monitoring. This integral and digital information is directly suitable for embedding in the Building Information Modeling (BIM) concept. In phase one, ADMO, an automatic data-driven modeling system based on numerical methods of linear algebra and system identification, was introduced as a process-oriented state estimator in the micro-time domain. Machine learning (ML) methods take into account the operating and environmental conditions (EOC). The output-only method for condition estimation, damage detection and localization has been successfully verified experimentally and methodically using the I4S technical center and a real bridge. In phase two, the focus will be on prognosis models that provide mechanically and physically oriented and predictive recommendations for action that can be derived from the existing theoretical process models and state estimators in the micro and macro time domain over the service life. The parameters and damage indicators of the output-only identification are method-related relative and system-theoretical reference variables of an idealized stochastic input process. Therefore, these model parameters must be transferred into robust mechanically and physically scaled prediction models in the macro-time domain of the service life. The resulting absolute mechanical-physical condition indicators can then be interpreted in the form of established engineering-oriented characteristic parameters, which can provide meaningful support to the construction owner with predictive-oriented decision-making based on condition estimators for maintenance. An iterative methodical approach and the associated verifying, robust method design is accompanied by experiments on real mechanical structures and bridge structures.
DFG Programme
Priority Programmes
