The project deals with the influences of secondary effects on the damage identification and localization of mechanical structures under ambient excitation (e.g. wind, waves, traffic, etc.). In addition to the usual static and dynamic loads, mechanical structures are also affected by environmental and operational conditions (EOC) such as temperature, humidity or motors of a ventilation system. In experimental laboratory tests, such influences can either be excluded or kept constant and therefore neglected. In contrast, for real mechanical structures, these influences are largely unknown in their interaction, their respective characteristics and their specific influence on the dynamic behavior. The environmental and operational influences are of significant importance for the structural monitoring of mechanical structures during operation. For example, changing environmental conditions can cause a change in the natural frequencies, which hinders or even makes it impossible to localize damage during a frequency analysis. The inclusion of EOC is essential for structural monitoring, as this is the only way to relate the externally induced fluctuations in structural properties to the changes in mechanical parameters. The problem lies in the fact that structural damage is recorded as a local change in the mechanical parameters (mass or stiffness), while the EOCs exert a global influence on these parameters (e.g. temperature-dependent modulus of elasticity). The EOCs are diverse in nature and are never fully known on real structures. Assuming a large number of EOC classes, the mechanical structure can be considered as a linear parameter-varying system (LPV). The automated condition identification of mechanical structures during the utilization phase is an active field of research in the area of structural monitoring.

Fig. 1: Change in natural frequencies with changing EOC conditions (α-factor) and damage.

A frequently used approach is to integrate the EOC exclusively via basic physical principles (e.g. via FEM models). This can be very time-consuming, if not impossible. In addition, this approach has the disadvantage that it is based solely on assumptions about the mechanical system and not on real findings.

Due to these circumstances, a measurement data-based approach to system identification seems advantageous. Data science methods, which are ideally suited to finding and quantifying complex relationships from large amounts of data, are ideal for this purpose. Data science methods offer various approaches, whereby one of the most promising approaches is classification. The algorithms belong to the supervised learning methods and derive background information from the patterns in the data in comparison to known data. The already published method for damage localization state projections estimation error (SP2E) has already been extended in simulation so that it is possible to include the EOC. This approach is based on output-only measurements of the structures under investigation as a result of ambient excitation. This means that the EOC is not taken into account subsequently via statistical analyses, but is included in the method in terms of system theory. These methods are to be expanded and stabilized for use on real structures and then verified experimentally.

[Translate to English:]

Das Einbinden der EOC erfolgt in drei Phasen. Zunächst wird auf Basis einer Output-Only Strukturmessung ein Normierungsmodell, beispielsweise mit bekannten SSI-Methoden, identifiziert. Die zweite Phase stellt eine Lernphase dar, in welcher die LPV-Systeme gemäß der auftretenden EOC in Klassen eingeteilt werden. Die dritte Phase repräsentiert die zu überwachende Lebensphase einer Struktur. Zuvor als maßgeblich identifizierte EOC werden fortlaufend gemessen und der Strukturzustand über Output-Only Identifikation überwacht. Mit den Referenzmodellen und den EOC-Klassen kann eine EOC-Prozessanalyse durchgeführt werden. Dazu wird ein EOC-Vektorraum gebildet, welcher die veränderlichen Eigenschaften abbildet und in der Überwachungsphase eine Schadenslokalisation mittels SP2E erlaubt.