Filtern
Dokumenttyp
- Beitrag zu einem Tagungsband (2)
- Zeitschriftenartikel (1)
- Vortrag (1)
Schlagworte
- Bayesian inference (1)
- Bayesian updating method (1)
- Brückeninfrastruktur (1)
- Concrete bridges (1)
- FE modelling (1)
- Innovative Messverfahren (1)
- Load identification (1)
- Model update (1)
- Risk analysis (1)
- Stereofotogrammetrie (1)
Organisationseinheit der BAM
Eingeladener Vortrag
- nein (1)
In this work, a method for vehicle load identification is proposed. It involves Bayesian
Analysis and (quasi-)static importance functions in order to estimate vehicle positions,
velocities and weights. The structure is modeled with finite elements in order to generate
model predictions for different load configurations. The model predictions are compared to
the actual measured data to identify the most probable loading configuration for that
measurement. This involves the use of enhanced Monte Carlo simulations such as MCMC to
reduce the computational effort.
In der aktuellen Projektphase liegt der Fokus auf der Datenaufnahme, -bearbeitung und -speiche-rung mit dem Ziel, automatisierte Auswerteverfahren einsetzen zu können. Aktuell wurden primär punktuelle Messungen an ausgewählten Messtagen aufgenommen. Die Systeme sollen so weiter-entwickelt werden, dass sie sich auch für kontinuierliche Messungen im Rahmen von Monitoring-aufgaben eignen.
Ein wichtiger Fokus bei der Auswertung ist die Kombination mit numerischen Modellen, die mithilfe von Bayesian Update Verfahren und den aufgenommenen Messdaten kalibriert und im Verlauf der Monitoringaufgabe angepasst werden sollen. Insbesondere sollen auch zeitabhängige Modelle, die eine zeitliche Entwicklung von Struktureigenschaften beinhalten (Kriechen, Schwinden, Ermüdung, Korrosion) dazu verwendet werden, die zukünftige Performance der Struktur bewerten zu können. Basierend darauf werden dann Konzepte zur Planung von Inspektion und Wartung erstellt.
In this work, a probabilistic framework for identification of traffic loads on concrete Bridge structures is presented using data from a FE structural model in combination with a finite volume approach for traffic load modelling. The identification approach uses Bayesian Inference to identify traffic loads from measured sensor data from travelling load experiments performed at BAM. The work focuses on the load identification part of the Framework utilizing global structural response measurements only. The obtained information on traffic loads can be forwarded to further analysis such as fatigue and structure state estimation or model updating.
Growth of vehicle traffic density can be observed in many countries all over the world. This accretion is caused by world-wide population growth on the one hand, but also by increasing freight volumes and, thus, freight transports on the streets on the other hand.
This increased exposure becomes more and more of a problem for the civil infrastructure such as bridges. Many of these bridges are nowadays stressed to their limits by higher loads than they were originally designed for and/or operating times beyond the initially predicted life span. This raises questions about structural safety and lifetime prediction, of course, and therefore illustrates the need for accurate structural monitoring.
Since the lifetime of bridge structures is primarily influenced by their traffic loading, an accurate identification of load configurations over the whole length of a structure is most desirable.
In this paper, a method for vehicle load identification is proposed. It involves Bayesian Analysis and (quasi-)static importance functions in order to estimate vehicle positions, velocities and weights. The structure is modeled with finite elements in order to generate model predictions for different load configurations. The model predictions are compared to the actual measured data to identify the most probable loading configuration for that measurement. This involves the use of enhanced Monte Carlo simulations such as MCMC to reduce the computational effort. The measured data from different kinds of sensors can (and should) be combined for accuracy gain – in this case a combination of measured displacements and inclinations.
Since the measurements take place over some time during the passage of the vehicle, these estimations are carried out for several time instants for which the estimation is carried out. The advantage of using Bayesian Updating Method is the embodied learning effect leading to an improvement of the estimation when adding new information in a new calculation step.
Using the estimates for the loading conditions of a bridge structure as well as measurements of the structural responses, Bayesian analysis is again used in order to estimate localized structural parameters such as Young's modulus or Moments of Inertia in form of probability density functions yielding most probable values for the parameters.
Considering the difficulties for load identification close to the support poles of the bridge and therefore for the proposed structural parameter identification procedure, it is clear that this problem is ill posed. Bayesian regularization methods also have proven to be very effective when handling ill posed problems.