Evaluation of bridge safety based on Weigh-in-Motion data

Probabilistic assessment of bridges has been the subject of various studies in recent decades. It has been widely agreed that evaluating an existing bridge according to the standards and codes used for new structures can lead... [ view full abstract ]

Probabilistic assessment of bridges has been the subject of various studies in recent decades. It has been widely agreed that evaluating an existing bridge according to the standards and codes used for new structures can lead to demolition of a safe bridge or unnecessary repairs, and thus to high economic cost and an increase in the associated environmental impact. This paper investigates several concerns, the sensitivities of and correlation between the different stochastic parameters influencing the load on a bridge and its resistance to that load. The usefulness of updating the bridge safety model using damage indicators from a Structural Health Monitoring system is also examined.

The proposed approach combines a number of aspects. Firstly, a probabilistic bridge load model is established based on Weigh-In-Motion (WIM) data to mimic a realistic traffic flow and hence, the loads and their effects on the bridge. Traffic loading is highly correlated as the same vehicles influence many parts of the bridge. This has a significant influence on the probability of failure.

To model the resistance of the bridge a probabilistic approach is used and full correlation between segments is assumed. Combining the load and resistance models, the probability of failure can be inferred. In the future work the bridge safety model, more precisely the resistance model, will be updated. Bayesian updating will be used in the current framework based on the information obtained from specific damage indicators.

This study aims at obtaining valuable information regarding the importance of the different aspects of bridge safety models and the sensitivity of the probability of failure (i.e. the level of safety) to them. It is also expected to confirm the applicability of a Bayesian approach to this problem.