Use of monitoring extreme strain data for failure probability analysis of concrete bridge girder: An optimal vine Gaussian-copula model

A critical part of ensuring the safety and functionality of structural systems in light of inevitable degradation and aging is periodically observing and monitoring the changes in the materials, properties and conditions of structural systems over time. This process, otherwise known as structural health monitoring (SHM), has been applied to a wide range of infrastructures, including bridges and skyscrapers, to predict impending structural damage, locate the damage, and identify the type of damage and its severity. A typical SHM process consists of structural components, sensors, data acquisition, transfer and storage mechanism, data management and data interpretation technologies.

Thousands of SHM systems have been deployed in different structures globally. Importantly, these systems leverage the capabilities of advanced sensory data acquisition, signal processing, data management and analysis technologies to ensure operational reliability of structural bridges. For bridge girders, the correlation between the output variables can be used to express the performance associated with the failure modes of the monitoring points. Considering such a correlation, reliability assessment of the bridges by computing the failure probability of the girder using the monitored data has drawn significant research attention.

Previous research on bridge reliability has mainly focused on using existing reliability analysis methods to analyze the system/component reliability based on general resistance and load effect information, an approach with several drawbacks. For instance, the copula function is commonly used to perform data-based bridge reliability, taking into account the correlation of the monitoring points. However, this approach becomes more complex and prone to limitations with an increase in the number of monitoring points. Thus, developing a reliable alternative approach is highly desirable.

To overcome the existing problems, Associate Professor Yuefei Liu and Associate Professor Xueping Fan from Lanzhou University presented a new method for analyzing the failure probability of a bridge girder. In their approach, the long-span bridge girder was considered a research project. To begin with, the optimal regular vine model was constructed by considering the correlation among the multiple monitored variables. This model was built based on the monitored daily extreme strain data at multiple monitoring points. Next, the failure probability of the bridge girder system was analyzed by combining the first-order second-moment method and bivariate Gaussian copula model with the proposed optimal regular vine model. Their work is currently published in the journal, Structures.

The authors reported that the proposed optimal regular vine analysis method exhibited strong performance with the ability to solve complex ergodicity problems associated with high dimensional random variables. The optimal model was used to describe the maximum correlation among the multiple monitoring variables as well as the correlation among the performance functions associated with the failure models of the monitoring points. Failure probability with and without considering the effect of correlation of the performance functions was analyzed and compared. It was reported that the failure probability obtained by considering the correlation of the performance functions was remarkably less than that found without considering such a correlation. This suggests the importance of such correlation in improving structural health monitoring of bridge structures.

In summary, Professors Liu and Fan reported the failure probability analysis of a bridge girder using a newly developed optimal vine Gaussian-copula model based on monitored extreme data. This approach addressed most of the limitations of the existing methods. The feasibility and potential practical applicability of the proposed model and methodology were validated by using the monitored extreme strain data of the bridge girder. In a statement to Advances in Engineering, the authors noted that their study provided new insights into failure probability analysis that would contribute to advanced SHM of bridges.