A software prototype is developed for assessing and updating the reliability of single-cell prestressed concrete box girders subjected to chloride-induced reinforcement corrosion. The underlying system model consists of two integrated sub-models: a condition model for predicting the deterioration state of the box girder and a structural model for evaluating the overall system reliability. The condition model is based on a dynamic Bayesian network (DBN) model which considers the spatial variation of the corrosion process. Inspection data are included in the calculation of the system reliability through Bayesian updating on the basis of the dynamic DBN model. To demonstrate the effect of partial inspections, the software prototype is applied to a case study of a typical highway bridge with six spans. The case study illustrates that it is possible to infer the condition of uninspected parts of the structure due to the spatial correlation of the corrosion process. Figure 1. Sections defining the element size of the deterioration model of the box girder. Each section consists of four elements: a deck plate, a bottom flange and two webs. All random variables describing the deterioration state of the individual elements at time are collected in a vector = ( ) representing the overall deterioration state of the box girder. The joint distribution ( ) of is computed based on a DBN model of the corrosion process, which considers the spatial correlation among the deterioration in individual elements. This model is described in more detail in Section 3. Following Straub and Der Kiureghian (2011), the deterioration state of the box girder is considered constant over a period Δ = 1 year. Conservatively, the deterioration state of the box girder in the period [ − Δ , ] is set equal to the state at time , . The event of structural collapse of the box girder in that time interval is denoted by = ( ). The probability of structural collapse conditional on the deterioration state of the box girder at time , Pr( | = ), is computed by means of a probabilistic structural model of the box girder. This model is described in more detail in Section 4. The overall probability of structural collapse in the period [ − Δ , ] is given by the total probability theorem as: Pr( ) = Pr( | = ) ( ) (1) Information on the deteriorating box girder obtained through inspections is included in the calculation of the probability of structural collapse through Bayesian updating of ( ) on the basis of the DBN model. The updated probability of structural collapse of the box girder is given by Pr( | : = : ) = Pr( | = ) ( | : ) (2) where ( | : ) is the updated joint distribution of the deterioration state of the box girder and : = : is a set of all inspection outcomes in the time period [0, ] related to the deterioration state of the box girder.
[1]
Daniel Straub,et al.
A Computational Framework for Risk Assessment of RC Structures Using Indicators
,
2006,
Comput. Aided Civ. Infrastructure Eng..
[2]
Daniel Straub,et al.
Stochastic Modeling of Deterioration Processes through Dynamic Bayesian Networks
,
2009
.
[3]
Cur.
STATISTICAL QUANTIFICATION OF THE VARIABLES IN THE LIMIT STATE FUNCTIONS
,
2000
.
[4]
Daniel Straub,et al.
Reliability Acceptance Criteria for Deteriorating Elements of Structural Systems
,
2011
.
[5]
Sebastian Thöns,et al.
Intelligente Bauwerke: Prototyp zur Ermittlung der Schadens- und Zustandsentwicklung für Elemente des Brückenmodells
,
2015
.
[6]
Jianjun Qin,et al.
Risk Management of Large RC Structures within Spatial Information System
,
2012,
Comput. Aided Civ. Infrastructure Eng..
[7]
Daniel Straub,et al.
A framework for the asset integrity management of large deteriorating concrete structures
,
2009
.
[8]
Peter Norvig,et al.
Artificial Intelligence: A Modern Approach
,
1995
.
[9]
Stuart J. Russell,et al.
Dynamic bayesian networks: representation, inference and learning
,
2002
.