Bayesian decision analysis for the optimization of inspection and repair of spatially degrading concrete structures. (1st October 2020)

• Record Type: Journal Article
• Authors: Vereecken, Eline; Botte, Wouter; Lombaert, Geert; Caspeele, Robby
• Is Part Of: Engineering structures. Volume 220(2020)
• Publication Date: 2020-10-01
• Subjects: Pre-posterior analysis -- Bayesian updating -- Value of information -- Spatial degradation -- Concrete structures; Structural engineering -- Periodicals; Structural analysis (Engineering) -- Periodicals; Construction, Technique de la -- Périodiques; Génie parasismique -- Périodiques; Pression du vent -- Périodiques; Earthquake engineering; Structural engineering; Wind-pressure; Periodicals; 624.105
• DOI: 10.1016/j.engstruct.2020.111028
• Languages: English
• ISSNs: 0141-0296

Abstract

Highlights: The pre-posterior framework is extended to spatially degrading structures. Comparison between inspection techniques, times and locations is enabled. Spatial correlation can be considered, e.g. by hyperparameters or random fields. Reduction of structural resistance in time can be accounted for. Abstract: Pre-posterior analyses can be used to determine the value of information (VoI) of inspection strategies. As such, decisions can be made on whether or not it is relevant to perform inspections and to what extent this is the case, before these are implemented. In this article, a pre-posterior analysis framework is developed for degrading (concrete) structures, accounting for time-dependent behaviour and spatial variability. To account for the time-dependent degradation, a deterioration model is implemented and failure probabilities and costs are evaluated at different points in time. In relation to the spatial nature of the degradation, the structure is discretized in zones and elements and spatial correlation is modelled by random fields and hyperparameters. Although the framework is generally applicable, it is applied to concrete structures subjected to corrosion. First, the framework is illustrated by a simple analytical example to demonstrate the calculation procedure and show how the distributions of hyperparameters and random fields in the corrosion model can be updated based on obtained inspection outcomes. More specifically, the information on non-inspected elements which can be extracted from measurements performed elsewhere in the structure is quantified. Finally, the framework is also applied to a case study in order to prove the feasibility to optimize the timing and location of inspections in relation to the corrosion state of a bridge girder, considering the VoI as a decision variable.