Infrastructures, Vol. 11, Pages 286: Value-of-Information-Based Material Testing for Existing Reinforced Concrete Members
Infrastructures doi: 10.3390/infrastructures11080286
Authors:
Vittorio Palma
Simone Celati
Agnese Natali
Walter Salvatore
Sebastian Thöns
This paper presents a predicted information and predicted action (PIPA) decision-analysis approach with Bayesian material updating for planning material testing in existing reinforced concrete members. The objective is to select the number of concrete and reinforcing-steel tests before testing is performed and before a management action is chosen, accounting jointly for the updated structural performance, information-acquisition costs, action costs, and expected failure consequences. Predicted future test outcomes are used to update the material-strength distributions; the updated distributions are then propagated through the shear, flexural, and system reliability analyses to inform outcome-dependent action selection. Management interventions are modelled as system-state actions through action-dependent system failure probabilities. The optimal testing option is identified by minimising a total predicted-information and predicted-action cost-and-risk measure that combines information and expected action costs with the expected consequences of the system states. The approach is applied to a benchmark reinforced-concrete member with transverse shear reinforcement and uncertain concrete compressive strength and reinforcing-steel yield strength, in which the same steel-strength population is adopted for the longitudinal and transverse reinforcement. The decision-optimal testing option consists of six concrete tests and three reinforcing-steel tests, reducing the total expected decision cost-and-risk measure by approximately 58.2% relative to the no-new-information decision. The decision value arises mainly from avoiding unnecessary intervention when favourable material information is acquired. The results formulate material-test planning as a decision-value problem, providing an alternative to fixed sample-size rules while retaining explicit dependence on structural, probabilistic, action, and cost assumptions.