Structural Test Design with Value of Information

Abstract

In order to assess the value of information of an experiment through a pre-posterior decision analysis for a specific structure, detailed analyses of the structural system performance and all relevant costs and benefits associated to the decisions through the entire life-cycle of the structure are carried out. These analyses result in expected utilities for all modeled decision options. Considered in the life-cycle analyses are: design, operation and maintenance, repair, replacement, monitoring, test costs, and finally the benefit of operating the structure. These contributions to the life-cycle benefits and costs are weighted against their probability of occurrence. With the probabilities of damage and failure computed by structural models, the direct and indirect risks along-side the expected benefit are quantified. The damage and failure probabilities are computed in structural reliability analyses that utilize structural models. The prior reliability of the structural models is updated with (pre-) posterior knowledge from experiments. Probabilistic models for the information acquisition strategies (1) pre-construction proof load testing, (2) hybrid simulation, and (3) recorded service experience are introduced accounting for the type of the information, their precision and costs. Owners, operators, and designers of structures face the same issue to find optimal decisions related to design, operation and maintenance, retrofitting, inspection and monitoring, or replacement. These decisions influence the structural safety and economic benefit provided by the structure. One of the main challenges in this complex environment is to spend limited resources in the optimal way for the previously listed decision options. These decisions influence each other and may be relevant at vastly different points in time. The design of the structure is at the very beginning, even before the service-life begins, and retrofitting or replacement will be issues arising later. Examples of modeling such interactions of structural design and decision making at various stages of the service-life are shown. Value of information (VoI) constitutes a theory that indicates if a test or experiment is expected to provide a higher utility through increased expected benefits due to subsequent informed decisions or by the reduction of risks, the required resources and monetary expenditures. By applying the VoI theory and the concept of expected value of sample information (EVSI) based on the Bayesian decision theory by Raiffa & Schlaifer (1964) to measurement information about the condition of structures, a decision maker is able to assess information and actions in an objective manner before information acquisition and action are performed. This thesis discusses methods to acquire knowledge about the condition of structures through the means of proof load testing in conjunction with structural health monitoring (SHM), recorded service life data about the structure, and experiments not directly conducted on the structure in question. Such off-structure experiments analyzed herein are hybrid simulations that interactively combine physical and numerical models in order to capture the influence of interactions of structural elements. The new concepts of pre-construction proof load testing and hybrid simulation as substitute to proof load testing are demonstrated in case studies. Pre-construction proof load tests may be advantageous if applied to structures that are usually not considered forconventional proof load tests. Reasons that can prohibit a conventional proof load test can be difficulties or the impossibility to artificially load the structure with a sufficiently high proof load, or the risk arising from such a test. Both reasons can apply to offshore structures. The conceptual application of proof load testing sub-groups from the structural system in conjunction with a suitable system model is demonstrated. One can obtain such information similar to a proof load test of an existing structure. Another concept that may help to acquire structural health information is the use of hybrid simulations in order to substitute proof load tests or long term SHM. The feedback-coupled combination of physical and numerical model of hybrid simulations can help to gather knowledge about a structure that cannot be obtained with other methods. Because the experiment is not conducted with the actual structure, no information about the strength realizations of load carrying components is acquired. However, the advantage of not directly testing the structure is that no damage can be inflicted to it, is a relevant factor in the hybrid simulation value of information. The concepts of pre-construction proof load testing, hybrid simulation as a substitute for proof load testing, and the new analysis concept ‘expected value of sample information and action analysis’ (EVSIA) are demonstrated with case studies. EVSIA is an extension to EVSI by including actions that are only possible with posterior knowledge. The case studies consider hypothetical decisions about the optimal test strategy in order to obtain information about the structural health for an improved reliability assessment, and decisions regarding action on structural design or use of the structure. Results of the case studies indicate that: • Higher VoI in an EVSI analysis is achieved through higher proof loads and larger tested fractions of the structure. By using the (pre-) posterior information for design improvements, the EVSIA analysis shows that the value of information and action becomes insensitive or almost independent of the proof load and can yield even higher value than an EVSI analysis. • Structural testing before construction requires system models that account for the separated elements. • Proof load tests usually provide high VoI to systems of high complexity i.e. constituted by many components spreading out and lowering the risk of damage.