On the assessment of robustness

Abstract

A framework for assessing robustness is proposed, taking basis in decision analysis theory. Robustness is assessed by computing both direct risk, which is associated with the direct consequences of potential damages to the system, and indirect risk, which corresponds to the increased risk of a damaged system. Indirect risk can be interpreted as risk from consequences disproportionate to the cause of the damage, and so the robustness of a system is indicated by the contribution of these indirect risks to total risk. A framework is presented for measuring robustness in this way, and implications for system modelling and acceptable levels of robustness are discussed. Numerical studies of idealized structural systems are performed using this framework, to demonstrate the use of the proposed robustness index and provide insight into system properties affecting robustness. Considered exposures include the design live load and an extraordinary exposure representing a fire or explosion that causes the loss of one or more system components. The results indicate that properties affecting system reliability, such as number of components or the stochastic properties of the load, also affect robustness. Perhaps more interestingly, it is seen that properties such as failure consequences and time to repair a damaged system also affect this measure of robustness. The assessment framework is applied here to study damage tolerance, but the procedure can be applied as well to other aspects of robustness such as tolerance to human error in design or construction.