Stochastic Modeling of Long-Term Degradation over Structural Lifetimes

Abstract

As structures and civil systems age, decision-making over structural lifetimes requires an understanding of the long-term degradation of these systems. Due to the extensive uncertainties in degradation processes, including the number of possible degradation modes acting on a structure, their effects, external factors, and the long time duration, estimating degradation is complex. Many previous studies investigate the impacts of individual degradation modes on reliability, providing insights into the forms of functions that can be used to estimate degradation due to these mechanisms. These have been used as a basis for more general structural degradation models through the use of, e.g., linear, polynomial, or exponential functions. These methods provide an estimate for degradation that can be used to estimate long-term reliability. However, these models can be limited in terms of the number of degradation mechanisms accounted for and often do not match known physical constraints of degradation. We propose a new stochastic model for long-term structural degradation. This model is based on the mechanical properties of individual degradation modes. The degradation at any instance is calculated as a random sum of a random number of degradation modes acting on the structure. The effect of a degradation mode is modeled as a stochastic function based on its mechanical effect. Individual degradation modes include those that start at time of initial construction and those beginning later in the lifetime of the structure. The effect of one degradation mode on the rate of another is also considered. We apply the proposed model to three years of field monitoring data and compare the resulting analyses with estimations from existing functions. How structural inspection data can be used to learn or update the parameters of the model is also described. The proposed model results in a more accurate estimation of long-term degradation, leading to improved predictions of system responses and supporting reliability-based decision-making over structural lifetimes.