Application of ANN for Estimating Time-Variant Structural Reliability of Reinforced Concrete Structures Using Approximate Approach

Abstract

Deterioration of RC structures with the passage of time due to aggressive environment leads to reduction of strength, stiffness, and reliability of the structure. Generally, for assessing reliability or failure probability of aging structures that are time-dependent, one must consider the uncertainty in the structural degradation in association with nonstationarity in the load distribution process. In this paper, an approximate approach has been used for evaluating the effects of structural degradation and live-load variations on time-dependent failure probability of aging structures. In order to assess the time-dependent failure probability, the service life of the structure is considered to be 50 years. Due to heavy computational requirements that need a long and tedious completion process, soft computing, e.g., artificial neural networks (ANNs), have been used, which act as a physical system for rapid prediction of the structural reliability with reasonable accuracy. The predicted outputs obtained from the neural network model were validated with the actual outputs and were found to yield good results, making it suitable for estimating long-term time-variant behavior of aged structures under a significant loading process.