Structural Reliability Assessment of Steel Four-Bolt Unstiffened Extended End-Plate Connections Using Monte Carlo Simulation and Artificial Neural Networks

Abstract

Connections account for a pivotal function in the safety of a structure. The structural design should be based on minimal probability of failure in its lifetime, which is addressed by the probabilistic methods for the evaluation of structural reliability. However, conventional reliability techniques, including Monte Carlo simulation (MCS), require plenty of time for cases with implicit limit state function (LSF). Accordingly, the present study aims to analyze the structural reliability of the steel four-bolt unstiffened extended end-plate connections using integrated artificial neural network (ANN) and MCS approaches. The ANN-based MCS exhibits a higher speed compared to the conventional application as the implicit LSF is estimated by the ANN model. The finite element modeling of the connections provides the data required for ANN training. Following the achievement of the LSF, the MCS method is used to assess the connection reliability. Considering a target reliability index of 3.5, a resistance reduction factor was obtained to be 0.82 which is smaller than the current system resistance factor used in AISC 358 connections.