Stress concentration factor (SCF) is an important parameter for the fatigue design of offshore joints. There are many empirical equations for quick estimation of SCF in tubular joints, based on experimental and numerical investigations. However, most of these equations apply at the crown and saddle points only, even though the maximum SCF may not always occur at these points, resulting in overestimated fatigue life. As the maximum SCF location varies due to multiplanar loads, damage, or reinforcement of joints, its location and magnitude are critical for a realistic fatigue life estimation. However, conventional statistical tools cannot approximate the complex behavior of SCF around the brace axis. On the other hand, artificial neural networks (ANN) can efficiently approximate complex phenomena. This study uses ANN to develop empirical models for determining SCF around the weld toe of KT-joints subjected to in-plane bending (IPB) loads. Eighteen hundred and fifty-eight (1858) designs were simulated using finite element analyses to generate data for training the ANN. Two IPB load conditions were focused on, and empirical equations were proposed for SCF around the chord side of the central brace-chord interface. These equations approximate maximum SCF with less than 5% error. This methodology applies to other joints and load configurations also.
Read full abstract