Artificial Neural Network based Elastic Response Limit Analysis of 90° Back-to-Back Pipe Bends

The 90° back-to-back pipe bends is critical in altering the flow direction in pipelines, often experiencing significant damage from internal pressure and bending moments during operation. This study aims to determine the elastic response limits of such pipes under combined loading conditions using AI-based methods. Internal pressure and 360° bending moments were set as input variables, generating linear analysis datasets using Abaqus and Python script. ANN models were trained to predict the maximum equivalent stress and employed as a surrogate model to identify elastic limits. Iterative algorithms were used to combine load conditions within the yield strength, enabling accurate determination of elastic limits. This research highlights the potential of AI-driven structural integrity evaluation techniques in enhancing the understanding of stress distributions under complex loads, contributing to improved safety and efficiency in pipeline design.