ENHANCING TUBULAR STRUCTURE INTEGRITY ASSESSMENT: A MACHINE LEARNING-INTEGRATED FAILURE ASSESSMENT DIAGRAM APPROACH

Failure Assessment Diagrams (FADs) are a cornerstone of structural integrity analysis, but their complexity can limit their adoption in certain engineering contexts. This study introduces a novel machine learning approach designed to make FAD analysis more accessible and computationally efficient. The method leverages artificial neural networks (ANNs) trained on FAD models derived from rigorously calculated numerical contour integrals. Finite Element Analysis (FEA) was used to generate the FAD envelopes that is used to train the ANNs. The resultant ANNs demonstrate several capabilities: generating precise FAD envelopes, predicting limit loads, and incorporating the effects of residual stress via the Φ parameter. This work highlights the potential of ANN-based FADs as a powerful and adaptable tool, streamlining complex structural integrity evaluations without sacrificing accuracy.