Abstract— <p>Machine learning is increasingly used to predict material behavior in scientific fields, outperforming traditional numerical techniques. In this study, an Artificial Neural Network model was integrated into a finite element formulation to establish the creep strain law of metallic materials, considering creep time, stress level, temperature, and multiaxial ductility factor. First, the Neural Network’s structure and principles were presented, and its ability to infer creep-strain derivatives without prior learning was highlighted. After selecting a 2-hidden-layer architecture, the trained model was implemented into Abaqus via a CREEP subroutine. A similar Artificial Neural Network structure estimates multiaxial ductility factor considering relevant factors which was used for a multiaxial creep condition. To validate the model, it was compared with the analytical Wen-Tu model for Sanicro25 alloy. The model’s predictive ability was demonstrated through uniaxial and small punch creep test simulations. Results suggested the Artificial Neural Network can replace analytical creep strain models in finite element codes and is competitive in simulation time.</p>