Hybrid physics–machine learning approach combining Gaussian process regression and Sines-style invariants for multiaxial fatigue life prediction of S355 steel | Synapse