A GRU‐Transformer With HLBO Based Method for Fatigue Life Prediction of Additively Manufactured Materials

ABSTRACT Accurate fatigue life assessment of materials is essential for ensuring the safety and reliability of critical engineering components. With the development of advanced materials and manufacturing processes, the construction of fatigue life prediction model for materials, especially additively manufactured materials, is difficult due to the complex fatigue evolution processes. Deep learning approaches show greater potential than traditional physical methods in capturing the complex, nonlinear relationships inherent in fatigue life prediction. This study proposes a fatigue life prediction method on the basis of an integrated Gated Recurrent Units (GRU) and Transformer architectures, combined with Hybrid Leader‐Based Optimization (HLBO). The proposed method is validated using a multi‐material fatigue dataset of additively manufactured components, primarily focusing on Ti‐6Al‐4 V titanium alloy, 316L stainless steel, AlSi10Mg aluminum alloy, and Inconel 718 nickel‐based alloy. Meanwhile, the proposed model was evaluated against several baseline models (Convolutional Neural Network (CNN), Multilayer Perceptron (MLP), Residual Network (ResNet), and GRU, Transformer) using multiple performance metrics, including MAE, MSE, R 2 , MAPE, and a comprehensive evaluation metric. Experimental results demonstrate that the GRU‐Transformer with HLBO based method outperforms all compared models in terms of accuracy and stability. The proposed model shows superior capability in handling complex fatigue data and offers a reliable tool for data‐driven fatigue life prediction under varied material and loading conditions.