PolyGraphCL: A Multiview Graph Contrastive Learning Framework for Grain-Level Fatigue Damage Prediction in Polycrystalline Materials

Accurately predicting fatigue damage at the grain scale in polycrystalline materials is challenging, primarily due to the complex microstructural topology, anisotropic deformation, and severe class imbalance caused by the rarity of slip-band-marked damage events relative to the vast population of intact grains. Conventional machine learning (ML) methods and single-view graph neural networks (GNNs) often lack the capacity to model such heterogeneity across scales. To bridge this gap, we introduce PolyGraphCL, a novel multiview graph contrastive learning (CL) framework integrating heterogeneous inductive biases from three backbones—graph convolutional network (GCN) for localized neighborhood aggregation, graph attention network (GAT) for globally attentive interactions, and graph sample and aggregate (GraphSAGE) for multiscale sampling. These diverse structural views, derived from applying different GNN architectures to the same input graph, are fused through a learnable attention mechanism, enabling dynamic weighting of view-specific representations per node to capture both fine-grained and holistic structural characteristics. To further address extreme label imbalance, we incorporate cross-view CL that aligns intranode representations across views while repelling internode embeddings, facilitating the formation of class-discriminative manifolds. Evaluated on a ferritic steel microstructure dataset comprising 7633 grains (311 damaged) with 100 descriptors per node, PolyGraphCL achieves an average ${F}1$ score of $0.8816~\pm ~0.0505$ and balanced accuracy (BA) of $0.7788~\pm ~0.1606$ under stratified fivefold cross-validation-surpassing both conventional ML baselines and single-view GNNs. Furthermore, GNNExplainer-based attribution reveals that PolyGraphCL’s predictions are predominantly governed by local stress concentration, with moderate influence from topological substructures, offering interpretable insights grounded in underlying physical mechanisms. Altogether, PolyGraphCL offers a robust, interpretable, and domain-adaptive framework for advancing data-driven fatigue prediction in computational materials science (MS).