As the core load-transfer medium in bonded structures, the adhesive layer critically governs overall reliability, with Mode I fracture representing its dominant failure mechanism under tensile loading. This study systematically compares the eXtended Finite Element Method (XFEM) and its two coupled variants—the XFEM-Cohesive Zone Model (CZM) and XFEM-Virtual Crack Closure Technique (VCCT)—in simulating Mode I fractures of adhesive joints. Key comparisons include predictions of stress distribution, load-transfer evolution, and crack propagation paths, all validated through Double Cantilever Beam (DCB) simulations and experiments. Results show that standard XFEM accurately predicts initial stiffness (error < 8%) but overestimates peak load by 10.7%. XFEM-CZM maintains errors below 8% for both stiffness and peak load, while XFEM-VCCT achieves exceptional peak-load accuracy (error < 1%) but overestimates stiffness. In crack evolution, standard XFEM yields an idealized propagation path, whereas the coupled methods reveal a distinct three-stage process. Stress/strain fields in standard XFEM remain stable during propagation, while the coupled approaches exhibit interfacial irregularities before crack arrival, followed by tip concentration and band-like transfer during stable growth. Each method offers distinct advantages, underscoring that selection should align with specific research objectives and modeling requirements.

The extended finite element method (XFEM) has been extensively employed in numerical simulations of fracture mechanics. Nevertheless, the standard XFEM still faces inherent challenges in solving complex problems, such as susceptible stiffness matrix ill-conditioning, insufficient local interpolation flexibility, and high sensitivity to mesh distortion. To address these inherent deficiencies, this study proposes an improved XFEM (IXFEM), which replaces the traditional crack-tip enrichment terms in standard XFEM with an enhanced radial point interpolation method (ERPIM) to fundamentally optimize the numerical performance of the XFEM. A systematic comparative analysis of various radial basis functions (RBFs) demonstrates that the developed polynomial-truncated multiquadric (PT-MQ) basis function maintains high interpolation accuracy while delivering superior numerical stability. Based on the PT-MQ basis function, the crack-tip enrichment formulation of standard XFEM is reconstructed to derive the ERPIM shape functions. Combined with an optimized stiffness matrix assembly strategy and the Level Set Templated Cover Cutting Method (LSTCCM) integration scheme, the numerical defects inherent in conventional XFEM are effectively eliminated. Systematic convergence and robustness tests verify that the proposed IXFEM exhibits excellent convergence performance and stable numerical accuracy even under complex working conditions involving random mesh distortion and drastic cross-scale variations of shape parameters. Three classical benchmark fracture examples further validate the proposed method, and the stress intensity factors (SIFs) predicted by IXFEM agree well with analytical solutions. Quantitative performance comparisons reveal that the proposed IXFEM achieves substantially higher computational efficiency than standard XFEM. Overall, the developed IXFEM realizes an excellent trade-off among computational accuracy, robustness, and computational cost, providing a competitive numerical framework for solving complex fracture problems.