Abstract: Objective Permanent ground displacements caused by strike-slip faults pose a significant threat to buried oil and gas pipelines, potentially inducing local deformations that may lead to tensile rupture or buckling failure, thereby severely compromising pipeline structural safety. Accurate prediction of pipeline deformation under strike-slip fault action is therefore critical. Method A method for solving pipeline deformation under fault dislocation based on a Physics-Informed Neural Network (PINN) is proposed. A deep neural network (DNN) is constructed to establish the mapping between axial and lateral displacements of the pipeline along the axial spatial coordinates. Using the principle of virtual work and von Kármán strains, combined with a nonlinear soil spring model to simulate axial and lateral nonlinear interactions between the pipeline and surrounding soil, and a linear-hardening constitutive model to characterize the nonlinear behavior of the pipe material, a physics-informed loss function for pipeline deformation under strike-slip fault action is formulated. Coupled with pipeline mechanical characteristics under fault loading and Dirichlet boundary conditions to define the boundary loss function, this approach achieves high-precision prediction of pipeline deformation caused by strike-slip faults. Results Taking a 1016 mm-diameter pipeline crossing a fault zone as an example, finite element (FE) model results are used as reference. Comparisons for crossing angles of 30°, 45°, 60°, 75°, and 90° with a fault displacement of 2.5 m show that the PINN model closely matches the FE results, with mean squared errors of displacement below 10⁻⁵ and mean squared errors of strain below 10⁻⁷. Further comparisons across multiple fault conditions and international methods for predicting pipeline strain peaks under strike-slip faults indicate that the PINN model consistently predicts strain peaks in agreement with FE results, with an average relative error of only 2.85%, significantly outperforming other strain peak prediction methods. Conclusion The proposed PINN approach demonstrates significant advantages in computational efficiency and accuracy, serving as a viable alternative to FE analysis and providing a novel method for assessing pipeline integrity under fault action.