Objective Permanent ground displacement from strike-slip faults poses a serious threat to buried oil and gas pipelines. It can induce local plastic deformation that leads to tensile fracture or buckling failure, thereby compromising pipeline structural integrity. Accurate prediction of pipeline deformation under strike-slip fault displacement is therefore essential. Existing analytical methods struggle to capture the nonlinear interaction between pipelines and surrounding soil and the elastic–plastic behavior of the pipes. Finite element modeling (FEM) is complex and often fails to strike an effective balance between computational accuracy and efficiency.

Methods Taking into account shortcomings of previous research, a method for solving pipeline deformation under strike-slip fault displacement based on a Physics-Informed Neural Network (PINN) is proposed. A Deep Neural Network (DNN) is constructed to map axial and lateral pipeline displacements within the axial spatial coordinates of pipelines. Using the principle of virtual work and the Green–Lagrange strain measure, a nonlinear soilspring model simulates axial and lateral nonlinear interactions between pipelines and the surrounding soil, while a linear-hardening constitutive model characterizes the pipe’s nonlinear material behavior. A physics-informed loss function is formulated to express pipeline deformation under strike-slip fault displacement. A boundary condition loss function is established, further integrating mechanical characteristics of pipelines under fault displacement and enforcing Dirichlet boundary conditions. The combined formulation enables high-precision prediction of pipeline deformation under strike-slip fault displacement.

Results In a case study based on a 1,016 mm pipeline section crossing a strike-slip fault zone, the PINN calculation results for crossing angles of 30°, 45°, 60°, 75°, and 90° under a fault displacement of 2.5 m were compared with the reference values from FEM calculations and found to be closely aligned. Subsequent quantitative error analysis yielded the following ranges of mean squared errors (MSE) for the PINN calculations: 3.82×10⁻⁷–1.85×10⁻⁶ for axial displacement and 1.73×10⁻⁶–6.18×10⁻⁶ for lateral displacement. For strain metrics derived from the PINN results, the axial tensile-strain MSE ranged from 1.20×10⁻¹⁰–6.06×10⁻⁹, and axial bending-strain MSE ranged from 8.01×10⁻¹⁰–2.73×10⁻⁸. To broaden the comparison conditions, comparisons were also made with international representative methods for solving peak pipeline strain under strike-slip fault displacement. Results showed that the PINN results remained consistent with the FEM results under different load conditions, with an average relative error of 2.88％, demonstrating its significant superiority to other methods for solving peak strain.

Conclusion The proposed PINN method exhibits significant advantages in computational efficiency and accuracy, providing a new approach to maintaining pipeline structural integrity under fault displacement.