Bell-and-spigot jointed ductile iron pipelines can accommodate differential ground settlement through both pipe-barrel bending and localized joint rotation, but these discontinuous kinematics are difficult to resolve with conventional continuous models. This study proposes a multi-domain physics-informed neural network (Multi-domain PINN) for rapid response prediction and quantitative response-regime identification of these pipelines. By decomposing the pipeline into subdomains according to joint locations, the proposed model captures the continuous deformation of pipe barrels together with the localized rotational discontinuities at flexible joints. Trained on high-fidelity three-dimensional finite element simulations validated against a full-scale field test, the model achieved R 2 values of 0.9842 and 0.9307 for vertical displacement and joint rotation, respectively. A joint-dominance fraction is introduced as a unified angular-deformation index that converts joint rotation and pipe-barrel bending into comparable measures of deformation participation. FEM observations first identified two fundamental response regimes: pipe-bending-dominated and joint-rotation-dominated responses. The trained Multi-domain PINN was then used as a physically interpretable surrogate for dense parametric analysis, yielding a statistically derived response-regime threshold of β Jc = 0.542 within the investigated parameter space, close to the mechanically neutral value of 0.5. Within the investigated parameter space, the two settlement-center conditions show the clearest systematic separation in the joint-dominance fraction, while the other investigated parameters mainly modulate the degree of joint participation and the proximity to the transition boundary. The proposed framework provides an efficient and physically interpretable surrogate for rapid prediction and quantitative mechanism assessment of the investigated buried jointed ductile iron pipeline system subjected to differential settlement.
No takes yet. Share an insight, caveat, or question.
Zhao et al. (2026) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: