Siltation of drainage pipelines is a major disaster-causing factor in urban flooding. Owing to critical bed shear stress and the possibility of settlement at the bottom of pipes, there are significant nonlinear differences in the hydraulic transport and retardation response characteristics of siltation particles under different initial flow velocities, siltation degrees, and siltation lengths. Studies on siltation in drainage pipelines lack a comprehensive understanding of the intricate coupling dynamics between siltation particles and water flow, and the regularity governing the siltation transport behaviour under varying water flow conditions remains unclear. To solve these problems, this study constructs computational fluid dynamics and discrete element (CFD-DEM) coupling method and adaptive penalty model for pipeline deposition based on probabilistic settlement function, adaptive genetic algorithm, and bidirectional long short-term memory neural network (PSF-AGA-BLSTM). A three-dimensional transient hydraulic model of pipeline siltation was adopted to clarify the critical bed shear stress of the silt particles, after which an adaptive genetic algorithm with a probability settlement function as fitness was constructed to generate a numerous silt particle sample sets with settlement result labels. Adaptive genetic algorithms were combined with bidirectional long short-term memory neural networks to establish a spatiotemporal adaptive penalisation mechanism for pipeline silt particle transport. The simulation model based on PSF-AGA-BLSTM and CFD-DEM was analysed using simulation and scale tests and was found to have strong robustness and reliability. The results showed that the effects of siltation degree, flow rate, and siltation length on the transport of silt particles were all non-negligible; when the siltation degree exceeded 0.2, the siltation degree had a more significant effect on particle transport than siltation length and flow rate. The transport of silt particles occurred in two phases, rapid growth and steady growth under erosion by water flow. The rate of change in silt length accounted for up to 45% of the reason for silt nudging and silt spreading. The study clarified the hydraulic transport law of silt particles and analyzed the effect of different influencing factors. These efforts provide technical support for subsequent pipe cleaning and maintenance as well as flood prevention and mitigation.