Intelligent generative design of TPMS scaffolds via BPNN-PSO bidirectional mapping for mechanical and hydrodynamic properties

The architectural design of implantable scaffolds represents a critical determinant, as it necessitates sufficient mechanical robustness and optimal porosity to facilitate cellular infiltration and nutrient transport. To address this challenge, a back propagation neural network (BPNN) was employed to capture the complex nonlinear mapping relationship between structural parameters and macroscopic performance. Subsequently, the particle swarm optimization (PSO) algorithm was utilised to solve the inverse problem by identifying optimal structural parameters that ensure adequate mechanical robustness while concurrently maintaining a favourable porosity essential for biological functionality. The optimised parameters were then employed to fabricate triply periodic minimal surfaces (TPMS) scaffolds. These scaffolds were subsequently subjected to mechanical compression testing, stiffness measurement, and microscopic analysis of PLA-HAp crystallinity morphology and of fracture crack propagation patterns. Additionally, in vitro cell adhesion assays were performed to evaluate cellular attachment and spreading behaviour on the scaffold surface. This computational framework synergistically optimises mechanical and mass transport properties, overcoming trial-and-error limitations and offering a new paradigm for load-bearing tissue scaffold design.