This study investigates and compares properties of various P-type Triply Periodic Minimal Surface (TPMS) porous structures for bone scaffold design. At first, six cases of homogeneous single/dual-layer structures, axial single/dual-layer gradient structures and radial single/dual-layer gradient structures with the same average porosity are developed. Dual-layer gradient structures are selected for further design due to more similar pore and stress distributions to human bones, reduced maximum stress, higher yield strength and greater variations in yield strength and elastic modulus (E). The mechanical and permeability properties of ten cases of axial and radial dual-layer gradient structures with the same overall porosity but different inner and outer layer porosities are then further designed and studied. The results show that yield strength is within 112.75–139.97 MPa, E ranges from 11.15 to 13.01 GPa, the permeability (K) falls within 1.51–10.01 × 10−9 m2 and the average wall shear stress (WSSavg) varies between 6.18 and 9.11 mPa. The yield strength, E and K of radial dual-layer gradient structures are higher and WSSavg is lower than those of axial dual-layer gradient structures. Moreover, with increase in inner average porosity () and decrease in outer , the yield strength, E and K gradually decrease while WSSavg gradually increases for both types of structures. In particular, the radial dual-layer structure with the lowest porosity of 27.5% in the inner layer and highest porosity of 42.5% in the outer layer has superior mechanical and permeability properties. The findings offer direct guidance for the structural design of bone implants, enabling performance customization for different applications.
27 August 2024 in Journal of Materials Engineering and Performance
For high-dimensional expensive multi-objective optimization, the irregular Pareto front shape is difficult to completely capture due to the complex mapping between mass design and objective spaces, and the traditional evolution mechanism neglects the varying real-time optimization states of different objectives, leading to search imbalance under limited expensive evaluations. This paper proposes the objective potential-driven collaborative evolution (called OPCE) algorithm. Specifically, based on objective values, the individual sets corresponding to dominant and poor objectives are determined, and the differential evolution mutation operation is integrated to construct an objective potential-driven classification collaboration strategy to generate candidate offspring that adapt to objective potential along evolutionary direction of balancing the conflicts between objectives. To fully explore the local regions of candidate offsprings, using RBF-based gradient information, the dimension effect probability is derived for each candidate offspring. The dimension perturbation mutation operation is performed on high-effect dimension, generating diversified offspring along the high-effect gradient direction. Then, high-potential subpopulations are reconstructed via non-dominated sorting to suit each subproblem. Finally, two dual reference-driven local search problems are constructed with the derived neighborhood advantage expansion and disadvantage compensation functions as optimization objectives, and the derived neighborhood disadvantage and advantage limit functions as constraints, respectively. This achieves rapid local convergence for each subproblem while maintaining multi-objective balance. Experimental results on benchmark problems show that OPCE significantly outperforms mainstream surrogate-assisted evolutionary algorithms. Moreover, when applied to airfoil aerodynamic shape optimization, OPCE demonstrates superior comprehensive performance and robustness in balancing aerodynamic efficiency conflicts and dealing with expensive simulations.