Explotation of 3D-printed bone scaffolds architecture: a Design of Experiment (DOE) and computational fluid dynamic (CFD)-finite element method (FEM) framework ranking study

In silico analysis of tissue engineering products is an efficient tool to predict and design the optimal substitute in developing scaffolds. The primary objective of this study is to computationally rank a set of twelve predefined internal architectures of PCL polymer-based 3D-printed bone scaffolds and identify the best-performing candidate, considering two essential aspects: (i) the magnitude of fluid-induced mechanical stimuli, namely wall shear stress (WSS) and pressure gradients, which are established biophysical regulators of osteogenic differentiation and proliferation, and (ii) the scaffold’s structural integrity and resistance to mechanical failure (quantified via Von Mises stress distribution), thereby avoiding stress-shielding and localized collapse under compressive loading. Initially, the scaffold designs were created according to specified design requirements, including form, the presence or absence of a central connection, and the profile type, including round, square, and octagonal, using a computerized drawing and planning tool. A total of 12 models were ultimately generated, using the Design of Experiments (DOE) methodology to thoroughly examine the mechanical characteristics and conduct simulations for perfusion bioreactor testing. In the second phase, the mechanical behavior of the scaffolds was analyzed under two distinct physiological loading scenarios using a one-way decoupled FEM approach. In the first scenario, fluid-induced tractions (pressure and wall shear stress) obtained from the CFD solution were mapped onto the scaffold surfaces via one-way FSI, representing the mechanical stimuli exerted by the perfused culture medium under bioreactor perfusion conditions; this scenario was used for the primary parametric study and ranking of the twelve scaffold designs. In the second, separate scenario, a uniaxial compression simulation was performed, in which a compressive pressure load of 0.5 MPa was applied to the scaffold’s top surface with the bottom surface fully constrained, representing the real-life mechanical pressure experienced by the scaffold during daily physiological activity. In addition, a dynamic analysis was conducted to ascertain the fluid pressure distribution, velocity profile, and shear stress of the scaffold walls by Computational Fluid Dynamics (CFD) analysis. The parameters, including Von Mises stress, shear stress, pressure, and velocity, were identified and assigned to each cost function of the 12 models. Finally, the Nelder-Mead algorithm was applied to rank the twelve models based on a weighted multi-criteria score. The results demonstrate that the conducted in silico framework serves as an effective high-throughput screening tool for selecting scaffold architectures that provide a favorable biophysical microenvironment balancing mechanical resilience with perfusion-driven mechanobiological stimuli before costly in vitro experimentation. While limited to a parametric set of twelve configurations, it provides a proof-of-concept demonstration of the effectiveness of combining DOE, CFD, FEM, and Nelder-Mead ranking for selecting the best-ranked internal architecture within the investigated set; specifically the combination of body-centered struts and square cross-sectional profiles, for bone tissue engineering 3D substitutes. The macroscopic envelope geometries (cubic and rectangular prismatic) serve only as parametric spacers and are not intended to represent clinically prescriptive external shapes; anisotropy was not investigated in this proof-of-concept study and remains a subject for future work with expanded design spaces. The results demonstrate the superiority of the body-centered scaffold with square cross-sectional struts (Model M11) across nearly all mechanical and fluid dynamic criteria. While the design space in this proof-of-concept study is limited to 12 categorical configurations, the ranking framework, formally defined with a clear design vector, objective function, and constraints, identified the best-ranked architecture among the twelve candidates. This framework can be readily extended to larger design spaces with continuous variables (e.g., pore size, strut thickness, porosity) and additional performance metrics. Specifically, it outperforms the majority of models in minimizing Von Mises stress and maximizing shear rate, pressure distribution, and fluid velocity. The results could be tested further in in vitro and in vivo studies.