The optimum scaffold for tissue engineering must guarantee the mechanical integrity in the damaged zone and ensure an appropriate stiffness to regulate the cellular function. For this to happen， scaffolds must be designed to match the stiffness of the native tissue. Moreover， the degradation rate in the case of bioresorbable materials must be also considered to fit the tissue regeneration rate. This paper presents a methodology based on Design of Experiments， Finite Element Analysis， metamodels and Genetic Algorithms to optimize the assignation of material in different sections of the scaffold to obtain the desired stiffness over time and comply with the constraints needed. The method applies an initial... More
The optimum scaffold for tissue engineering must guarantee the mechanical integrity in the damaged zone and ensure an appropriate stiffness to regulate the cellular function. For this to happen， scaffolds must be designed to match the stiffness of the native tissue. Moreover， the degradation rate in the case of bioresorbable materials must be also considered to fit the tissue regeneration rate. This paper presents a methodology based on Design of Experiments， Finite Element Analysis， metamodels and Genetic Algorithms to optimize the assignation of material in different sections of the scaffold to obtain the desired stiffness over time and comply with the constraints needed. The method applies an initial sampling focused on a modified Latin Hypercube strategy to obtain data from the simulations. These data are used in the next stages to generate the metamodels by using Kriging. The predictions of the metamodels are used by the genetic algorithms to find the best estimated solutions. Different runs of the genetic algorithm drive the sampling， improving the accuracy of the surrogate models over the optimization process. Once the accuracy of the metamodels estimates is sufficient， a final genetic algorithm is applied to obtain the optimum design. This approach guarantees a low sampling effort and convergence to carry out the optimization process. The method allows the combination of discrete and continuous design variables in the optimization problem， and it can be applied both in solid and in hierarchical-based geometries.
