Multi-objective Optimization of Graft Configuration using Genetic Algorithms and Artificial Neural Network

Optimization of graft geometric configuration with regard to blood dynamics is the major target of this research. A developed multi-objective genetic algorithm is considered in order to reach optimal graft geometries for idealized arterial bypass systems of fully occluded host arteries. An artificial neural network simulating hemodynamic specific conditions is introduced in order to reduce the genetic search computational time. Input data values are constrained within pre-defined boundaries for graft geometric parameters and the correspondent target values are solutions for blood velocity and shear stress functions calculated with a finite element simulator. Optimal solutions are presented as Pareto fronts covering a range of best possible solutions. Key-Words: Shape optimization, Multi-objective optimization, Genetic algorithms, Artificial Neural Networks, Biomedical engineering