Floating Point Genetic Algorithm Approach for Optimizing Unstructured Meshes

Abstract

This paper presents a method for optimizing unstructured triangular meshes using a floating-point genetic algorithm. A mesh generation algorithm based on a modified advancing front method and sets of heuristic rules are used to generate the initial non-smooth triangular meshes for complex shapes. The developed mesh is then smoothed using a floating-point genetic algorithm that is more flexible than the usual binary genetic algorithms, and can handle non-smooth regions containing several local extrema. Three approaches are used in selecting the fitness function to be optimized in the genetic algorithm, namely, the triangle aspect ratio, the maximum angle at each node of the triangular mesh, and a weighted linear combination of both functions. The genetic algorithm has been tested and validated for a number of test cases covering a wide range of complex geometry applications. The results have shown a high degree of improvement in the quality of the smoothed meshes and an ability to handle non-convex regions.