An Automatic Isotropic Triangular Grid Generation Technique Based on an Artificial Neural Network and an Advancing Front Method

Table 1

Comparison of four automatic grid generation methods.

Performance/method	Delaunay method	Quadtree/octree method	AFM	ANN_AFM
The computation complexity (N is the total number of grids)	O(Nlog(N)) for 2D/3D grid generation.	O(Nlog(N)) for 2D/3D grid generation is faster than Delaunay and AFM	O(Nlog(N)) for 2D/3D grid generation.	O(Nlog(N)); the actual test result is 30% higher than that of the AFM.
Grid quality	The grid quality is better in 2D, and grid generation may produce thin elements in 3D.	The inner grid quality is good, but the boundary grid quality is poor.	The grid quality is better than those of both the Delaunay method and quadtree/octree method.	The grid quality is better than that of the original AFM.
Grid density controllability	The grid density can be controlled and local controllability is mediocre.	The grid density can be controlled and local controllability is good.	The grid density can be controlled and local controllability is good.	The grid density can be controlled and local controllability is good.
Program implementation complexity	Relatively easy	Comparatively complex	Medium difficulty	Medium difficulty
Anisotropic grid generation	Medium	Difficulty	Excellent	Excellent