Parameter Optimization on a Tessellation Model for Crack Pattern Simulation

Abstract

The capabilities of a parametric model for crack patterns simulation are presented. Planar tessellations are partitions of the plane into convex polygons (called cells) without overlapping. The Voronoi tessellations and Poisson line tessellations are the most prominent models; however, to model crack patterns, it is more appropriate to deal with tessellations that are generated by a cell division process. We describe the STIT tessellation as a reference model for crack patterns and introduce several modifications. Having described a variety of 40 parametric models and appropriate simulation algorithms, we delineate and specify tuning methods to optimize the adaption of the model to real crack pattern data. An example of a metalized polydimethylsiloxane demonstrates the capability of our approach. The results indicate that this approach yields a considerable improvement in modeling compared to previous studies.

Keywords

• Crack pattern
• random tessellation
• STIT tessellation
• spatial statistics
• metaheuristic tuning methods