Evaluation of the effectiveness of neighbors' selection algorithms in the random cellular automata model of unconstrained grain growth

Full-field simulations used to enhance the predictive capability of classical finite element models require the preparation of high-quality initial digital material representation (DMR) data. Depending on the selected method, such a procedure can be time-consuming and, in subsequent model runs, result in a significant increase in computational time. An unconstrained grain growth algorithm based on the random cellular automata (RCA) method can be used as a potential solution to provide high-quality DMR. However, the most time-consuming part of that approach, the neighbor-search algorithm, should be optimized from an algorithmic point of view before practical application in full-field analysis. Therefore, four neighbor-search algorithms dedicated to the RCA method were developed and benchmarked within the paper for the algorithm's complexity and impact of individual parameters on computational efficiency. The research has shown that significantly reducing RCA simulation time is possible with the properly developed neighbor-search algorithm and problem-specific code optimizations.

Automated summary

Full-field simulations require high-quality initial digital material representation (DMR) data to improve the predictability of classical finite element models. An unconstrained grain growth algorithm based on the random cellular automata (RCA) method can provide high-quality DMR. This study develops and benchmarks four neighbor-search algorithms for the RCA method, evaluating their complexity and impact on computational efficiency. The results show that significantly reducing RCA simulation time is possible with optimized neighbor-search algorithms and problem-specific code optimizations.