Abstract:The evaluation methods of cone’s global angular sizes were promoted based on the roundness profile extraction strategy, and the evaluation models of the least square global angular size and the minimax global size of cone were built. Based on the evaluation models above, the Moth-flame Optimization algorithm for evaluation of cone’s global angular sizes were given. The least square and minimax global angular sizes of four cone simulated samples S-1, S-2, S-3 and S-4 were evaluated by using the developed program, and their evaluation results and optimization characteristics values obtained based on different evaluation methods, different optimization initial values and different optimization algorithms were compared. The evaluation results showed that the results obtained based on linear and non-linear least square methods are basically the same for S-1, S-2 and S-3, but the results obtained based on two kinds least square methods are different for S-4, the difference between two optimization characteristics values and the difference between two least square global angular sizes of which are 4e-8mm2 and 0.999”, respectively; For four cone samples, the evaluated results(including optimization characteristics values and minimax global angular sizes) , obtained under two kinds of least square cone parameters being taken as the optimization initial values of MFO, of each cone sample are different, which showed that the evaluation results of MFO are not globally optimal, and that the optimization initial values have some influence on the evaluation results. The minimax global angular sizes of S-4 were evaluated using Crow Search Algorithm, Artificial Ecosystem-based Optimization, Equilibrium Optimizer and Particle Swarm Optimization, and their results were compared with those of MFO. In addition to the influence of optimization initial values, the evaluation results were influenced by the search interval, the population number and the maximum number of iterations, but their influence regularities were not strong. In general, the results evaluated by MFO are better than those evaluated by the other four optimization methods