论文部分内容阅读
针对目前RTM注射压力选取多凭经验,主观依赖性太强,数值计算又局限于简单模型,不适于形状复杂或已知参数不足的模型求解,以及遗传算法初始种群多为随机产生,容易导致算法早熟或不收敛等问题,提出基于粗糙集的改进遗传算法。用区分矩阵法对RTM注射压力决策表进行属性约简,并设定相似度阀值以提取知识库中符合相似度要求的样本来产生初始种群。实例表明,属性约简前后注射时间几乎相等,但约简后知识检索和提取速度比约简前提高了41%,说明对RTM注射压力决策表进行属性约简,不会改变知识分类,但可以明显提高计算速度。此外,改进遗传算法在17代时已经收敛,而标准遗传算法直到26代才收敛,表明改进遗传算法比标准遗传算法收敛更快更稳定,证明用改进方法优化RTM注射压力是可行和有效的。
According to the current RTM injection pressure selection more experience, the subjective dependence is too strong, the numerical calculation is limited to a simple model, not suitable for complex shapes or known parameters of the model to solve, and the initial population of genetic algorithms are mostly randomly generated, easily lead to algorithms Premature or not convergent, this paper proposes an improved genetic algorithm based on rough sets. Discriminant matrix method is used to reduce the attribute of RTM injection pressure decision table, and the similarity threshold is set to extract the samples that meet the similarity requirement in the knowledge base to generate the initial population. The example shows that the injection time before and after attribute reduction is almost the same, but the speed of knowledge retrieval and retrieval after reduction is 41% higher than that before reduction, which means that the attribute reduction of RTM injection pressure decision table will not change the knowledge classification, Significantly increase the speed of calculation. In addition, the improved genetic algorithm converged at the 17th generation, whereas the standard genetic algorithm did not converge until the 26th generation, indicating that the improved genetic algorithm converged faster and more stable than the standard genetic algorithm. It proves that it is feasible and effective to improve the RTM injection pressure with the improved method.