【摘要】 结合人工智能方法和大涡模拟(LES)对湍流燃烧场中的反应区域进行识别,可以发挥LES能够提供三维空间中不同物理场数据的优势,也有效解决火焰面识别中传统方法不够客观和依赖于经验的问题。以湍流条件下MILD燃烧的HM1工况为例,引入无监督学习中的聚类算法,对多个变量的模拟数据进行聚类分析。结果表明,在考虑了多个物理变量之后,三种不同的聚类算法对反应区域识别有着很好的一致性,说明多维聚类分析与LES结合,能够将复杂的湍流燃烧场进行区域类别的划分。这一新的方法对于进一步分析湍流燃烧场中各个区域(尤其是反应区域)的特性,提供了一个支撑。更多还原

【Abstract】 The integrated analysis of artificial intelligence method and large-eddy simulation (LES)data for flame recognition in turbulent combustion flames can take advantage of LES to provide a large number of different physical fields in three-dimensional space,and also effectively solve the traditional method of flame recognition is not objective and dependent in a-priori knowledge.This study adopts a well-known turbulent MILD combustion case,the HM1 case,and the clustering algorithm in unsupervised learning to perform cluster analysis on the LES data of multiple variables.Results show that after considering multiple physical variables,three different clustering algorithms have a good consistency in the recognition of flame surface,indicating that multidimensional cluster analysis combined with LES can uniformly classify complex turbulent combustion fields.This new method provides a support for further analysis of the characteristics of various regions (especially the combustion region)in turbulent combustion field.更多还原