基于人工神经网络的MILD燃烧区域识别

Combustion Region Identification in MILD Based on Artificial Neural Network

基于人工神经网络(artificial neural network,ANN),并结合聚类分析方法,对大涡模拟仿真出的湍流燃烧场的区域进行识别,利用了聚类分析可以通过分类来提供标签的优势,充分发挥出ANN的自学习能力.以湍流MILD(moderate&intense low oxygen dilution)燃烧的HM1工况为例,搭建ANN,选择燃烧场中的物理特征,对由聚类分析提供的分类结果进行学习,来识别燃烧区域.结果表明,该方法有效提高了燃烧区域识别的准确率,并减少了大量的数据需求.为实际工业中燃烧区域的识别提供了更简单、快捷、准确的方法.

Based on artificial neural network(ANN),and combined with cluster analysis method,the turbulent combustion region simulated by large eddy simulation is identified.Advantage was taken of the fact that cluster analysis can provide labels through classification,and full use was made of the self-learning capability of ANN.The HM1 working condition of turbulent MILD combustion was used as an example.An ANN was built,and the physical features in the combustion field were selected to identify the combustion region through the self-study of the classification results provided by cluster analysis.The results show that the method effectively improves the accuracy of combustion region identification and reduces large data requirements,providing a simpler,faster and more accurate method of identifying the combustion region for practical industry.