Logistic-Regression-Based Meta-Analysis of Factors Affecting Flame Stability in a Shock Tube
Logistic-Regression-Based Meta-Analysis of Factors Affecting Flame Stability in a Shock Tube
复制标题
DOI：
10.1080/00102202.2022.2142046
复制
发表时间：
2022-11
影响因子：
1.9
通讯作者：
A. J. Susa;A. Ferris;Lingzhi Zheng;Ronald K. Hanson
中科院分区：
文献类型：
--
作者：
A. J. Susa;A. Ferris;Lingzhi Zheng;Ronald K. Hanson
这项工作描述了一个荟萃分析，以更好地了解激波管火焰实验中控制火焰稳定性的参数。分析中使用的数据集是从多个先前的研究采用激波管火焰速度的方法，以及以前未报告的结果汇总。火焰首先被定性分类为平滑，扭曲，或起皱的端壁图像的基础上。然后从每组实验中选择一个子集的试验捕捉稳定过渡从稳定到不稳定的火焰。在所有的数据集，稳定和不稳定的火焰之间的过渡进行了分析，作为一个二元分类问题，使用逻辑回归。由此产生的分类模型，铸造在一个多变量的幂律表达式的临界温度（），火焰变得不稳定，被发现是高度预测的稳定过渡。考虑训练的模型系数允许识别不同实验参数的影响。模型参数提供了显着的新的见解，在实验中的稀释剂的选择，使用的测试气体具有高的热容比被发现有最有益的效果增加，而刘易斯数和分子量的影响被确定为可以忽略不计。点火位置和压力的影响是不太好量化，由于有限的可用数据，并确定为进一步研究的关键领域。
This work describes a meta-analysis performed in an effort to better understand the parameters controlling flame stability in shock-tube flame experiments. The data set used in the analysis was aggregated from multiple prior studies employing the shock-tube flame speed method, as well as previously unreported results. Flames are first qualitatively classified as being smooth, distorted, or wrinkled based on end-wall images. A subset of trials capturing the stability transition from stable to unstable flames are then selected from each set of experiments. Across all data sets, the transition between stable and unstable flames is analyzed as a binary classification problem using logistic regression. The resulting classification model, cast in terms of a multivariate power-law expression for the critical temperature () at which flames become unstable, is found to be highly predictive of the stability transition. Consideration of the trained model coefficients allows the effects of different experimental parameters onto be identified. Model parameters provide significant new insight into the selection of diluents in experiments; the use of a test gas with a high heat capacity ratio is found to have the most beneficial effect toward increasing, while the effects of Lewis number and molecular weight are determined to be negligible. Ignition-location and pressure effects are less well quantified due to limited available data and are identified as key areas of further study.