基于机器学习模型的制冷剂燃爆特性预测
Prediction of Refrigerant Ignition and Detonation Characteristics Based on Machine Learning Model
费腾 1杨昭 1陈裕博 1张勇 1李杰1
作者信息
- 1. 天津大学,中低温热能高效利用教育部重点实验室 天津 300072
- 折叠
摘要
本文以常见可燃工质为对象,基于Gaussian 16W的M06-2X/6-311+G(d,p)优化计算,得出工质分子结构在微观模型下的分子描述符数据.同时采用多元线性回归(MLR)、随机森林(RF)、人工神经网络(ANN)三种不同的机器学习方法,将微观数据与宏观实验数据关联起来,从而预测这些工质的最小点火能,总体R2分别达到了 0.853、0.782和0.906,表明预测有着良好的准确度和鲁棒性,该预测结果可以为新工质的实用性和安全性提供理论依据.
Abstract
In this paper,common flammable working medium is taken as the object.Based on M06-2X/6-311+G(d,p)optimization calculation of Gaussian 16 W,molecular descriptor data of working medium molecular structure under microscopic model are obtained.At the same time,three different machine learning methods,namely Multiple Linear Regression(MLR),Random Forest(RF)and Artificial Neural Network(ANN),were used to correlate the micro data with the macro experimental data,so as to predict the minimum ignition energy of these working media.The overall R2 reached 0.853,0.782 and 0.906,respectively.The results show that the prediction has good accuracy and robustness.The prediction results can provide theoretical basis for the practicability and safety of the new working medium.
关键词
燃爆特性/机器学习/最小点火能/可燃工质Key words
ignition and detonation characteristics/machine learning/minimum ignition energy/combustible working medium引用本文复制引用
出版年
2024
NETL
NSTL
万方数据