基于机器学习的氢基燃料爆炸下限预测研究

doi:10.16265/j.cnki.issn1003-3033.2026.08.1140

中国安全科学学报 ›› 2026, Vol. 36 ›› Issue (8): 179-186.doi: 10.16265/j.cnki.issn1003-3033.2026.08.1140

基于机器学习的氢基燃料爆炸下限预测研究

张智超¹(), 李艳超¹, 张锎²^,^**()

¹ 大连理工大学化学工程学院, 辽宁大连 116024
² 中国安全生产科学研究院化工园区安全研究所, 北京 100012

收稿日期:2026-04-10 修回日期:2026-06-16 出版日期:2026-08-28
通信作者:
**张锎(1990—),男,河南济源人,博士,工程师,主要从事危险化学品安全技术及工业介质爆炸安全防护研究。E-mail:californiumzhang@163.com。
作者简介:
张智超 (2001—),男,河南郑州人,硕士研究生,主要研究方向为可燃气体的爆炸下限预测、机器学习及大数据分析等。E-mail:13523568905@163.com。
李艳超副教授
基金资助:
国家自然科学基金资助(52404249); 国家自然科学基金资助(52274178); 中央引导地方科技发展资金项目(2025JH6/101000001)

Research on predicting lower explosion limits of hydrogen-based fuels using machine learning

Zhang Zhichao¹(), Li Yanchao¹, Zhang Kai²^,^**()

¹ School of Chemical Engineering, Dalian University of Technology, Dalian Liaoning 116024, China
² Chemical Industrial Park Safety Research Institute, China Academy of Safety Science and Technology, Beijing 100012, China

Received:2026-04-10 Revised:2026-06-16 Published:2026-08-28

摘要/Abstract

摘要：

为保障氢基燃料在生产、储运与使用过程中的安全,收集H₂、NH₃、CH₃OH在不同初始温度、初始压力下的爆炸下限(LEL)试验数据,构建包含498组样本的机器学习数据库;基于该数据集开展特征相关性分析,按8∶2比例划分训练集与测试集,并采用5折交叉验证策略提升神经网络预测模型训练的稳定性和可靠性;建立多元线性回归模型与深度神经网络模型,对比不同激活函数和优化算法以及不同神经元数量,确定神经网络超参数与网络结构;采用一维层流预混火焰模型与详细反应机制开展燃烧反应动力学计算,将3种方法进行系统对比与精度验证。结果表明:神经网络最优结构为5-10-12-1,采用修正线性单元(ReLU)激活函数与自适应矩估计算法优化器的预测性能最佳,模型的决定系数R²为0.983 6,均方根误差(RMSE)为0.248 6,平均绝对误差(MAE)为0.201 2,与试验值的最大MAE为0.25%;燃烧反应动力学方法精度次之,试验值的最大MAE为0.58%;多元线性回归模型精度最低、泛化能力较差,模型的R²为0.911 9,RMSE为0.296 5,MAE为0.225 1,试验值的最大MAE达1.89%。

关键词: 机器学习, 氢基燃料, 爆炸下限(LEL), 预测模型, 神经网络, 燃烧反应动力学

Abstract:

To ensure the safety of hydrogen-based fuels during production, storage, transportation and usage, the LEL test data of H₂, NH₃ and CH₃OH under different initial temperatures and pressures were collected, and a machine learning database consisting of 498 samples was constructed. Based on this dataset, feature correlation analysis was conducted, and the data was split into training and testing sets in an 8∶2 ratio. A five-fold cross-validation strategy was employed to enhance the stability and reliability of neural network model training. Multivariate linear regression models and deep neural network models were established separately. By comparing different activation functions, optimization algorithms, and numbers of neurons, the hyperparameters and network architecture of the neural network were determined. Concurrently, combustion reaction kinetics calculations were performed using a one-dimensional laminar premixed flame model and a detailed reaction mechanism, and the three methods were systematically compared and validated for accuracy. The results indicate that the optimal structure of the neural network is 5-10-12-1. The prediction performance is the best when using Rectified Linear Unit(ReLU) activation function and the adaptive moment estimation algorithm as the optimizer. The coefficient of determination R² of the model is 0.983 6, the root mean square error (RMSE) is 0.248 6, the mean absolute error (MAE) is 0.201 2, and the maximum MAE compared with the experimental values is 0.25%. The combustion reaction kinetics method has the second-best accuracy, with the maximum MAE of the experimental values being 0.58%. The multiple linear regression model has the lowest accuracy and poor generalization ability. R² of the model is 0.911 9, the RMSE is 0.296 5, and the MAE is 0.225 1, and the maximum MAE of the experimental values reaches 1.89%.

Key words: machine learning, hydrogen-based fuels, lower explosive limit(LEL), prediction model, neural network, combustion reaction kinetics

中图分类号:

X932爆炸安全与防火、防爆

张智超, 李艳超, 张锎. 基于机器学习的氢基燃料爆炸下限预测研究[J]. 中国安全科学学报, 2026, 36(8): 179-186.

Zhang Zhichao, Li Yanchao, Zhang Kai. Research on predicting lower explosion limits of hydrogen-based fuels using machine learning[J]. China Safety Science Journal, 2026, 36(8): 179-186.

图/表 9

表1

图1

图2

图3

图4

图5

图6

图7

图8

参考文献 19

[1]	Wang Xianming, Law C K. An analysis of the explosion limits of hydrogen-oxygen mixtures[J]. The Journal of Chemical Physics, 2013, 138(13):DOI:10.1063/1.4798459.
[2]	Jaimes D J. Determination of lower flammability limits of mixtures of air and gaseous renewable fuels at elevated temperatures and pressures[D]. Irvine: University of California, 2017.
[3]	Vidal M, Wong W, Rogers W J, et al. Evaluation of lower flammability limits of fuel-air-diluent mixtures using calculated adiabatic flame temperatures[J]. Journal of Hazardous Materials, 2006, 130(1/2):21-27. doi: 10.1016/j.jhazmat.2005.07.080
[4]	李国梁, 蒋军成, 潘勇. 基于绝热火焰温度混合气体爆炸下限的预测[J]. 中国安全科学学报, 2011, 21(7): 57-61.
	Li Guoliang, Jiang Juncheng, Pan Yong. Prediction of lower explosive limit of mixed gas based on adiabatic flame temperature[J]. China Safety Science Journal, 2011, 21(7): 57-61.
[5]	Zheng Ligang, Fan Zhiqiang, Yu Minggao, et al. Predictive model on explosion limits of explosive gas mixture containing H₂, CH₄ and CO based on neural network[J]. Progress in Safety Science and Technology, 2006, 6: 1227-1231.
[6]	时静洁, 赵薇, 陈小林, 等. 基于BP人工神经网络的有机化合物爆炸下限预测[J]. 消防科学与技术, 2023, 42(5): 609-614.
	Shi Jingjie, Zhao Wei, Chen Xiaolin, et al. Prediction of lower explosive limit of organic compounds based on BP artificial neural network[J]. Fire Science and Technology, 2023, 42(5): 609-614.
[7]	Cheikhravat H, Chaumeix N, Bentaib A, et al. Flammability limits of hydrogen-air mixtures: nuclear technology[J]. Nuclear Technology, 2017, 178(1): 5-16. doi: 10.13182/NT12-A13543
[8]	Zhang Kai, Luo Tianpei, Li Yanchao, et al. Effect of ignition, initial pressure and temperature on the lower flammability limit of hydrogen/air mixture[J]. International Journal of Hydrogen Energy, 2022, 47(33):15107-15 119.
[9]	Zhang Kai, Shang Sheng, Li Xiaolin, et al. Lower flammability limits of NH₃/H₂ mixtures under different initial temperatures and initial pressures[J]. Fuel, 2023, 331: DOI:10.1016/j.fuel.2022.125982.
[10]	Zhang Kai, Li Xiaolin, Jiang Yuting, et al. Lower flammability limits of H₂/NH₃/CHv3OH mixtures under elevated pressures, temperatures and different blending ratios[J]. Fuel, 2023, 337:DOI:10.1016/j.fuel.2022.127159.
[11]	Liu Xueling, Zhang Qi. Influence of initial pressure and temperature on flammability limits of hydrogen-air[J]. International Journal of Hydrogen Energy, 2014, 39(12): 6774-6782. doi: 10.1016/j.ijhydene.2014.02.001
[12]	Kondo S, Takizawa K, Takahashi A, et al. On the temperature dependence of flammability limits of gases[J]. Journal of Hazardous Materials, 2011, 187(1): 585-590. doi: 10.1016/j.jhazmat.2011.01.037
[13]	Luo Zhenmin, Wang Xuqing, Zhang Siqi, et al. Experimental study on the flammability limit parameters of premixed methanol-gasoline vapor-air mixtures[J]. Journal of Loss Prevention in the Process Industries, 2022, 80: DOI:10.1016/j.jlp.2022.104856.
[14]	Ciccarelli G, Jackson D, Verreault J. Flammability limits of NH₃-H₂-N₂-air mixtures at elevated initial temperatures[J]. Combustion and Flame, 2006, 144(1): 53-63. doi: 10.1016/j.combustflame.2005.06.010
[15]	Su Bin, Tan Yunsong, Zhang Lidong, et al. Prediction models for flammability limits of syngas/air mixtures based on machine learning approach[J]. International Journal of Hydrogen Energy, 2025, 98: 1356-1365. doi: 10.1016/j.ijhydene.2024.12.040
[16]	张利冬, 宋泽阳, 罗振敏, 等. 基于机器学习的煤自然发火期预测[J]. 中国安全科学学报, 2022, 32(12): 118-124. doi: 10.16265/j.cnki.issn1003-3033.2022.12.0134
	Zhang Lidong, Song Zeyang, Luo Zhenmin, et al. Prediction of spontaneous combustion period of coal based on machine learning[J]. China Safety Science Journal, 2022, 32(12): 118-124. doi: 10.16265/j.cnki.issn1003-3033.2022.12.0134
[17]	Udaybhanu G, Jameel A G A, Roberts W L, et al. Development of laminar burning velocity prediction model and correlation of iso-octane air mixtures using artificial neural network[J]. Energy, 2024, 307: DOI:10.1016/j.energy.2024.132639.
[18]	GRI-MECH home page[EB/OL]. (2002-10-30). http://combustion.berkeley.edu/gri-mech/.
[19]	Okafor E C, Naito Y, Colson S, et al. Experimental and numerical study of the laminar burning velocity of CH₄-NH₃-air premixed flames[J]. Combustion and Flame, 2018, 187: 185-198. doi: 10.1016/j.combustflame.2017.09.002

文献	H₂	NH₃	CH₃OH	温度/℃	压力/kPa
[7]	√			25~15	100~250
[8]	√			20~100	50~250
[9]	√	√		20~100	50~250
[10]	√	√	√	20~100	50~250
[11]	√			21~90	100~400
[12]		√		5~100	100
[13]			√	40~120	100
[14]	√	√		20~300	100

基于机器学习的氢基燃料爆炸下限预测研究

Research on predicting lower explosion limits of hydrogen-based fuels using machine learning

RichHTML

PDF

可视化

摘要/Abstract

引用本文

使用本文

图/表 9

参考文献 19

相关文章 15

编辑推荐

Metrics

本文评价

[1]	张庆峰, 曹宇飞, 周悦, 付川云. 数据不均衡下教练机致命事故识别与分析[J]. 中国安全科学学报, 2026, 36(8): 124-132.
[2]	喻豪, 王超, 贺子旺, 刘宇, 金子浚, 亓帅. 基于机器学习的PBM细观参数标定及卸荷岩爆模拟研究[J]. 中国安全科学学报, 2026, 36(8): 260-268.
[3]	陈晓红, 蔡诚成, 刘晓亮, 安庆贤. 基于多尺度残差卷积神经网络的矿山微震识别[J]. 中国安全科学学报, 2026, 36(8): 55-64.
[4]	纪文涛, 王雅歌, 肖海利, 孟令宣, 王燕. 气粉两相体系爆炸下限变化特征及量化模型研究进展[J]. 中国安全科学学报, 2026, 36(7): 1-11.
[5]	陈昱光, 海凌滔, 杨彬, 郭延永, 肖海承. 用于车辆运行风险预测的ISMA-BP神经网络算法[J]. 中国安全科学学报, 2026, 36(7): 199-206.
[6]	张春晓, 闫春雨, 邢志伟, 冯博文, 王芸, 丛子棋. 考虑服役环境的RBF-ANN飞机襟翼结构损伤预测[J]. 中国安全科学学报, 2026, 36(6): 24-32.
[7]	张景林, 陈圣群. 基于少样本数据驱动的危化品爆炸事故风险定级模型[J]. 中国安全科学学报, 2026, 36(6): 40-50.
[8]	张建国, 王文唱, 任连伟, 邹友峰, 顿志林. 基于SAA-GRNN优化模型的厚松散层条件下概率积分法参数求解[J]. 中国安全科学学报, 2026, 36(5): 18-26.
[9]	余修武, 肖林, 叶莱. 基于铀尾矿库核素监测的SAPSO-BP数据融合算法[J]. 中国安全科学学报, 2026, 36(5): 48-55.
[10]	王睿, 张讯, 邓祥辉, 王平安, 王旭, 张巍. 爆炸作用下地铁盾构隧道韧性评估与预测[J]. 中国安全科学学报, 2026, 36(4): 123-131.
[11]	王丽, 王喆, 关文玲, 张嘉琪, 刘超林, 孟玉莹. 面向应急救援的BP神经网络人员定位-追踪-静止报警系统[J]. 中国安全科学学报, 2026, 36(4): 204-210.
[12]	吴凡, 赖密密, 李明阳. 职业韧性对民航飞行员安全绩效的影响机制[J]. 中国安全科学学报, 2026, 36(3): 1-8.
[13]	张杰, 杨科, 范超尘. 基于MTAM-LSTM的采煤工作面支架载荷预测方法[J]. 中国安全科学学报, 2026, 36(3): 144-152.
[14]	谢尊贤, 马浩浩, 江松, 武潇云. 基于GA-BP神经网络的露天矿山排土场边坡失稳预测[J]. 中国安全科学学报, 2026, 36(3): 81-88.
[15]	田水承, 李红妍, 石炎彬, 田方圆, 王亚娟, 段梦菲. 警觉度影响下矿工风险感知能力识别模型[J]. 中国安全科学学报, 2026, 36(2): 1-8.