Heat flux prediction of electrical cabinet fires using a physics-based model combined with machine learning methods

IF 6.9 2区 工程技术 Q2 ENERGY & FUELS
Qiuju Ma , Zhennan Chen , Jianhua Chen , Yubo Sun , Nan Chen , Mengzhen Du
{"title":"Heat flux prediction of electrical cabinet fires using a physics-based model combined with machine learning methods","authors":"Qiuju Ma ,&nbsp;Zhennan Chen ,&nbsp;Jianhua Chen ,&nbsp;Yubo Sun ,&nbsp;Nan Chen ,&nbsp;Mengzhen Du","doi":"10.1016/j.applthermaleng.2025.128576","DOIUrl":null,"url":null,"abstract":"<div><div>Accurate heat flux prediction in cabinet fires is essential for evaluating thermal risks. However, traditional methods become constrained due to flame obstruction, unavailable key parameters on-site, and limited future prediction capability. This study proposes a hybrid framework combining extreme gradient boosting (XGBoost), a physics-based model, and bidirectional long short-term memory (BiLSTM) for heat flux inference and forecasting. XGBoost is employed to infer target temperatures from cabinet wall thermocouples, which are then converted into heat flux through a physics-based model, establishing a real-time inference pathway from accessible measurements to heat flux. Then, BiLSTM is introduced to forecast heat flux 60 s ahead, thereby capturing the future evolution of the target variable. Results demonstrate that the predicted heat flux achieves errors below 0.17 kW/m<sup>2</sup> in training, validation, and test sets. When using predicted heat flux as input, BiLSTM maintains errors below 5.9 %. Compared with other models, our methods achieve lower errors. In addition, a feature selection strategy based on correlation analysis and hierarchical clustering is developed to identify the optimal thermocouple combination when constructing the dataset. The proposed method provides a practical and efficient insight for predicting heat flux in cabinet fires, contributing to enhanced fire risk management in nuclear industry.</div></div>","PeriodicalId":8201,"journal":{"name":"Applied Thermal Engineering","volume":"280 ","pages":"Article 128576"},"PeriodicalIF":6.9000,"publicationDate":"2025-10-01","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Applied Thermal Engineering","FirstCategoryId":"5","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S1359431125031680","RegionNum":2,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q2","JCRName":"ENERGY & FUELS","Score":null,"Total":0}
引用次数: 0

Abstract

Accurate heat flux prediction in cabinet fires is essential for evaluating thermal risks. However, traditional methods become constrained due to flame obstruction, unavailable key parameters on-site, and limited future prediction capability. This study proposes a hybrid framework combining extreme gradient boosting (XGBoost), a physics-based model, and bidirectional long short-term memory (BiLSTM) for heat flux inference and forecasting. XGBoost is employed to infer target temperatures from cabinet wall thermocouples, which are then converted into heat flux through a physics-based model, establishing a real-time inference pathway from accessible measurements to heat flux. Then, BiLSTM is introduced to forecast heat flux 60 s ahead, thereby capturing the future evolution of the target variable. Results demonstrate that the predicted heat flux achieves errors below 0.17 kW/m2 in training, validation, and test sets. When using predicted heat flux as input, BiLSTM maintains errors below 5.9 %. Compared with other models, our methods achieve lower errors. In addition, a feature selection strategy based on correlation analysis and hierarchical clustering is developed to identify the optimal thermocouple combination when constructing the dataset. The proposed method provides a practical and efficient insight for predicting heat flux in cabinet fires, contributing to enhanced fire risk management in nuclear industry.
结合机器学习方法的基于物理模型的电柜火灾热流预测
准确的柜体火灾热通量预测是评估柜体火灾热风险的关键。然而,传统的方法由于火焰障碍、现场关键参数无法获得以及未来预测能力有限而受到限制。本研究提出了一个结合极端梯度增强(XGBoost)、基于物理的模型和双向长短期记忆(BiLSTM)的混合框架,用于热通量推断和预测。利用XGBoost从机柜壁热电偶中推断出目标温度,然后通过基于物理的模型将其转换为热流密度,建立了从可访问测量到热流密度的实时推断路径。然后,引入BiLSTM提前60 s预测热通量,从而捕捉目标变量的未来演变。结果表明,在训练集、验证集和测试集上,预测的热流密度误差均在0.17 kW/m2以下。当使用预测热通量作为输入时,BiLSTM将误差保持在5.9%以下。与其他模型相比,我们的方法误差更小。此外,提出了一种基于相关分析和层次聚类的特征选择策略,在构建数据集时识别出最优的热电偶组合。该方法为柜体火灾热通量预测提供了实用有效的方法,有助于加强核工业火灾风险管理。
本文章由计算机程序翻译,如有差异,请以英文原文为准。
求助全文
约1分钟内获得全文 求助全文
来源期刊
Applied Thermal Engineering
Applied Thermal Engineering 工程技术-工程:机械
CiteScore
11.30
自引率
15.60%
发文量
1474
审稿时长
57 days
期刊介绍: Applied Thermal Engineering disseminates novel research related to the design, development and demonstration of components, devices, equipment, technologies and systems involving thermal processes for the production, storage, utilization and conservation of energy, with a focus on engineering application. The journal publishes high-quality and high-impact Original Research Articles, Review Articles, Short Communications and Letters to the Editor on cutting-edge innovations in research, and recent advances or issues of interest to the thermal engineering community.
×
引用
GB/T 7714-2015
复制
MLA
复制
APA
复制
导出至
BibTeX EndNote RefMan NoteFirst NoteExpress
×
提示
您的信息不完整,为了账户安全,请先补充。
现在去补充
×
提示
您因"违规操作"
具体请查看互助需知
我知道了
×
提示
确定
请完成安全验证×
copy
已复制链接
快去分享给好友吧!
我知道了
右上角分享
点击右上角分享
0
联系我们:info@booksci.cn Book学术提供免费学术资源搜索服务,方便国内外学者检索中英文文献。致力于提供最便捷和优质的服务体验。 Copyright © 2023 布克学术 All rights reserved.
京ICP备2023020795号-1
ghs 京公网安备 11010802042870号
Book学术文献互助
Book学术文献互助群
群 号:604180095
Book学术官方微信