混合变分推理深度学习驱动的可燃气体扩散数值模型的不确定性量化

IF 4.2 3区 工程技术 Q2 ENGINEERING, CHEMICAL
Weikang Xie , Qing Wang , Jiyuan Li , Zonghao Xie , Jihao Shi , Xinyan Huang , Asif Usmani
{"title":"混合变分推理深度学习驱动的可燃气体扩散数值模型的不确定性量化","authors":"Weikang Xie ,&nbsp;Qing Wang ,&nbsp;Jiyuan Li ,&nbsp;Zonghao Xie ,&nbsp;Jihao Shi ,&nbsp;Xinyan Huang ,&nbsp;Asif Usmani","doi":"10.1016/j.jlp.2025.105758","DOIUrl":null,"url":null,"abstract":"<div><div>Accurate modeling of flammable gas dispersion is essential for fire and explosion risk assessment. However, CFD models that rely on fixed hyperparameters preclude uncertainty quantification, leading to overconfidence prediction. This work proposed a hybrid deep learning framework with variational Bayesian inference to inversely solve distributions of numerical model parameters. The gas dispersion database under different Froude numbers <em>Fr</em> is developed, which contains repetitive experimental data and corresponding numerical simulation values. CNN-AM architecture is developed to capture nonlinear relationship between model parameters and concentration outputs. Using experimental data, ADVI is employed to derive posterior distributions of the optimal model parameters. The results indicate that the parameter-optimized model obviously improves prediction accuracy for 80 % scenarios, with overall error below 5 %. Furthermore, spatial distribution characteristics of plumes are characterized probabilistically. Near leakage nozzles, local concentration fluctuations peak when gravity and initial momentum jointly dominate plume dynamics at <em>Fr</em> = 74.38. In terms of plume morphology, variability in horizontal extent increases monotonically with <em>Fr</em>, while uncertainty in vertical drop attains a maximum at 0.060 when <em>Fr</em> = 55.79. These findings demonstrate the robustness of the proposed method for uncertainty quantification in gas distribution modelling, thereby enhancing risk evaluation in industries.</div></div>","PeriodicalId":16291,"journal":{"name":"Journal of Loss Prevention in The Process Industries","volume":"98 ","pages":"Article 105758"},"PeriodicalIF":4.2000,"publicationDate":"2025-08-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":"{\"title\":\"Uncertainty quantification of flammable gas dispersion numerical models driven by hybrid variational inference deep learning\",\"authors\":\"Weikang Xie ,&nbsp;Qing Wang ,&nbsp;Jiyuan Li ,&nbsp;Zonghao Xie ,&nbsp;Jihao Shi ,&nbsp;Xinyan Huang ,&nbsp;Asif Usmani\",\"doi\":\"10.1016/j.jlp.2025.105758\",\"DOIUrl\":null,\"url\":null,\"abstract\":\"<div><div>Accurate modeling of flammable gas dispersion is essential for fire and explosion risk assessment. However, CFD models that rely on fixed hyperparameters preclude uncertainty quantification, leading to overconfidence prediction. This work proposed a hybrid deep learning framework with variational Bayesian inference to inversely solve distributions of numerical model parameters. The gas dispersion database under different Froude numbers <em>Fr</em> is developed, which contains repetitive experimental data and corresponding numerical simulation values. CNN-AM architecture is developed to capture nonlinear relationship between model parameters and concentration outputs. Using experimental data, ADVI is employed to derive posterior distributions of the optimal model parameters. The results indicate that the parameter-optimized model obviously improves prediction accuracy for 80 % scenarios, with overall error below 5 %. Furthermore, spatial distribution characteristics of plumes are characterized probabilistically. Near leakage nozzles, local concentration fluctuations peak when gravity and initial momentum jointly dominate plume dynamics at <em>Fr</em> = 74.38. In terms of plume morphology, variability in horizontal extent increases monotonically with <em>Fr</em>, while uncertainty in vertical drop attains a maximum at 0.060 when <em>Fr</em> = 55.79. These findings demonstrate the robustness of the proposed method for uncertainty quantification in gas distribution modelling, thereby enhancing risk evaluation in industries.</div></div>\",\"PeriodicalId\":16291,\"journal\":{\"name\":\"Journal of Loss Prevention in The Process Industries\",\"volume\":\"98 \",\"pages\":\"Article 105758\"},\"PeriodicalIF\":4.2000,\"publicationDate\":\"2025-08-14\",\"publicationTypes\":\"Journal Article\",\"fieldsOfStudy\":null,\"isOpenAccess\":false,\"openAccessPdf\":\"\",\"citationCount\":\"0\",\"resultStr\":null,\"platform\":\"Semanticscholar\",\"paperid\":null,\"PeriodicalName\":\"Journal of Loss Prevention in The Process Industries\",\"FirstCategoryId\":\"5\",\"ListUrlMain\":\"https://www.sciencedirect.com/science/article/pii/S0950423025002165\",\"RegionNum\":3,\"RegionCategory\":\"工程技术\",\"ArticlePicture\":[],\"TitleCN\":null,\"AbstractTextCN\":null,\"PMCID\":null,\"EPubDate\":\"\",\"PubModel\":\"\",\"JCR\":\"Q2\",\"JCRName\":\"ENGINEERING, CHEMICAL\",\"Score\":null,\"Total\":0}","platform":"Semanticscholar","paperid":null,"PeriodicalName":"Journal of Loss Prevention in The Process Industries","FirstCategoryId":"5","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S0950423025002165","RegionNum":3,"RegionCategory":"工程技术","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q2","JCRName":"ENGINEERING, CHEMICAL","Score":null,"Total":0}
引用次数: 0

摘要

准确的可燃气体扩散模型对火灾和爆炸风险评估至关重要。然而,依靠固定超参数的CFD模型排除了不确定性量化,导致过度自信的预测。本文提出了一种带有变分贝叶斯推理的混合深度学习框架,用于反求解数值模型参数的分布。建立了不同弗劳德数下的气体分散数据库,该数据库包含重复实验数据和相应的数值模拟值。采用CNN-AM结构捕捉模型参数与浓度输出之间的非线性关系。利用实验数据,采用ADVI方法推导出最优模型参数的后验分布。结果表明,参数优化模型在80%情景下的预测精度明显提高,总体误差在5%以下。此外,对羽流的空间分布特征进行了概率表征。在泄漏喷嘴附近,当重力和初始动量共同主导羽流动力学时,局部浓度波动在Fr = 74.38处达到峰值。在羽流形态方面,水平程度的不确定性随Fr单调增加,而垂直下降的不确定性在Fr = 55.79时达到最大,为0.060。这些发现证明了所提出的方法在气体分布建模中的不确定性量化的鲁棒性,从而提高了工业风险评估。
本文章由计算机程序翻译,如有差异,请以英文原文为准。
Uncertainty quantification of flammable gas dispersion numerical models driven by hybrid variational inference deep learning
Accurate modeling of flammable gas dispersion is essential for fire and explosion risk assessment. However, CFD models that rely on fixed hyperparameters preclude uncertainty quantification, leading to overconfidence prediction. This work proposed a hybrid deep learning framework with variational Bayesian inference to inversely solve distributions of numerical model parameters. The gas dispersion database under different Froude numbers Fr is developed, which contains repetitive experimental data and corresponding numerical simulation values. CNN-AM architecture is developed to capture nonlinear relationship between model parameters and concentration outputs. Using experimental data, ADVI is employed to derive posterior distributions of the optimal model parameters. The results indicate that the parameter-optimized model obviously improves prediction accuracy for 80 % scenarios, with overall error below 5 %. Furthermore, spatial distribution characteristics of plumes are characterized probabilistically. Near leakage nozzles, local concentration fluctuations peak when gravity and initial momentum jointly dominate plume dynamics at Fr = 74.38. In terms of plume morphology, variability in horizontal extent increases monotonically with Fr, while uncertainty in vertical drop attains a maximum at 0.060 when Fr = 55.79. These findings demonstrate the robustness of the proposed method for uncertainty quantification in gas distribution modelling, thereby enhancing risk evaluation in industries.
求助全文
通过发布文献求助,成功后即可免费获取论文全文。 去求助
来源期刊
CiteScore
7.20
自引率
14.30%
发文量
226
审稿时长
52 days
期刊介绍: The broad scope of the journal is process safety. Process safety is defined as the prevention and mitigation of process-related injuries and damage arising from process incidents involving fire, explosion and toxic release. Such undesired events occur in the process industries during the use, storage, manufacture, handling, and transportation of highly hazardous chemicals.
×
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