Fire source localization for long short-term memory networks based on Bayesian optimization
-
摘要:
民航飞机货舱火灾多发于高空低温低压的环境，对飞机安全飞行造成了巨大的威胁。为快速定位货舱火灾源点和采取针对性区域灭火措施，提出一种基于贝叶斯优化（BO）的长短期记忆（LSTM）神经网络火源定位模型（BO-LSTM）。该模型使用LSTM神经网络充分挖掘多种火灾特征时序数据（烟雾、温度、CO浓度）与火灾源点的时空关联特性，同时采用贝叶斯算法搜寻LSTM神经网络的最优超参数组合以提高模型的鲁棒性和准确性。通过仿真研究验证BO-LSTM模型，使用Pyrosim火灾模拟软件以1∶1比例建立了8个常用民航飞机货舱模型，并在每个模型中随机选取10个火源点进行低温低压环境的火灾仿真。实验结果表明：所建模型预测火源中心点距离实际火源中心点的直线距离误差皆小于0.1 m，预测火源二维坐标皆处于真实火源的范围内。贝叶斯优化过的LSTM神经网络极大提高了传统LSTM神经网络的性能，适用于低温低压状态下的飞机货舱火源定位。
Abstract:Civil aviation aircraft cargo compartment fires often occur in high-altitude, low-temperature and low-pressure environments, posing a significant threat to aircraft safety. A Bayesian optimization (BO) based long short-term memory (LSTM) neural network model (BO-LSTM) is proposed to quickly locate fire source and take specific means for fire suppression. In BO-LSTM model, time series data are fully analyzed to investigate the spatiotemporal correlation between fire characteristics (smoke, temperature, CO concentration) and fire source by using LSTM network. Meanwhile, the Bayesian algorithm is employed to search for the optimal LSTM network hyperparameter combination to improve model robustness and accuracy. In this paper, simulation study was used to validate the proposed BO-LSTM method: 8 widely used aircraft cargo compartment models were built by Pyrosim® fire simulation software at a scale of 1∶1, and 10 fire source positions were randomly selected in each cargo compartment model to simulate fire data at low-temperature and low-pressure conditions. The experimental results demonstrate that the distance error between predicted fire location and true fire source was less than 0.1 meters, and the predicted two-dimensional fire source localized within the true fire source range. The performance of LSTM neural network has been improved dramatically by the Bayesian optimization method and make it a suitable tool for aircraft cargo compartment fire source localization at the environment of low-temperature and low-pressure.
-
图 11 不同机型前后舱的平均绝对误差分析
Figure 11. Mean absolute error analysis of front and lower cabins of different aircraft models
表 1 LSTM网络域空间的超参数
Table 1. Hyperparametric domain space of LSTM networks
超参数 域空间 神经元数量1(units1) （50,100） 神经元数量2(units2) (1,50) 弃权系数1(dropout1) (0,0.5) 弃权系数2(dropout2) (0,0.5) 批处理大小(batch_size) (0,100) 训练步数(Epochs) (300,600) 优化器(optimizer) ['Adam','RMSProp','AdaGrad'] 激活函数(activation) ['sigmoid','relu'] 
下载: 导出CSV
-
[1] 刘鹏. 新时期民航飞机火灾的特点和扑救战术原则探讨[J]. 湖南安全与防灾, 2019(1): 48-51. doi: 10.3969/j.issn.1007-9947.2019.01.019LIU P. Discussion on the characteristics of civil aviation aircraft fire and the principle of fighting tactics in the new period[J]. Hunan Safety and Disaster Prevention, 2019(1): 48-51(in Chinese). doi: 10.3969/j.issn.1007-9947.2019.01.019 [2] ANDREW A M, ZAKARIA A, SAAD S M, et al. Multi-stage feature selection based intelligent classifier for classification of incipient stage fire in building[J]. Sensors, 2016, 16(1): 31. doi: 10.3390/s16010031 [3] 王洋, 李向国, 梅志千, 等. 基于红外阵列传感器的火源定位方法[J]. 系统仿真技术, 2022, 18(2): 73-76.WANG Y, LI X G, MEI Z Q, et al. Fire source positioning method based on infrared array sensor[J]. System Simulation Technology, 2022, 18(2): 73-76(in Chinese). [4] 刘帅, 吴梦军, 左远正. 基于图像处理的大跨隧道火灾定位技术试验研究[J]. 公路交通技术, 2018, 34(S1): 41-44.LIU S, WU M J, ZUO Y Z. Experimental study on fire locating technology based on image processing in large-span tunnel[J]. Technology of Highway and Transport, 2018, 34(S1): 41-44(in Chinese). [5] 胡淋翔, 安子樱, 李伟, 等. 基于神经网络的单室火源位置识别模型[J]. 建筑热能通风空调, 2022, 41(8): 53-57. doi: 10.3969/j.issn.1003-0344.2022.08.012HU L X, AN Z Y, LI W, et al. Compartment fire location identification model based on neural network[J]. Building Energy & Environment, 2022, 41(8): 53-57(in Chinese). doi: 10.3969/j.issn.1003-0344.2022.08.012 [6] WAHLQVIST J, VAN HEES P. Implementation and validation of an environmental feedback pool fire model based on oxygen depletion and radiative feedback in FDS[J]. Fire Safety Journal, 2016, 85: 35-49. doi: 10.1016/j.firesaf.2016.08.003 [7] 杨建忠, 邵资焱, 陈希远. 通风对飞机货舱烟雾探测影响研究[J]. 中国安全科学学报, 2019, 29(2): 69-75.YANG J Z, SHAO Z Y, CHEN X Y. Research on influence of ventilation on detection of smoke in an aircraft cargo compartment[J]. China Safety Science Journal, 2019, 29(2): 69-75(in Chinese). [8] BROHEZ S, CARAVITA I. Fire induced pressure in airthigh houses: experiments and FDS validation[J]. Fire Safety Journal, 2020, 114: 103008. doi: 10.1016/j.firesaf.2020.103008 [9] 涂芝润. 基于LSTM的VOD精炼炉终点预报模型研究[D]. 西安: 西安理工大学, 2021.TU Z R. Research on endpoint prediction model of VOD refining furnace based on LSTM[D]. Xi’an: Xi’an University of Technology, 2021(in Chinese). [10] FRAZIER P I. A tutorial on Bayesian optimization[EB/OL]. (2018-07-08)[2023-07-21]. http://arxiv.org/abs/1807.02811. [11] SHIELDS B J, STEVENS J, LI J, et al. Bayesian reaction optimization as a tool for chemical synthesis[J]. Nature, 2021, 590(7844): 89-96. doi: 10.1038/s41586-021-03213-y [12] VARGAS HERNÁNDEZ R A. Bayesian optimization for calibrating and selecting hybrid-density functional models[J]. The Journal of Physical Chemistry A, 2020, 124(20): 4053-4061. doi: 10.1021/acs.jpca.0c01375 [13] 蔡文杰. 基于机器学习的广告推荐技术研究[D]. 南京: 南京邮电大学, 2021.CAI W J. Research on advertising recommendation technology based on machine learning[D]. Nanjing: Nanjing University of Posts and Telecommunications, 2021(in Chinese). [14] 中国国际货运航空有限公司. 中国国际货运航空有限公司机型手册[EB/OL]. (2015-04-13)[2023-07-21]. http://max.book118.com/html/2017/0701/119584966.shtm.Air China Cargo Co. , Ltd. Aircraft type manual of air China cargo Co., Ltd[EB/OL]. (2015-04-13)[2023-07-21]. http://max.book118.com/html/2017/0701/119584966.shtm(in Chinese). [15] 熊枭, 陈达, 韩宙, 等. 基于传感器十字交叉式排布的飞机货舱早期火源定位研究[J]. 消防科学与技术, 2022, 41(12): 1632-1636. doi: 10.3969/j.issn.1009-0029.2022.12.004XIONG X, CHEN D, HAN Z, et al. Early fire source location in aircraft cargo hold based on crossover arrangement of sensors[J]. Fire Science and Technology, 2022, 41(12): 1632-1636(in Chinese). doi: 10.3969/j.issn.1009-0029.2022.12.004 [16] Federal Government. Code of federal regulations: CFR Part 25(Title 14)-2022[S]. Washington, D. C, : Government Printing Office, 2022: 858. [17] QU C, HOUSTON P L, YU Q, et al. Machine learning classification can significantly reduce the cost of calculating the Hamiltonian matrix in CI calculations[J]. Journal of Chemical Physics, 2023, 159(7): 071101. doi: 10.1063/5.0168590 -
下载:
点击查看大图
计量
- 文章访问数: 265
- HTML全文浏览量: 75
- PDF下载量: 11
- 被引次数: 0
