基于深度神经网络和趋势特征向量的舱室火灾蔓延预测  

Prediction of cabin fire spread based on deep neural networks and trend feature vectors

作　　者：李捷 黄文斯 Li Jie;Huang Wensi(Shipbuilding Technology Research Lnstitute,Shanghai 200030,China;Wuhan Institute of Marine Electric Propulsion,Wuhan 430064,China)

机构地区：[1]上海船舶工艺研究所,上海市200030 [2]武汉船用电力推进装置研究所,武汉430064

出　　处：《船电技术》2024年第6期73-78,共6页Marine Electric & Electronic Engineering

摘　　要：为了增强CFAST模型的预测精度和应对模型误差,本文提出一种基于LSTM的深度神经网络模型,结合趋势特征向量算法,预测火灾烟气流动的趋势和温度分布。该模型考虑了舱室火灾蔓延度和火灾扩散度等指标,通过提取CFAST预测的火灾时间序列中的趋势特征,并通过LSTM网络的迭代训练过程,校准CFAST的预测偏差,实现对火灾烟气流动和温度分布的高效预测。最终基于船舱火灾典型案例的仿真试验,验证了本文所设计优化方法对舱室火灾蔓延预测的性能。In order to enhance the prediction accuracy of the CFAST model and deal with model errors,this paper proposes a deep neural network model based on LSTM,combined with the trend feature vector algorithm,to predict the trend and temperature distribution of fire smoke flow.The model takes into account indicators such as cabin fire spread and fire diffusion,extracts trend features in the fire time series predicted by CFAST,and calibrates the prediction deviation of CFAST through the iterative training process of the LSTM network to achieve fire smoke flow and efficient prediction of temperature distribution.Finally,the simulation test based on a typical case of cabin fire verifies the performance of the optimization method designed in this article in predicting cabin fire spread.

关 键 词：火灾预测 趋势特征 神经网络 烟气流动 

分 类 号：U647[交通运输工程—船舶及航道工程]