一种基于火灾图像和人工智能的热释放速率计算方法  

A Method for Calculating Heat Release Rates Based on Fire Images and Artificial Intelligence

作　　者：王自龙 张天航 黄鑫炎 Wang Zilong;Zhang Tianhang;Huang Xinyan(Department of Building Environment and Energy Engineering,The Hong Kong Polytechnic University,Hong Kong 900777,China)

机构地区：[1]香港理工大学建筑环境与能源工程学系,中国香港900777

出　　处：《燃烧科学与技术》2024年第1期69-74,共6页Journal of Combustion Science and Technology

基　　金：香港研究资助局资助项目(T22-505/19-N).

摘　　要：火焰的热释放速率是描述火灾行为和危险性的重要参数之一,然而由于火灾场景的复杂性,传统的火灾热释放速率测量方法无法适用.本文旨在探索一种基于人工智能和图像的火灾热释放速率预测方法,用于火灾场景下火灾强度的识别,进而为消防应急救援提供依据.用于火灾热释放速率识别的深度学习模型(Swin Transformer)由美国国家标准技术研究所公开的火灾图像量热数据库训练而成,其中包含火灾测试89例,不同时刻的5万多张火灾图像.所得模型可根据火灾场景下拍摄的火灾图像实时识别火灾的大小.其预测结果表明,即使在不同火源的场景下,深度学习模型可以通过当前的火灾图像对火灾热释放速率进行有效识别,这将为未来的智慧消防系统开发提供有效的支持.The heat release rate(HRR)of a fire is one of the critical parameters in describing the fire behavior and hazard.However,due to the complexity of fire scenarios,traditional methods of measuring HRR cannot be applied.This study aims to explore a HRR prediction method based on artificial intelligence in fire scenarios,which will facilitate the identification of fire intensity and provide support for fire emergency rescue.The deep learning model(Swin Transformer)for HRR calculation is trained on the publicly available fire calorimetric database from the National Institute of Standards and Technology,which contains 89 fire tests and over 50000 fire images taken at different times.The obtained model can identify the size of the fire in real time based on the fire images taken in the fire scenario.The results show that even in different fire scenarios,the deep learning model can effectively identify HRR based on the fire images,which will provide effective support for the development of future intelligent fire systems.

关 键 词：火灾图像 人工智能 火灾量热 Swin Transformer 

分 类 号：TK16[动力工程及工程热物理—热能工程]