• Home
  • About CICEE
  • Editorial Board
  • Publication Ethics
  • 中文

    • Investigation on High-Pressure Methane Jet Turbulent Combustion Flame Image Processing Method Based on Deep Learning
    DOI：10.13949/j.cnki.nrjgc.2022.04.003
    Key Words:deep learning  image segmentation  methane  high-pressure jet  turbulent flame
    Author NameAffiliationE-mail
    SHI Lei* State Key Laboratory of Engine Reliability， Weichai Power Co.， Ltd.， Weifang 261061， China shilei02@weichai.com 
    LEI Yan* Department of Automotive Engineering， Beijing University of Technology， Beijing 100124， China leiyan@bjut.edu.cn 
    LIANG Xiaojie Department of Automotive Engineering， Beijing University of Technology， Beijing 100124， China  
    LIN Chen Department of Automotive Engineering， Beijing University of Technology， Beijing 100124， China  
    WANG Lei State Key Laboratory of Engine Reliability， Weichai Power Co.， Ltd.， Weifang 261061， China  
    WANG Xiaoyan State Key Laboratory of Engine Reliability， Weichai Power Co.， Ltd.， Weifang 261061， China  
    ZHOU Dingwu Hunan Automotive Engineering Vocational College， Changsha 412001， China  
    Hits: 1675
    Download times: 1121
    Abstract:Based on a constant volume combustion chamber（CVC）， an optical experiments of high-pressure nature gas （methane） jet combustion was carried out， and the flame image processing was completed by the deep learning method and the edge detection algorithm， respectively. The comparative results showed that because there were both gas jet and flame in these images which had many targets with great difference， the edge detection algorithm could not clearly identify the gas jet and the flame， so it is suitable for single target image processing. The deep learning method could gain the flame contour， and further effectively obtain the macroscopic property of the jet turbulent flame such as the jet turbulent flame propagation distance and flame propagation velocity， so it is suitable for multi-target image processing. The results based on deep learning image processing method show that when the high-pressure methane jet meets the premixed fireball， the flame speed rapidly increases from the initial laminar speed （<3 m/s） up to greater speed with the maximum flame speed of more than 300 m/s， which forms the turbulent flame. The turbulent flame penetrates forward along the jet direction and the area of the jet flame increases. The maximum flame speed linearly decreases as the interval between the jet and the spark ignition.
    View Full Text  View/Add Comment  Download reader