中国安全科学学报 ›› 2023, Vol. 33 ›› Issue (10): 16-22.doi: 10.16265/j.cnki.issn1003-3033.2023.10.1744

• 安全社会科学与安全管理 • 上一篇    下一篇

基于深度学习的检修作业过程风险智能预警

王金江1(), 关鹏婷1, 陈卓1, 葛伟凤2,3, 鞠茜4   

  1. 1 中国石油大学(北京) 安全与海洋工程学院,北京 102249
    2 中海油安全技术服务有限公司,天津 300452
    3 中海油能源发展股份有限公司 安全环保分公司,天津 300452
    4 中国石油西南油气田公司 华油公司,重庆 401120
  • 收稿日期:2023-04-11 修回日期:2023-07-13 出版日期:2023-10-28
  • 作者简介:

    王金江 (1981—),男,山东寿光人,博士,教授,主要从事设备安全监测与智能诊断、安全大数据与人工智能等方面的研究。E-mail:

    葛伟凤 工程师

    鞠茜 工程师

  • 基金资助:
    国家重点研发计划项目(2020YFB1709702); 国家自然科学基金重点项目(52234007)

Intelligent warning of risk during maintenance operations based on deep learning

WANG Jinjiang1(), GUAN Pengting1, CHEN Zhuo1, GE Weifeng2,3, JU Qian4   

  1. 1 School of Safety and Ocean Engineering, China University of Petroleum-Beijing, Beijing 102249, China
    2 CNOOC Safety Technology Service Co., Ltd., Tianjin 300452, China
    3 Safety & Environmental Protection Branch, CNOOC EnerTech Co., Ltd.,Tianjin 300452, China
    4 CNPC Petro China Southwest Oil & Gas Field Company HuayouCompany, Chongqing 401120, China
  • Received:2023-04-11 Revised:2023-07-13 Published:2023-10-28

摘要:

为提高炼化企业检维修作业过程的安全性,利用目标检测技术,构建基于深度学习的智能风险识别模型;首先,结合贝叶斯网络(BN)及模糊集理论,建立动态风险评估模型;其次,采用分级动态预警方法,实现检修作业过程风险要素的智能监控预警;然后,以压缩机检修作业过程为例,采用失效模式和影响分析(FMEA)与作业条件危险性分析方法,辨识作业过程风险,再利用基于深度学习的风险智能识别模型,监控现场作业过程并识别风险;最后,采用BN和模糊集理论相结合的方法,评估作业过程中的动态风险,并采用分级预警的方法智能预警作业过程风险。结果表明:该方法可以有效识别压缩机检修作业过程中的风险因素,识别准确率为93%,在此基础上进行动态风险评估,当观测到事件概率发生变化,且超出相应预警阈值时,依据风险等级进行报警,实现风险要素的智能监控。

关键词: 深度学习, 检修作业, 过程风险, 智能预警, 贝叶斯网络(BN)

Abstract:

In order to improve the safety of the inspection and maintenance operation process in refining and chemical enterprises, a deep learning-based intelligent risk identification model was constructed by using the target detection technology. Firstly, a dynamic risk assessment model was established by combining the BN and fuzzy set theory. Secondly, a hierarchical dynamic early warning method was adopted to realize the intelligent monitoring and warning of the risk elements of the overhauling operation process; then the compressor overhauling operation process was taken as an example to identify the operation process risks by means of Failure Mode and Effect Analysis (FMEA) and hazard analysis of operation conditions, and then the intelligent identification model of risk based on deep learning was used for on-site monitoring of the operation process and identification of risk. Finally, the dynamic risk assessment of the operation process was carried out by using the method of combining the BN and the fuzzy set theory, and the grading warning method was used to realize intelligent monitoring and early warning of the risk of the operation process. The results show that the method effectively identify the risk factors in the compressor overhauling operation process, with an identification accuracy rate of 93%. Based on this, a dynamic risk assessment is carried out. When the probability of the event changed and exceeded the corresponding warning thresholds, alarms are issued based on the risk level to realize the intelligent monitoring of the risk elements.

Key words: deep learning, inspection operations, process risk, intelligent warning, Bayesian network(BN)