摘要
针对行车传统运维方式中存在的可管可控状态不明、故障时定位分析耗时久、设备宕机时间长等问题,为实现行车状态运维的智能化转型升级,基于数字孪生技术和数据驱动方法,提出一种用于行车多机构联动、多因素耦合的隐性级联故障诊断分析方法。以数据驱动为核心,构建了数据驱动的行车数字孪生模型,详细阐述了行车数字孪生系统的组成和交互行为;对显性故障和隐性故障进行分类和定义,设计了SDAE—MCSVM—FBN方法解决多因素隐性级联故障;构建了数据驱动的行车数字孪生模型的原型系统,以某大型国企钢厂某车间内的行车作业过程为例,与传统运维方式中的人工点检对比,在所提方法应用前后,故障维修时长和设备宕机时长分别减少的时间比例区间为24.5%~32.8%,20.5%~32.4%,证实了所提方法对隐性级联故障诊断分析的有效性和可行性。
To realize the intelligent transformation and upgrading of the state operation and maintenance of the crane,an implicit cascade fault diagnosis and analysis method was proposed based on digital twin technology and data-driven method for the multi-agency linkage and multi-factor coupling of the crane.With data-driven as the core,a data-driven digital twin model of the crane was constructed,and the composition and interaction behavior of the digital twin system of the crane were elaborated.The explicit and implicit faults were classified and defined,and the SDAE-MCSVM-FBN method was designed to solve multi-factor implicit cascade faults.A prototype system of data-driven digital twin model of the crane was constructed.Taking the train operation process in the workshop of a large state-owned enterprise steel plant as an example,compared with the manual spot check in the traditional operation and maintenance mode,before and after the application of the method in this paper,the proportion interval of time for fault maintenance and equipment downtime was 24.5%~32.8%and 20.5%~32.4%respectively.The validity and feasibility of the proposed method for the diagnosis of hidden cascading faults were verified.
作者
张辉辉
郑龙辉
孙奕程
杨芸
李婕
黄德林
鲍劲松
张丹
ZHANG Huihui;ZHENG Longhui;SUN Yicheng;YANG Yun;LI Jie;HUANG Delin;BAO Jinsong;ZHANG Dan(School of Mechanical Engineering,Donghua University,Shanghai 201620,China;Department of Mechanical Engineering,York University,Toronto M4Y1M7,Canada)
出处
《计算机集成制造系统》
EI
CSCD
北大核心
2023年第6期2086-2101,共16页
Computer Integrated Manufacturing Systems
基金
中央高校基本科研业务费专项资金资助项目(2232022D-20)
上海市自然科学基金资助项目(21ZR1400800)
上海市扬帆计划资助项目(21YF1400500,19YF1401600)。
关键词
数字孪生
数据驱动
堆栈降噪自编码器
多类支持向量机
模糊贝叶斯网络
digital twin
data-driven
stacked denoising auto-encoder
mutil-class support vector machine
fuzzy Bayesian network
