摘要
列控车载设备故障排查与维护多依赖于人员经验,存在一定的片面性且效率较低。提出一种基于贝叶斯网络的列控车载设备故障智能诊断方法,基于历史故障数据得到故障征兆,利用粗糙集理论对故障征兆进行属性约简,降低训练模型的复杂度;将专家知识与故障数据训练相结合,改进贝叶斯网络模型,并将故障征兆关联关系融入模型中。以武广线列控故障数据为例,验证该模型的诊断效果。该方法对提升列控系统故障诊断的智能化水平具有借鉴意义。
The conventional method for trouble shooting and maintenance of on-board equipment of train control system, with over-reliance on the personnel experience, is characterized by one-sidedness and low efficiency. This paper proposed a Bayesian network based intelligent fault diagnosis method for on-board equipment. Fault symptoms were derived from the historical fault data by data mining, and then the attribute reduction of the fault symptoms based on Rough Set theory was conducted to reduce the complexity of the training model. Fur-ther, with the combination of expert knowledge and fault data training, the Bayesian network model was im-proved to incorporate the correlation between the fault symptoms into the model. Finally, the effects of this di-agnosis model were analyzed and verified based on the train control fault data of Wuhan-Guangzhou high-speed railway. The method proposed in this paper provides reference for the improvement of the intelligence level of on-board equipment fault diagnosis of train control system.
出处
《铁道学报》
EI
CAS
CSCD
北大核心
2017年第8期93-100,共8页
Journal of the China Railway Society
基金
国家自然科学基金(61473029)
国家重点基础研究发展计划(973计划)(2014CB340703)
中国铁路总公司科技研究开发计划(2014X008-A)
关键词
车载设备
故障诊断
贝叶斯网络
属性约简
故障征兆
on-board equipment
fault diagnosis
Bayesian network
attribute reduction
fault symptom