摘要
针对船舶综合监控系统中存储有海量的设备正常运行时的数据没有得到充分利用,此外设备退化时的故障数据难以获取,无法训练传统的多分类退化检测模型,提出利用单分类算法OSVM来建立模型,从而实现退化检测,在该过程中只需用正常样本数据来训练模型,并在一个经过实船数据验证过的模拟器产生的数据集上进行了试验。结果显示,只需要400个正常样本就可训练出准确的退化检测模型,该模型在精确度、召回率、特异性、正确率、AUC这5个指标都有很好表现。此外,该退化检测模型有很好的扩展性,也可用于其他机械设备的状态评估。
In view of the large amount of data stored in the integrated monitoring and control system of ships in normal operation is not fully utilized, in addition, it is difficult to obtain the fault data when the equipment is decayed, and it is unable to train the traditional multi classification decay detection model, a one-class algorithm OSVM is proposed to build the model, so as to realize the decay detection. During the process, only normal sample data is used to train the model, and the experiment is carried out on a data set generated by a simulator which has been verified by real ship data. The results show that only 400 normal samples are needed to train an accurate decay detection model. The model has good performance in five indexes such as precision, recall, specificity, accuracy and AUC. In addition, the decay detection model has good expansibility and can also be used in the condition evaluation of other mechanical equipment.
作者
田慧
林叶锦
张均东
TIAN Hui;LIN Yejin;ZHANG Jundong(College of Marine Engineering,Dalian Maritime University,Liaoning Dalian 116026,China)
出处
《船舶工程》
CSCD
北大核心
2020年第7期152-156,共5页
Ship Engineering
基金
工信部高技术船舶科研资助项目(工信部装函[2018]473号)。
关键词
燃气轮机
OSVM算法
单分类
状态评估
退化检测
gas turbine
OSVM algorithm
one-class classification
condition evaluation
decay detection
