样本不均衡条件下设备健康度评估方法

Method for evaluating equipment health under unbalanced conditions

摘要:

为了保障设备稳定可靠运行,减少设备故障,针对大多数设备采集样本不均衡的现状,提出了利用动态权重的支持向量数据描述(SVDD)方法对设备健康度进行评估。该方法首先利用设备正常健康数据进行SVDD单类学习；然后利用少量各种健康状态的数据样本计算SVDD模型超球面距离,结合其评估的健康度,使用二项式回归算法得到健康度拟合曲线,实现了健康度准确评估模型。在计算过程中,所有样本进行了指数变权的动态权重处理以提高准确性。最后以某型雷达发射机为例进行了测试验证。结果表明,该方法可实现设备健康状态准确评估,具有不错的实用价值。

Abstract:

In order to ensure stable and reliable operation of equipment and reduce equipment failure, a method based on Support Vector Data Description (SVDD) is proposed to evaluate equipment health for the current situation of unbalanced equipment samples. The method firstly uses the normal health data of the device to conduct SVDD single class learning. Then using a small number of data samples of various health states to calculate the SVDD model hypersphere distance, combined with the health degree of its evaluation, using the binomial regression algorithm to obtain the fitness fitting curve, the health degree accurate evaluation model is realized. In the calculation process, all samples are added the dynamic weight processing of exponential variational weight to improve the accuracy. Finally, the test is verified by taking a radar transmitter as an example. The experimental results show that the method has good practical value for accurate assessment of equipment health status.