基于动态TDFP模型与Wasserstein距离的滚动轴承运行状态评估

Operation condition evaluation of rolling bearing based on dynamic TDFP model and Wasserstein distance

针对滚动轴承运行状态评估问题,提出一种基于动态二维特征概率(TDFP)模型与Wasserstein距离的评估方法。首先通过滚动轴承健康数据样本的均方根和线性峭度指标构建二维性能指标序列,利用多元核密度估计构建TDFP基准模型;然后通过不断累积的监测数据样本对TDFP模型进行动态更新,并计算Wasserstein距离测度获取基准模型与更新模型之间的量化差异;最后结合自适应健康阈值策略实现对滚动轴承运行状态的动态评估。滚动轴承全寿命周期实验数据验证结果表明:该评估方法能够准确跟踪轴承的退化过程,并及时捕捉到轴承初始退化时刻,对实际工程应用具有一定参考及借鉴价值。

Aiming at the problem of operation condition evaluation for rolling bearings,an evaluation method based on dynamic two dimension feature probability(TDFP)model and Wasserstein distance is proposed.Firstly,a two dimension performance index sequence is constructed by the root mean square and the linear kurtosis of rolling bearing health data samples,then the TDFP benchmark model is constructed by multivariate kernel density estimation.Furthermore,the TDFP model is dynamically updated by the accumulated monitoring data samples,and the quantitative difference between the benchmark model and the updated model is obtained through the Wasserstein distance measure.Finally,the operation condition dynamic evaluation of rolling bearing is realized by combining the proposed adaptive health threshold strategy.The verification results of rolling bearing whole life cycle experimental data show that,the proposed method can accurately track the degradation process and timely capture the initial degradation time of bearing,and it has a certain reference value for practical engineering application.