摘要
针对现有阶段划分方法存在退化信息利用不充分且效率较低的问题,本文提出了基于一种新型健康指标的轴承退化阶段自适应划分方法。首先,采用高斯混合模型估计原始振动信号数据的概率分布,依据轴承不同健康状态下混合高斯分布之间的方向与距离差异构建健康指标(hedth indicaror,HI),可在无标签情形下准确获得反映轴承健康状态变化的HI。其次,利用切比雪夫不等式算法识别轴承的状态变化点并去除异常值,结合Transformer网络实现自适应在线阶段划分,并通过Softmax层输出分类概率,能够更精确地反映不同退化阶段的信号分布状态,对于机器学习模型在预测过程中实现可靠决策具有重要作用。PHM 2012轴承数据集上的实验表明:所提方法克服了现有HI对初期缺陷不敏感且鲁棒性差的问题,更加适应轴承的多阶段识别。
To address the insufficient utilization issue of degradation information and low efficiency in the existing stage division methods,an adaptive stage division method of bearing degradation based on a new health index(HI)is introduced in this paper.First,the Gaussian mixture model is applied to estimate the probability distribution of the original vibration signal data.An HI is then constructed based on the direction and distance differences between the mixed Gaussian distribution in different health states of the bearing.This HI,which reflects changes in the health state of the bearing,can be accurately obtained without the need for labels.Next,the Chebyshev inequality algorithm is used to identify the state change point of the bearing and remove the outliers.The adaptive online stage division is achieved by combining this method with a Transformer network,and the classification probability is output through the SoftMax layer,allowing for an accurate reflection of the signal distribution across different degradation stages.This method plays an important role in realizing reliable decision-making in the machine learning prediction process.Experiments conducted on PHM 2012 bearing datasets show that the proposed method overcomes the problems of insensitivity to initial defects and poor robustness in existing HIs,thereby increasing its suitability for multistage bearing recognition.
作者
陈东楠
胡昌华
郑建飞
郑红倩
裴洪
CHEN Dongnan;HU Changhua;ZHENG Jianfei;ZHENG Hongqian;PEI Hong(School of Missile Engineering,Rocket Force University of Engineering,Xi′an 710025,China;Library,Rocket Force University of Engineering,Xi′an 710025,China)
出处
《哈尔滨工程大学学报》
北大核心
2025年第9期1816-1828,共13页
Journal of Harbin Engineering University
基金
国家自然科学基金项目(62227814,62103433,62101579,62373369,62203462)
陕西省科协青年人才托举计划项目(20230127)
中国博士后科学基金面上项目(2023M734286).
