基于广义S变换与隐马尔可夫的滚动轴承性能评估

Performance Evaluation of Rolling Bearings Based on Generalized S-Transform and Hidden Markov

考虑到滚动轴承故障信号的非平稳性、强噪声性,导致状态评估结果不确定性高,提出一种基于广义S变换特征提取和变分贝叶斯-隐马尔可夫模型的滚动轴承性能评估方法。针对滚动轴承监测获得的振动信号,对其进行广义S变换后,分别进行时间、频率、时频的信息熵特征值运算,提取健康指数作为性能评估的特征向量,并使用变分贝叶斯-隐马尔可夫模型建立实时性能评估模型,用健康样本训练模型,以模型输出对数似然概率值作为性能退化的评估指标。利用数学模型仿真和辛辛那提大学提供的轴承数据验证特征指标和评估模型的可行性,结果表明广义S变换熵值优于常规的特征指标,在轴承早期微弱故障时灵敏度高,性能评估模型仅需要正常数据就可以准确表征轴承性能退化趋势,为设备的维修和故障检测提供了参考。

Considering the non-smoothness and strong noise of the rolling bearing fault signal,the condition evaluation results are highly uncertain a rolling bearing performance evaluation method based on generalized S-transform feature extraction and variational bayesian-hidden markov model is proposed.After the generalized S-transformation of the collected rolling bearing vibration signal,the information entropy eigenvalue operations of time,frequency and time-frequency are calculated respectively.Extracting health indices as feature vectors for performance evaluation and using a variational bayesian-markov model to build a real-time performance evaluation model.The model is trained with healthy samples,and the log-likelihood probability value output of the model is used as an evaluation metric for performance degradation.Mathematical model simulations and bearing data provided by the university of Cincinnati are used to validate the characteristic index and assess the feasibility of the model.The results show that the generalized S-transform entropy value is better than the conventional characteristic index,with high sensitivity in case of early weak bearing fault,and the performance evaluation model requires only normal data to accurately characterize the bearing performance degradation trend,which provides the reference for equipment maintenance and fault detection.