融合动力学模型的滚动轴承性能退化评估方法OA

Aiming to address the issue that single health indicators are insufficient to accurately reflect the real-time health status of rolling bearings in online scenarios,a performance degradation assessment method for rolling bearings based on a dynamic model is proposed.Built on the existing bearing dynamic model,the influence of sliding friction forces induced by speed variations on vibration response is considered,and a dynamic model capable of characterizing the health state of rolling bearings at different speeds is constructed.A model updating method based on convolutional neural networks is designed to solve the problem of online synchronous updating of dynamic models.By integrating the online synchronous dynamic model,the contact force characteristics of the dynamic model are analyzed to construct a sensitive index for characterizing the health status of the rolling bearing.Furthermore,an adaptive health status assessment strategy is established based on the dynamic correction threshold of the similarity between the measured and simulated signal.This strategy enables autonomous division of the degradation degree of the rolling bearing and dynamic update of the health index threshold.Finally,the validity of the method is verified using rolling bearing vibration experimental datasets from Xi'an Jiaotong University(XJTU-SY),the National Aeronautics and Space Administration(NASA),and a laboratory self-built platform.The experimental results demonstrate that,compared with the mainstream methods such as multivariate multi-scale fuzzy entropy and feature similarity ratio,the proposed method improves accuracy by 1.67%,reduces memory usage by 30.4%,and decreases floating-point computations by 4.55%.