Data-driven Remaining Useful Life Prediction For Rolling Bearings

Rolling bearings are important components of large equipment such as high-speed trains and engines.After prolonged use under extreme working conditions,bearings often get damaged,which affects the normal operation of the equipment.Accurately predicting the remaining useful life of rolling bearings is beneficial for formulating effective maintenance plans,ensuring the safety and stability of the equipment,and preventing accidents.With the development of high-tech technologies such as the Industrial Internet of Things,big data,and cloud computing,data-driven prediction of the remaining useful life of rolling bearings has significant research significance and practical value,attracting the attention of numerous scholars.The main work of this article includes:On the one hand,in response to the problem of low prediction accuracy of existing methods for predicting the remaining useful life of rolling bearings under single working conditions,a prediction method based on the fusion of relevance vector machine and particle filtering is proposed.Health indicator construction is a common and effective method for accurately describing the degradation of rolling bearings.Firstly,based on the root mean square(RMS)of the bearing vibration signal in the time domain,the relevance vector machine is used for life prediction.To further improve the prediction accuracy,the degradation trajectory of the relevance vector machine is used to construct the state transition equation of the particle filter,thereby predicting the bearing degradation trajectory and calculating the remaining useful life.Since the parameters of the kernel function in the relevance vector machine have a significant impact on the prediction performance,the sparrow search algorithm is used to optimize them.Finally,simulations are performed on the PHM2012 rolling bearing dataset and the XJTU-SY rolling bearing dataset.By comparing with other prediction methods,it is demonstrated that the proposed method in this paper has a higher prediction accuracy.On the other hand,in response to the inability of single working condition prediction methods to achieve accurate predictions for multiple working conditions,this paper proposes a remaining useful life prediction method based on adaptive continuous deep belief networks and kernel extreme learning machine.Firstly,a rolling bearing health indicator extraction method based on adaptive continuous deep belief network is proposed.By designing an adaptive learning rate adjustment mechanism,stable feature extraction is achieved,and a health indicator is constructed based on the distance between the initial operating state of feature extraction and real-time degradation state.Based on this,a remaining useful life prediction model for rolling bearings based on kernel extreme learning machine is studied.The sparrow search algorithm is used to optimize the kernel function parameter and regularization coefficient of the kernel extreme learning machine to improve their prediction accuracy.Finally,the proposed method is tested on the PHM2012 rolling bearing dataset and the XJTU-SY bearing dataset,and compared with existing methods,confirming that the proposed method achieves high accuracy in predicting the remaining useful life of bearings under multiple working conditions.