Research Of Remaining Useful Life Prediction For Fan Slewing Bearing

Large forgings are the core components of major mechanical equipment,and they work continuously under complex working conditions,which can easily lead to increased wear and tear,thus causing failures.To ensure the reliable operation of the equipment and optimize the maintenance strategy,it is necessary to effectively predict the remaining useful life of the equipment.With the continuous progress of the Internet of Things and Sensor technology,a wealth of operating data reflecting equipment degradation can be obtained in real-time through various sensors,promoting the rapid development of equipment reliability analysis based on data-driven method.Large slewing bearing is a new type of large casting and forging developed in recent decades,and it is an important connecting part in wind power generation system,which consists of rolling body and collar with raceway.Usually operating at low speeds and heavy loads,most failures involve raceway failure,which seriously affects the safe operation of wind turbines.Therefore,it has become an important research aspect to research how to construct a degradation model based on the operation characteristics of the slewing bearing of the fan and the monitored degradation information to predict accurately remaining useful life.In this thesis,according to the operation characteristics of large fan slewing bearing,based on the real-time degradation information of multi-sensor monitoring,the remaining useful life prediction model is researched and constructed.The validity of the model is verified by some running data.The specific work includes the following aspects:(1)For the degradation process of fan slewing bearing has nonlinear and non-monotonic characteristics,and the random changes in the external environment(such as tropical high temperature,lightning,snow,wind and sand,random impact,etc.)lead to its accelerated degradation.It is proposed to use the stress effect of complex working conditions on the slewing bearing as a random covariate,which is introduced into the system degradation model in the form of an additive hazard model.A nonlinear Wiener process model for the remaining useful life of the fan slewing bearing is established.A closed expression for the probability density function of the remaining useful life is derived by considering the randomness of system degradation and sample difference.The validity of the proposed model is verified by the monitored vibration data of the fan slewing bearing.(2)For the problem that the traditional monitoring data is susceptible to noise interference,and the limited sample data and single sample data cannot fully characterize the running state of the fan slewing bearing,the multi-sensor data fusion method is researched,and the fused composite health index is used to analyze the system performance degradation state.For the problem of inaccurate prediction results caused by difficulty in obtaining a large amount of physical failure fault data,a hybrid prediction model for the remaining useful life of the fan slewing bearing is proposed based on the physical model method and the data-driven method.The parameters are estimated and updated by Maximum Likelihood Estimation method and Bayesian method.The closed expression of the probability density function of the remaining useful life is derived.The validity of the proposed model is verified by monitoring the surface temperature data of the fan slewing bearing raceway.(3)To extract more effective features representing the degradation state of the system and avoid the problem that the redundancy between different features leads to inaccurate prediction of the remaining useful life,an adaptive weight feature selection method based on multiple evaluation indexes is proposed.In addition,by considering the random correlation between different characteristics and the interaction between different degradation states,a nonlinear state space prediction model of wind turbine slewing bearing is established.The Auxiliary Particle Filter method is used to estimate the state and update the degradation model.The Diagonal Matrix Weighting algorithm is used to realize the distributed fusion of degradation states.The validity of the proposed method is verified by monitoring the vibration data and temperature data of the fan slewing bearing.(4)For the problem of uneven wear distribution affecting system degradation caused by different stresses on the raceways of fan slewing bearings,a nonlinear state space prediction model considering the random influence of different characteristics on different degradation states under the random stress of the raceway is proposed.For the problem that the traditional Auxiliary Particle Filter algorithm has sample degradation and poverty,which leads to inaccurate state estimation results,an adaptive enhanced Auxiliary Particle Filter algorithm is proposed to improve the accuracy of state estimation.The Bayesian super-parameter estimation method is used to estimate the parameters of the random stress model.The proposed model is verified by the monitored vibration data,temperature data,and torque data.