Study On Intelligent Fault Diagnosis And RUL Prediction Of Fan Bearing

The fan is widely used and is the key equipment in the modern production process.Due to the high speed and heavy load of the fan,it will be dangerous to the surrounding equipment and workers if the fan fails.As an important part of large rotating machine,the main bearing is called "joint of industry".Its health state is closely related to the operation of mechanical equipment.Monitoring the health status of bearings is the optimal strategy to avoid loss of rotating equipment and reduce maintenance costs.Research on bearing fault diagnosis and RUL prediction can reduce the hidden trouble caused by equipment fault effectively.Repairers can judge the failure and severity of machinery equipment through online monitoring,and take corresponding measures to solve the problem in advance so as to minimize the risk cost.The study of bearing degradation mechanism and life prediction has important theoretical and practical significance.In this paper,a method based on wavelet frequency band subdivision and multi-level clustering is proposed to solve the problems of low accuracy,weak generalization ability.Firstly,the bearing data is preprocessed,and then the timefrequency features of bearing data are extracted to form the feature parameter set.The dimension of data is reduced by principal component analysis.The experimental results show that the accuracy of the system is up to 99.3%,and the system has strong generalization ability and robustness.The main innovation points are to improve the wavelet packet method and enhance the ability to segment the signal frequency;Multi-level clustering algorithm is proposed to improve the ability of clustering algorithm to deal with the amount of data.A prediction model based on the fusion network of res-DD and GRU is proposed to solve the problems of weak feature extraction ability and large prediction error in bearing RUL prediction.First,the bearing monitoring data is imported,and the bearing signal is extracted in time and frequency domain.Then,the parameter set is imported into the neural network model for training,and the optimal network model and parameters are obtained.The experimental results show that the system has high accuracy and robustness.The main innovation is to introduce the DD network into the RUL prediction network model and improve it,which improves the prediction ability of the model.Finally,the system interface is built based on the matlab environment,which makes the program modular and visual,and it reduces the difficulty of operation effectively,expands the scope of application,and has practical significance.