Life Cycle Prediction System For Large Induced Draft Fan Bearings

The bearings of the induced draft fans play the role of bearing,reducing the friction between the relative moving parts,supporting the rotation and improving the service life of the induced draft fans,and once the bearings are destroyed,the fans will be shut down and affect the whole power generation system.In this thesis,two induced draft fans in operation in a thermal power plant are used as the research background,and the bearings of the induced draft fans are used as the research object.The embedded technology and data-driven technology are combined to build a bearing life prediction model using long and short-term memory neural network to develop a complete fan bearing remaining life prediction system,including the life prediction of fan bearings and real-time bearing operation status monitoring.The structure and installation of the induced draft fan bearing are explained and the causes of fan bearing failure and the types of bearing failure are summarized according to the working environment and working characteristics of the fan bearing,and the common vibration signal time-frequency domain analysis methods are introduced.The vibration signal collected in the field will hav e noise due to the resolution of the sensor,industrial frequency interference and other factors,which will affect the accuracy of the analysis results,so the noise reduction process is needed first.The vibration signal is decomposed by using variationa l modal decomposition,and the components with higher correlation with the original vibration signal are selected for signal reconstruction to achieve the purpose of noise reduction,and the signal-to-noise ratio is calculated to measure the noise reductio n effect.The time domain and frequency domain feature values are extracted from the noise-reduced signal,and the feature values that are sensitive to the changes of the vibration signal are selected as the input values of the life prediction model.The model building for life prediction is primary,and the accuracy of the model directly affects the life prediction results.The proposed life prediction model is validated by simulations using the full-life bearing dataset obtained from the PHM Data Challenge held in 2012,and the remaining life prediction model of the bearing is built using Long Short-Term Memory(LSTM)neural network and the initial prediction of the remaining life of the bearing is performed,followed by the use of the Slime Mould Algorithm(SMA)is used to optimize the hyperparameters of the Long Short-Term Memory neural network,so as to build SMA-LSTM life prediction model for bearing life prediction,and the two models are compared for experiments,and the experimental results show that SMA-LSTM is better than the prediction effect of LSTM.The fan bearing life prediction system is designed,and the prediction system is mainly divided into three aspects: data acquisition,data transmission,and software design.The vibration signal as well as temperature signal of a thermal power plant induced draft fan is connected to the developed life prediction system for testing and analysis,and the test results prove that the system is functional and can display the remaining service life of the fan bearing and the vibration signal as well as temperature signal of the bearing in real time,with real-time,accuracy and reliability,which meets the actual demand.