Research On Vibration Signal Analysis And Fault Diagnosis Of Rotating Machinery

With the rising level of industrialization,the complexity of rotating machinery is further deepened,and the level of automation is constantly improved.When rotating machinery failure can not be found in time,it often causes huge economic losses.When the fault develops seriously,it will even cause huge safety hazards to the factory.In order to ensure the stable and reliable operation of rotating machinery,the research on vibration signal analysis and fault diagnosis of rotating machinery is of great practical significance to ensure the safe operation of plant equipment.Firstly,this paper analyzes the vibration mechanism and basic characteristics of rotating machinery.Four vibration forms of rotating machinery are introduced,and the basic vibration characteristics of rotating machinery are analyzed by taking the unit disk rotor model as an example.Then,four common fault mechanisms and vibration characteristics of rotating machinery,such as rotor unbalance,rotor misalignment,dynamic and static rubbing,oil whirl,are analyzed.Secondly,based on the specific analysis of the common faults of rotating machinery,this paper selects Lab VIEW software as the development platform,combined with the relevant hardware to design the vibration signal analysis system.The system is mainly designed for system login page,data acquisition,signal processing and analysis,and fault feature extraction module based on wavelet packet.The vibration signal analysis system collects the vibration signal of rotating machinery in industrial field through 4370 acceleration sensor,then amplifies the collected signal through 2635 charge amplifier,and then stores the vibration signal in the computer through sound card,and then reads,filters,analyzes the time domain and frequency spectrum of the signal through Lab VIEW.Wavelet packet transform is used to decompose and reconstruct the fault vibration signal of rotating machinery,and the frequency band energy value obtained is used as the corresponding fault feature vector,which provides fault data samples for subsequent state recognition.Finally,the BP neural network is used to identify the fault state.Aiming at the shortcomings of the traditional BP neural network random initialization weights and thresholds,and easy to cause the network to fall into local minimum,slow convergence and other problems,a fault state identification method based on GA-BP neural network is proposed,Firstly,the global search ability of genetic algorithm is used to obtain a group of optimal weights and thresholds,and then they are sent to BP neural network for training.The experimental results show that GA-BP neural network has more advantages than single BP neural network in the iterations of network training and the accuracy of fault recognition.