Research On Fatigue Crack Localization Method For Operating Rotors Based On Dynamic Response

Fatigue cracks are the most dangerous faults in rotor systems,are highly hidden,and can lead to sudden and destructive accidents if not detected and dealt with in time.Considering the actual needs of rotor system crack diagnosis,the ability to quickly locate rotor cracks without disassembling or interrupting operation can have significant implications for rotor system maintenance decisions.This study centers on the dynamic modeling and verification of fatigue cracks in a rotor system,as well as to develop a methodology for detecting such cracks.The research was carried out using a single-span double-disc rotor-bearing system as the subject of investigation.Firstly,a dynamic model of a rotor-bearing system without cracks was established using the finite element method and Timoshenko beam elements.The stiffness matrix for crack elements was obtained using the principles of linear fracture mechanics and the strain energy release rate.A dynamic model of a rotor-bearing system was developed,which accounted for both cracked and crack-free units.The zero stress intensity factor method was employed to simulate the breathing behavior of the rotor cracks.Verification of the dynamic model was performed through static and dynamic means,guaranteeing the accuracy of both the modeling method and the solution.This established model is suitable for exploring methods of locating cracks.In the static verification process,in order to ensure the accuracy of the finite element method modeling and to adjust the model parameters,both ANSYS simulation and modal experiments were utilized for verification.The natural frequency of the rotor system and the bearing parameters were mapped through the use of ANSYS finite element analysis software,which utilized both the BP neural network surrogate model and the GA-BP neural network surrogate model.In addition,the rotor system’s measured geometric parameters were also taken into account during this process.Using modal experiments,the natural frequency of the first four bending of the rotor system under test is measured,and using the equivalent stiffness of the bearing as input,two methods were employed to identify the equivalent stiffness of the bearings.The results indicated that both methods were effective,however,the GA-BP neural network method had better recognition effect than the traditional BP neural network method,and its maximum error is 1.52%.Finally,the equivalent stiffness of the rotor system bearing under test was identified as9.43045×10~5 N/m.In the dynamic verification process,the dynamic model of the crack-bearing system with cracks was verified by using three-dimensional contact simulation of fatigue cracks and ANSYS transient response analysis.Model healthy rotor systems using ANSYS APDL command flow parametric programmable modeling.By simulating the shaft part of BEAM188 unit,the bearing part of COMBIN214 unit and the rotating table part of MASS21 unit,the healthy rotor-bearing system model was established.A SOLID95 cell with a center node is used to simulate respiratory cracks by establishing a frictionless contact pair on the cell,and a bonded connection between the BEAM188unit and the SOLID95 unit is established using the MPC contact algorithm.ANSYS transient solution was used to calculate the dynamic response of both the healthy and cracked rotors,and the experimental findings were compared with the corresponding analytical predictions obtained from the theoretical model using the finite element method.The findings indicate that the response obtained from ANSYS simulation is in agreement with the computational response generated by the finite element method,thereby validating the accuracy of the theoretical model of the rotor system containing cracks.Furthermore,an analysis was conducted on the vibration response characteristics of the rotor system with cracks.The results indicate that the primary characteristics of a cracked rotor are its superharmonic characteristics,and the lateral vibration of the rotor exhibits prominent 1X,2X,3X,and other high frequency doubling components.Therefore,the super-harmonic composition of the rotor system can be used to diagnose whether cracks appear in the rotor and achieve crack positioning.Secondly,using the superharmonic characteristics of the rotor containing cracks,a crack localization method based on higher-order kinetic mode decomposition(HODMD)is proposed.An improved HODMD method is introduced,which combines the total least squares method with the standard HODMD algorithm,and the order parameters are selected adaptively by optimizing the superharmonic frequency component vector.On this basis,the superharmonic feature components of multiple measurement points are extracted synchronously,and the damage index based on superharmonic transmission rate and fractal dimension is obtained.Based on numerical simulations and experimental analysis,the proposed method was studied and validated,demonstrating its capability to accurately locate multiple cracks in the rotor while mitigating the effects of stiffness reduction due to shaft steps.Finally,combined with the CNN-LSTM deep convolutional neural network,the full-spectrum characteristics of the rotor of dual-point respiratory cracks are combined to transform the localization problem of cracks into target classification problems,and realize the intelligent positioning of rotor cracks.Compared with classical neural networks such as GA-BP,RBF,CNN,etc.,the results show that the full-spectrum features of crack rotor perform well in 1D-CNN-LSTM,and the recognition accuracy of classification is as high as 94.93%,which can realize the reliable positioning of cracks.In conclusion,this paper focuses on the modeling and verification of fatigue crack rotor-bearing system,the improved HODMD rotor crack positioning method,and the rotor crack intelligent positioning method based on the 1D-CNN-LSTM deep convolutional network,establishes an accurate crack rotor dynamics model,and proposes a reliable rotor crack positioning method,which can be applied to the status monitoring of cracks in rotating machinery,and has theoretical and engineering application value.